Lecture_Notes.pdf

Fundamentals Of Computer
Networking And Internetworking
Prof. Douglas Comer
Purdue University
http://www.cs.purdue.edu/people/comer
 Copyright 2014 by Douglas Comer And Pearson Education. All rights reserved.

MODULE I
Introductions, Course Overview,
Approaches To Networking, Open And
Closed Systems, Protocols, And Layering
Computer Networks and Internets -- Module 1 1 Spring, 2014
Copyright  2014. All rights reserved.

Introductions
d Professor
d TAs
d Students

Topic And Scope
Computer networks and internets: an overview of concepts,
terminology, and technologies that form the basis for digital
communication in individual networks and the global Internet

You Will Learn
d Fundamental principles
d Concepts
d Terminology (lots of it)
d Key aspects of networking

The Five Key Aspects Of Networking
d Data communications: signals over wires and bits over
signals
d Networks: packets over bits
d Internets: datagrams over packets
d Network programming: application data over the Internet
d Cross-functional concepts and technologies: network
configuration, control, and management

Features Of The Course
d Covers all of networking and internetworking from media to
applications
d Examines each of the underlying technologies
d Focuses on concepts and principles that apply across
vendors and products
d Provides perspective and shows how the pieces fit together
d Explains how an Internet is built from heterogeneous
networks

What You Will Not Learn
d Commercial aspects
– Vendors
– Products
– Prices
– Markets and marketing
d How to engineer network equipment
d How to configure/operate networks
d How to design new protocols

Practice Sessions (Aka Labs)
d Form an important part of the course
d You will
– Build network programs
– Capture and analyze packets
– Learn about protocols

Background Expected
d Our goal is breadth rather than depth
d Only a few basics are needed
– Ability to program in C
– A glancing acquaintance with data structures and
pointers
– A minor brush with algebra
– A basic understanding of operating systems
d The major requirement is a desire to learn

Summary Of The Course
d Explores all aspects of networking and internetworking
d Gives concepts and principles
d Focuses on the big picture
d Includes lots of programming exercises
d Moves rapidly and covers lots of vocabulary

Historic Approaches
To Networking

How Should A Network Be Structured?

d The early phone company answer
– Data networking is like telephone calls
– We will devise and offer various data services
– Charges will depend on distance and duration
– You only need 128 Kbps

d The early phone company answer
– Data networking is like telephone calls
– We will devise and offer various data services
– Charges will depend on distance and duration
– You only need 128 Kbps
d The early computer vendor answer
– A network connects computers in your organization
– We will devise all the necessary equipment and software
– You only need to connect our computers
– You only need to run our applications

(continued)
d The early network equipment vendor answer
– The network is independent of computers
– We will create network equipment and interface
hardware that connects computers to our network
– We will build device drivers for your operating system
– You only need to use our network

Some Resulting Commercial Network Systems
d Apple Computer Appletalk
d Banyan Vines
d Digital Equipment Corporation DECNET
d IBM SNA
d Novell Netware
d Ungermann Bass NET/ One
d Xerox XNS

The Researcher’s Answer
d Although we have computers at multiple sites, we reject the
phone company’s approach
d Because we use diverse computer architectures, we reject
the computer vendors’ approach
d Because a variety of network technologies are possible, we
reject the network vendors’ approach
d A variety of applications are possible
d Let’s experiment with new technologies (short distance as
well as long distance) and new applications

Some Resulting Research Projects
d Xerox Palo Alto Research Center
– Ethernet
d MIT and elsewhere
– Token passing ring networks
d Department of Defense
– ARPANET
– SATNET
– Packet radio net
– The global Internet

Open Vs. Closed Networking
d Closed networks
– Vertical approach
– Each vendor designs/ builds their own
– Given technology owned by vendor
– Vendor may license technology to other groups
d Open networks
– Competitive approach
– Multiple groups collaborate to define a technology
– To insure interoperability, specifications written in
standards documents that are available to everyone
– Companies build products according to standards

Protocol Standards
And Protocol Design

Why Standardize?
d Networking supports communication among multiple
entities
d Agreement needed to make communication correct, efficient,
and meaningful

Which Organizations Issue Standards?
d IEEE (Institute of Electrical and Electronics Engineers)
d IETF (Internet Engineering Task Force)
d ITU (International Telecommunications Union)
d ISO (International Organization for Standardization)
d W3C (World Wide Web Consortium)
d ...and many others

Standards And Standardization
d Joke: why is networking so difficult?

Standards And Standardization
d Joke: why is networking so difficult?
d Because there are so many standards from which to choose

Protocol
d Each protocol specifies how to handle one aspect of
communication
d A protocol can specify
– Low-level details such as voltage and frequency
– High-level details such as format visible to a user
d Many individual communication protocol standards exist
d Set of protocols designed to work together is known as a
suite
– Example: TCP/ IP Internet protocol suite

Two Key Properties That Protocols Specify
d Syntax
– Format of each message
– Representation of data items
– Encoding of bits in electromagnetic signals
d Semantics
– Meaning of each message
– Procedures used to exchange messages
– Actions to take when an error occurs

Steps In Protocol Design
d Look at the facilities the underlying hardware provides

d Imagine an abstract communication mechanism as a user
would like it to work

d Design an efficient implementation of the abstraction

d Design an efficient implementation of the abstraction
d The key to success: choose a good abstraction

Why Protocol Design Is Difficult
d Multiple implementations of a protocol will exist
d Implementations will be created by a multiple
individuals/organizations
d There are many details to consider
d Key tradeoff
– A specification that dictates all possible details restricts
implementations
– A specification that does not specify enough details is
ambiguous and leads to incompatible implementations

Maximizing Interoperability
d Design principle that maximizes interoperability (due to
Postel)
Be conservative in what you send
and be liberal in what you accept.

Protocol Layering
and Layering Models

Protocol Layering
d Needed because communication is complex
d Intended primarily for protocol designers
d Divides communication into intellectually manageable
pieces
d Provides a conceptual framework that can help us
understand protocols
d Ideally, layering is invisible once protocols have been
designed
d Notes:
– Layering gives a guideline, not a rigid framework
– Optimizations may violate strict layering

Two Layering Models
d Internet protocols use a 5-layer reference model
d ISO and the ITU defined a 7-layer model

Internet Reference Model
Application
Transport
Internet
Network Interface
Physical
LAYER 1
LAYER 2
LAYER 3
LAYER 4
LAYER 5
d Descriptive model formed after TCP/IP protocols were
devised
d Used in practice

Physical Layer
d Underlying transmission media
d Electromagnetic energy and its use
d Representation of information in signals
d Electrical properties such as radio frequencies and voltage
d Associated hardware

Network Interface Layer
d Communication between a computer and network hardware
d Also called data link or MAC layer
d Mechanisms for gaining access to shared media
d Hardware (MAC) addressing
d Packet (frame) formats
d Packet (frame) types and demultiplexing
d Error detection

Internet Layer
d Communication between a pair of computers across the
Internet
d Internet packet format (datagram)
d Internet addressing model and address assignment
d Forwarding of Internet packets
d Dividing an Internet packet into smaller packets for
transmission
d Error detection and reporting

Transport Layer
d Communication between a pair of applications
d Demultiplexing among multiple destinations on a computer
d Reliable delivery and retransmission
d Mechanisms to control data rate and avoid congestion

Application Layer
d Format and representation of data and messages
d Procedures applications follow to
– Transfer data
– Handle errors or unexpected conditions
d Meaning of messages exchanged
d Internet infrastructure such as routing and DNS

General Idea
d Each computer contains an entire set of layered protocols
d When an application sends a message
– The message passes down through the layered protocols
– A given layer adds information and forms a packet
– The computer transmits the final packet
d When a packet arrives
– The packet passes up through the protocol layers
– A given layer performs processing and passes the packet
up to the next layer
– The application receives the message that was sent

Illustration Of Protocol Software On A Computer
Application
Transport
Internet
Net. Interface
Application
Transport
Internet
Net. Interface
Physical Network
Computer 1 Computer 2
d Protocols on a computer arranged in a conceptual stack

Packet Headers As A Packet Passes
Across The Internet
d One header prepended by each layer when message sent
d Result: headers are nested with lowest-layer header
appearing first
message the application sent
1: Physical header (possible, but not typical)
2: Network Interface header
3: Internet header
4: Transport header

Layering Principle
d Layered protocols enforce an invariant:
Layer N at the destination receives an exact copy of the
message sent by layer N at the source. All headers and other
modifications added by lower layers at the source must be
removed by lower layers at the destination.
d Allows protocol designer to focus on one layer at a time

Illustration Of The Layering Principle
Application
Transport
Internet
Net. Interface
Application
Transport
Internet
Net. Interface
SOURCE DESTINATION
Physical Network
identical message
identical packet
identical datagram
identical frame

Do We Understand Layering?

Do We Understand Layering?
No!

A Few Subtle Complications Of Layering

d Layering diagrams are abstract and simplistic

d Details and exceptions complicate practical systems

d Details and exceptions complicate practical systems
d Four examples
– Cross-layer communication
– Multiple protocols per layer
– Layering in an Internet
– Technologies that intertwine layers

Example Of Cross-Layer Communication
d Facts
– A transport protocol selects amount of data to send in
each packet
– To optimize performance, ensure packets are full
d Unfortunately
– To find maximum packet size, transport protocol must
interact with a lower layer

Multiple Protocols Per Layer
d Consider a typical computer
d User can run multiple applications simultaneously
– Email
– Web browser
d Computer can connect to multiple physical networks
– Wired Ethernet
– Wi-Fi wireless network
d Other layers have multiple protocols as well

Illustration Of Multiple Protocols At Each Layer
appl1 appl2 appl3
TCP UDP
IPv4 IPv6
wired interface wireless interface
Application
Transport
Internet
Net. Interface
Ethernet Wi-Fi Network
COMPUTER

Layering In An Internet
d Our layering diagrams only show two computers connected
to a network
d The Internet contains multiple networks interconnected by
routers
d Routers only need layer 2 and layer 3 software to forward
packets across the Internet

Illustration Of Layers Used To Forward
Packets Across The Internet
application application
transport transport
Internet Internet
Internet
net interface net interface
net interface
net 1 net 2
Host A Host B
router
d In practice, routers do more than forward packets
d We will learn more later in the course

Technologies That Intertwine Layers
d Cross-layer functions
– Routing protocols operate at layer 5 but change layer 3
forwarding tables
– Address resolution maps layer 3 addresses to layer 2
addresses
d Layer circularities
– Tunneling can be used to send IPv6 (a layer 3 protocol)
over IPv4 (another layer 3 protocol)
– Virtual Private Networks (VPNs) send IP over IP

Illustration Of Layering Used By A VPN
VPN
appl1 appl2
transport
Internet Internet
net interface
Physical net

ISO 7-Layer Reference Model
Application
Presentation
Session
Transport
Network
Data Link
Physical
LAYER 1
LAYER 2
LAYER 3
LAYER 4
LAYER 5
LAYER 6
LAYER 7
d Prescriptive model formed before protocols were devised
d Created by committee vote

ISO 7-Layer Reference Model
(continued)
d Model was defined when data networks connected dumb
terminals to large mainframes
d Session layer
– Handled details of login and control of send/ receive
– Provided opportunity for billing and accounting
d Presentation layer
– Defined data representation
– Primary intention was to map character sets
d Both layers now superfluous

Unfortunately
d Marketing organizations decided seven is better than five
d Many textbooks and vendors claim to use “all seven layers”

Summary
d Network systems can be open or closed
– Closed systems are created and owned by a single
company
– Open systems require that technology be specified in
standards documents that allow multiple companies to
build products
d A protocol standard can specify data and message
representation, rules for message exchange, error handling,
or low-level details such as voltage

Summary
(continued)
d A layering model provides a conceptual framework that
helps protocol designers create a suite of protocols
d Implementation of layered protocols known as a stack
d Internet uses a 5-layer reference model
d Remainder of the course explores each layer

An Alternative To Layering

An Alternative To Layering
Hire really, really smart people and have them design a single,
large protocol that handles all aspects of communication
without dividing the problem into smaller subproblems

A List Of All Practical
Alternatives To Layering
(this page intentionally left blank)

MODULE II
Network Programming And Applications

Topics
d Internet services and communication paradigms
d Client-server model and alternatives
d Network programming with a simplified API
d The socket API
d Application layer protocols
d Examples of standard application protocols

Internet Services And
Communication Paradigms

General Principle: Intelligence At The Edge
The Internet does not provide services. Instead, the
Internet only provides communication, and application
programs provide all services.

General Principle: Intelligence At The Edge
The Internet does not provide services. Instead, the
Internet only provides communication, and application
programs provide all services.
d Consequence
– Every Internet communication, including voice and
video teleconferencing, involves communication among
application programs

Communication Paradigms
d The Internet offers two communication paradigms
2
222222222222222222222222222222222222222222222222222222222222222222
Stream Paradigm Message Paradigm
2
222222222222222222222222222222222222222222222222222222222222222222
Connection-oriented Connectionless
2
222222222222222222222222222222222222222222222222222222222222222222
1-to-1 communication Many-to-many communication
2
222222222222222222222222222222222222222222222222222222222222222222
Sequence of individual bytes Sequence of individual messages
2
222222222222222222222222222222222222222222222222222222222222222222
Arbitrary length transfer Each message limited to 64 Kbytes
2
222222222222222222222222222222222222222222222222222222222222222222
Used by most applications Used for multimedia applications
2
222222222222222222222222222222222222222222222222222222222222222222
Built on TCP protocol Built on UDP protocol
2
222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Each paradigm has surprising characteristics

Stream Paradigm (TCP)
d Transfers a sequence of bytes
d Connection-oriented: data sent between two applications
d Bidirectional (one stream in each direction)
d No meaning attached to data and no boundaries inserted in
data
d Surprising characteristic:
Although it delivers all bytes in sequence, the stream
paradigm does not guarantee that the chunks of bytes
passed to a receiving application correspond to the
chunks of bytes transferred by the sending application.

Message Paradigm (UDP)
d Connectionless: network accepts and delivers individual
messages
d If the sender places N bytes in a message, a receiver will
find exactly N bytes in the incoming message
d Paradigm allows unicast, multicast, or broadcast delivery
(one destination, multiple destinations, or all destinations)
d Surprising characteristic:
Although it preserves boundaries, the message paradigm
allows messages to be lost, duplicated, or delivered out-
of-order; neither the sender nor receiver is informed
when such errors occur.

Stream Transport And Data Chunks
d The protocol system may
– Divide the data from the sender into multiple segments
and deliver a few bytes at a time to the receiver
– Combine data from multiple transmissions into a single
large chunk and deliver it to the receiver all at once
d Consequence: receiving application cannot know exactly
which pieces were sent

Example #1
d Assume a stream connection between two applications
d Sender
– Places 1000-byte message in buffer buf
– Makes a single request to send all 1000 bytes
d Receiver
– Allocates a buffer b with 1000 bytes
– Reads 1000 bytes from the stream into buffer b
d The OS may return between 1 and 1000 bytes
d Application must make repeated calls until all 1000 bytes
have been acquired

Example #2
d Assume a stream connection between two applications
d Sender transmits a sequence of four messages that are each
100 bytes long
d Receiver allocates a large buffer b of 1000 bytes and
requests that up to 1000 bytes from stream be read into
buffer b
d The OS may choose to return all four messages (400 bytes)
with a single read request
d Receiving application must be able to separate received data
into four separate messages

Programming Hints

Programming Hints
d When using the stream paradigm

Programming Hints
– Devise a way that a receiver knows where a message
ends
– Read from a socket until the entire message has been
acquired

Programming Hints
ends
acquired
d When considering using the message paradigm

Programming Hints
ends
acquired
d When considering using the message paradigm
– Don’t (at least not yet)

Identifying Individual Messages In A Stream
d Possibilities
– Send exactly one message followed by end of file (EOF)
– Send multiple messages with an integer length before
each message
– Send multiple messages with a termination character (or
sequence) following each message
d Notes
– Any technique can be used as long as both sides agree
– If sending a multi-byte length value or multi-byte
termination sequence, remember that the application may
need multiple calls receive all bytes

Questions

Questions
d In a realistic setting
– Is division of a message likely to occur?
– Is aggregation of multiple messages likely to occur?

Questions
d Answers

Questions
d Answers yes! (depending on the size of the messages)
– Messages larger than 1400 characters are usually divided
into multiple packets for transmission, and may be
delivered together or separately
– The stream service is designed to aggregate small
messages before making them available to a receiving
application

Buffering In The Stream Paradigm
d Aggregation, which makes bulk transfer more efficient, can
occur on the sending or receiving side
d The stream paradigm includes a push operation that an
application can use to force transmission and delivery
d Unix convention: automatically push for each individual
write call
d Programming hints
– To ensure a small message is transmitted and delivered
without delay, use a separate write
– Even with push, network delays mean applications must
be written to tolerate aggregation
d More details later in the course

Client-Server Model
And Alternatives

Client-Server Model Of Interaction
d Used by applications to establish communication
d One application acts as a server
– Starts execution first
– Awaits contact
d The other application becomes a client
– Starts after server is running
– Initiates contact

Client-Server Model Of Interaction
d Used by applications to establish communication
d One application acts as a server
– Starts execution first
– Awaits contact
d The other application becomes a client
– Starts after server is running
– Initiates contact
d Important concept: once communication has been
established, data (e.g., requests and responses) can flow in
either direction between a client and server

Characteristics Of A Client
d Arbitrary application program that becomes a client
temporarily
d Usually invoked directly by a user, and usually executes
only for one session
d Actively initiates contact with a server, exchanges messages,
and then terminates contact
d Can access multiple services as needed, but usually contacts
one remote server at a time
d Runs locally on a user’s personal computer or smart phone
d Does not require especially powerful computer hardware

Characteristics Of A Server
d Special-purpose, privileged program dedicated to providing
a service
d Usually designed to handle multiple remote clients at the
same time
d Invoked automatically when a system boots, and continues
to execute through many client sessions
d Waits passively for contact from arbitrary remote clients and
then exchanges messages
d Requires powerful hardware and a sophisticated operating
system
d Runs on a large, powerful computer

Characteristics Of A Server
d Special-purpose, privileged program dedicated to providing
a service
d Usually designed to handle multiple remote clients at the
same time — complicates the design
d Invoked automatically when a system boots, and continues
to execute through many client sessions
d Waits passively for contact from arbitrary remote clients and
then exchanges messages
d Requires powerful hardware and a sophisticated operating
system
d Runs on a large, powerful computer

Server Programs And Server-Class Computers
d Confusion exists between scientific and marketing
terminology
d Scientific: a client and a server are each programs
d Marketing: a server is a powerful computer
Internet
connection
client runs
in a standard
computer
server runs in
a server-class
computer

Summary Of Client-Server Interaction
2
222222222222222222222222222222222222222222222222222222222222222222222
Server Application Client Application
2
222222222222222222222222222222222222222222222222222222222222222222222
Starts first Starts second
2
222222222222222222222222222222222222222222222222222222222222222222222
Does not need to know which client Must know which server to
will contact it contact
2
222222222222222222222222222222222222222222222222222222222222222222222
Waits passively and arbitrarily long Initiates a contact whenever
for contact from a client communication is needed
2
222222222222222222222222222222222222222222222222222222222222222222222
Communicates with a client by Communicates with a server by
sending and receiving data sending and receiving data
2
222222222222222222222222222222222222222222222222222222222222222222222
Stays running after servicing one May terminate after interacting
client, and waits for another with a server
2
222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Illustration Of Steps Taken By Client And Server
Internet
Client Side
d Agree a priori on a
port number, N
d Start after server is
already running
d Obtain server name
from user
d Use DNS to translate
name to IP address
d Contact server using IP
address and port N
d Interact with server and
then exit
Server Side
d Agree a priori on a
port number, N
d Start before any of
the clients
d Register port N with
the local system
d Wait for contact
from a client
d Interact with client
until client finishes
d Wait for contact from
the next client...

Alternatives To Client-Server

d Broadcast
– Sender broadcasts message and all stations receive it
– Does not scale well (becomes inefficient)
– Difficult to restrict data access

d Broadcast
– Sender broadcasts message and all stations receive it
– Does not scale well (becomes inefficient)
– Difficult to restrict data access
d Rendezvous point
– Intermediary connects communicating applications
– In essence, there are two clients and a server
– Rendezvous point becomes a bottleneck

(continued)
d Peer-To-Peer Interaction
– Designed to avoid central server bottleneck
– Data divided among N computers
– Each computer acts as a server for its data and as a
client for other data
– Given computer receives 1 / N of the traffic
Internet
1/ N of all traffic

Network Programming With
A Simplified API

Network Programming
d General term that refers to the creation of client and server
applications that communicate over a network
d Programmer uses an Application Program Interface (API)
– Set of functions
– Include control as well as data transfer functions (e.g.,
establish and terminate communication)
d Defined by the operating system; not part of the Internet
standards
d Socket API has become a de facto standard

A Simplified API
d Will help you get started
d General idea
– Server is identified by pair (computer, application)
– Server starts first and waits for contact
– Client specifies server’s location
– Once a connection is established, client and server can
exchange data
d Only seven functions in the simplified API

Our Simplified API
2222222222222222222222222222222222222222222222222222222222222222222222
Operation Meaning
2222222222222222222222222222222222222222222222222222222222222222222222
await_contact Used by a server to wait for contact from a
client
2222222222222222222222222222222222222222222222222222222222222222222222
make_contact Used by a client to contact a server
2222222222222222222222222222222222222222222222222222222222222222222222
appname_to_appnum
Used to translate a program name to an
equivalent internal binary value
2222222222222222222222222222222222222222222222222222222222222222222222
cname_to_comp
Used to translate a computer name to an
equivalent internal binary value
2222222222222222222222222222222222222222222222222222222222222222222222
send Used by either client or server to send data
2222222222222222222222222222222222222222222222222222222222222222222222
recv Used by either client or server to receive data
2222222222222222222222222222222222222222222222222222222222222222222222
send_eof
Used by both client and server after they have
finished sending data
2222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Client And Server Using The API
d Sequence of calls for a trivial exchange in which a client
sends a single request and the server responds
Server
await_contact
recv
send
send_eof
Client
make_contact
send
recv
send_eof
d Both sides must call send_eof because communication is
bidirectional

Data Types For Our Simplified API
22222222222222222222222222222222222222222222222222222222222
Type Name Meaning
22222222222222222222222222222222222222222222222222222222222
appnum A binary value used to identify an application
22222222222222222222222222222222222222222222222222222222222
computer A binary value used to identify a computer
22222222222222222222222222222222222222222222222222222222222
connection A value used to identify the connection
between a client and server
22222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1

An Extra Function For Convenience
d Simplified API includes an extra function, recvln
d Not required, but convenient
d Similar to recv
– Receives data from a connection
– Places data in a buffer
d Difference
– Reads exactly the amount requested
– Technique: repeatedly call recv until specified length has
been acquired

Argument Types For Our API
2
222222222222222222222222222222222222222222222222222222222222222222222222
Function Type Type of Type of Type of
Name Returned arg 1 arg 2 args 3–4
2
222222222222222222222222222222222222222222222222222222222222222222222222
await_contact connection appnum – –
2
222222222222222222222222222222222222222222222222222222222222222222222222
make_contact connection computer appnum –
2
222222222222222222222222222222222222222222222222222222222222222222222222
appname_to_appnum appnum char * – –
2
222222222222222222222222222222222222222222222222222222222222222222222222
cname_to_comp computer char * – –
2
222222222222222222222222222222222222222222222222222222222222222222222222
send int connection char * int
2
222222222222222222222222222222222222222222222222222222222222222222222222
recv int connection char * int
2
222222222222222222222222222222222222222222222222222222222222222222222222
recvln int connection char * int
2
222222222222222222222222222222222222222222222222222222222222222222222222
send_eof int connection – –
2
222222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d You will learn more in the PSOs

Sockets
d Originally part of BSD Unix
d Now standard in the industry
d AT&T defined an alternative named TLI (Transport Layer
Interface), but TLI is now extinct
d Almost every OS includes an implementation
d MS Windows chose to make minor changes (annoying)

Socket Characteristics
d Socket can be used for
– Connectionless communication (UDP message)
– Connection-oriented communication (TCP stream)
d Many functions in the API
d Approach
– Create a socket
– Make many function calls to specify type of
communication, remote computer’s address, port number
to be used, etc.
– Use socket to send / receive data
– Close the socket (terminate use)

Example Socket Calls For Stream Communication
CLIENT SIDE SERVER SIDE
socket
connect
send
recv
close
socket
bind
listen
accept
recv
send
close

Terminology
d Availability of an application protocol
– Closed — vendor defines a protocol for their products
– Open — standardized and available for all vendors
d Basic protocol types
– Data representation — message and data formats
– Data transfer — procedures for exchanging messages
and handling unexpected / error conditions
d Notes
– Application may define separate protocol for each type
– Term Transfer in a protocol title indicates the latter

Defining An Application Layer Protocol
d Programmer specifies representation
– Format of each message and each data item
– Meaning of each item in a message
d Programmer specifies transfer
– Which side sends first
– Which side closes the connection first
– What to do if one side crashes unexpectedly

State In An Application Protocol
d Big decision: should state information be kept?
d Stateful protocol assumes previous requests have been
honored
d Stateless protocol assumes each request is independent
d Example of stateful interaction
– Request 1 specifies “read from file X”
– Request 2 specifies “read next 128 bytes”
d Example of stateless interaction
– Request 1 specifies “read bytes 0-127 from file X”
– Request 2 specifies “read bytes 128-255 from file X”

Examples Of Standard
Application Protocols

Application Protocol Examples
d Web browsing
d Email
d File transfer
d Remote login and remote desktop
d Domain Name System (name lookup)

Application-Layer Protocols For The Web
2
222222222222222222222222222222222222222222222222222222222222222222222
Standard Purpose
2
222222222222222222222222222222222222222222222222222222222222222222222
HyperText Markup A representation standard used to specify the
Language (HTML) contents and layout of a web page
2
222222222222222222222222222222222222222222222222222222222222222222222
Uniform Resource A representation standard that specifies the
Locator (URL) format and meaning of a web page identifier
2
222222222222222222222222222222222222222222222222222222222222222222222
HyperText Transfer A transfer protocol that specifies how a browser
Protocol (HTTP) interacts with a web server to transfer data
2
222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Reminder: keyword Transfer in the name of a protocol
means the protocol specifies message exchange

HyperText Markup Language (HTML)
d Representation standard for multimedia documents
d Specifies document is entirely in printable text
d Uses declarative rather than procedural approach
d Document includes metadata that can link to arbitrary item
d Document contains markup guidelines rather than precise,
detailed formatting or typesetting instructions
– Page can be displayed on arbitrary device
– Appearance depends on device
d Embedded tags control display
– Form is <tag_name> and </tag_name>

Uniform Resource Locator (URL)
d Representation standard
d A text string with punctuation characters separating the
string into (optional) subfields
d General form is:
protocol:// computer_name : port / document_name ? parameters
d Example where protocol, port, and parameters are omitted:
www . cs . purdue . edu / people / comer

HyperText Transfer Protocol (HTTP)
d Transfer protocol used with the Web
d Specifies format and meaning of messages
d Each message represented as text
d Transfers arbitrary binary data
d Can download or upload data
d Incorporates caching for efficiency
d Browser sends request to server

Four Major HTTP Request Types
2
222222222222222222222222222222222222222222222222222222222222222222222222
Request Description
2
222222222222222222222222222222222222222222222222222222222222222222222222
GET
Requests a document; server responds by sending status
information followed by a copy of the document
2
222222222222222222222222222222222222222222222222222222222222222222222222
HEAD
Requests status information; server responds by sending
status information, but does not send a copy of the document
2
222222222222222222222222222222222222222222222222222222222222222222222222
POST
Sends data to a server; the server appends the data to a
specified item (e.g., a message is appended to a list)
2
222222222222222222222222222222222222222222222222222222222222222222222222
PUT
Sends data to a server; the server uses the data to completely
replace the specified item (i.e., overwrites the previous data)
2
222222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d GET request has the form:
GET /item version CRLF
d Version is HTTP/1.0 or HTTP/1.1

HTTP Response
d Response begins with a header in text, optionally followed
by an item (which can be binary)
d Header uses keyword: information form like email header
d Header ends with a blank line

HTTP Header Format
d General form
HTTP/1.0 status_code status_string CRLF
Server: server_identification CRLF
Last-Modified: date_document_was_changed CRLF
Content-Length: datasize CRLF
Content-Type: document_type CRLF
CRLF
... item begins here and contains datasize bytes ...

Telnet Example (Apache Web Server)
$ telnet www.cs.purdue.edu 80
Trying 128.10.19.20...
Connected to lucan.cs.purdue.edu.
Escape character is ’^]’.
GET /homes/comer/ HTTP/1.0
HTTP/1.1 200 OK
Date: Sun, 10 Nov 2013 11:38:27 GMT
Server: Apache/2.2.11 (Unix) mod_ssl/2.2.11 OpenSSL/0.9.8r
Last-Modified: Mon, 17 Oct 2011 22:21:41 GMT
ETag: "bafb0-a50-4af8607f7c740"
Accept-Ranges: bytes
Content-Length: 2640
Connection: close
Content-Type: text/html
...data from the web page follows here

d Web browsing
d Email
d File transfer

Original End-To-End Email Paradigm
Internet
direct transfer
d Each computer runs
– Email server to accept incoming email
– Email client to send outgoing email
d Incoming mail deposited in user’s mailbox
d Outgoing mail placed in queue
d User interface to read or compose messages separate from
transfer applications

Current Email Paradigm
server
at ISP
server
at ISP
Internet
email transfer
protocol used
email access
protocol used
email access
protocol used
d User’s mailbox located on separate computer (usually at an
ISP)
d Mail transfer application deposits message in mailbox
d User interface application accesses remote mailbox
– A web browser may be used as an access mechanism
– Special-purpose applications also exist

Simple Mail Transfer Protocol (SMTP)
d Standard for email transfer
d Follows a stream paradigm
d Uses textual control messages
d Only transfers text messages
d Terminates message with <CR> <LF> . <CR> <LF>
d Allows a sender to specify recipients’ names and checks
each name
d Sends only one copy of a message to a computer, even if
destined to multiple recipients on the computer

Example SMTP Session
S: 220 somewhere.com Simple Mail Transfer Service Ready
C: HELO example.edu
S: 250 OK
C: MAIL FROM:<John_Q_Smith@example.edu>
S: 250 OK
C: RCPT TO:<Mathew_Doe@somewhere.com>
S: 550 No such user here
C: RCPT TO:<Paul_Jones@somewhere.com>
S: 250 OK
C: DATA
S: 354 Start mail input; end with <CR><LF>.<CR><LF>
C: ...sends body of mail message, which can contain
C: ...arbitrarily many lines of text
C: <CR><LF>.<CR><LF>
S: 250 OK
C: QUIT
S: 221 somewhere.com closing transmission channel

Mail Access Protocols
d Two standard protocols
– Post Office Protocol version 3 (POP3)
– Internet Mail Access Protocol (IMAP)
d Functionality
– Provide access to a user’s mailbox
– Permit user to view headers, download, delete, or send
individual messages
– Client runs on user’s personal computer
– Server runs on a computer that stores user’s mailbox

RFC2822 Mail Message Format
d Email representation standard
d Name derived from the Internet standard in which it is
defined
d Specifies
– Email message consists of text file
– Blank line separates header from body
– Header lines have the form:
Keyword: information

(continued)
d Some keywords have defined meanings:
– From:
– To:
– Subject:
– Cc:
d Keywords starting with uppercase X have no effect

(continued)
d Some keywords have defined meanings:
– From:
– To:
– Subject:
– Cc:
d Keywords starting with uppercase X have no effect
d Examples:
X-Best-networking-Course: CS422 at Purdue
X-Spam-Check-Results: bulk spam 90% likely
X-Worst-TV-Shows: any reality show

Multimedia Email

Multimedia Email
d Observe
– Email was standardized when computers only had
character-oriented (textual) interfaces
– SMTP is limited to transferring plain text messages
– Users want to email photos, spreadsheets, messages with
special fonts and color

Multimedia Email
d Observe
d Question: can SMTP be used to transfer such email?

Multimedia Email
d Observe
d Question: can SMTP be used to transfer such email?
d Answer: it is possible because one can encode arbitrary
binary items in plain text (think of a hex dump)

Sending Non-Text Email
d Standard is MIME (Multimedia Internet Mail Extensions)
d Backward compatible with RFC2822 mail and SMTP
d Sender
– Encodes arbitrary binary item in plain text
– Adds lines to email header to specify MIME
– Places additional headers before each item in the
message (including plain text items)
d Sender can specify content type and encoding
d Standard includes Base64 encoding

Examples Of Mime Headers
d MIME header lines added to other RFC2822 headers
MIME-Version: 1.0
Content-Type: Multipart/Mixed; Boundary=xyz123
d Each part of the message has a MIME header that starts
with the separator and specifies content type and encoding
d Example
--xyz123
Content-Type: image/jpeg
←blank line ends header

d Web browsing
d Email
d File transfer

File Transfer
d Standard is the File Transfer Protocol (FTP)
d Once accounted for the most packets on the Internet
d Interesting communication paradigm
– Client forms a control connection to send requests
– Server forms data connection for each file transferred
– Server closes data connection after transfer complete
d Notes
– Using a separate connection allows arbitrary data
transfer
– For data connections, the server becomes a client and the
client becomes a server (important for NAT)

Illustration Of FTP Communication
server
client
client forms a control connection
client sends directory request over the control connection
server forms a data connection
server sends directory listing over the data connection
server closes the data connection
client sends download request over the control connection
server forms a data connection
server sends a copy of the file over the data connection
server closes the data connection
client sends a QUIT command over control connection
client closes the control connection

d Web browsing
d Email
d File transfer

Remote Login And Remote Desktop
d Remote login
– Intended for systems with command-line interface
– Internet standard is TELNET
– Secure shell (ssh) encrypts transfers
– To appreciate the complexity of application protocols
look at the TELNET standard

Remote Login And Remote Desktop
d Remote login
– Intended for systems with command-line interface
– Internet standard is TELNET
– Secure shell (ssh) encrypts transfers
– To appreciate the complexity of application protocols
look at the TELNET standard
d Remote desktop
– Intended for systems that have a Graphical User
Interface (GUI)
– No Internet standards
– Move to thin client has revived interest

d Web browsing
d Email
d File transfer

Domain Name System (DNS)
d Important piece of Internet infrastructure
d Runs at the application layer
d Translates human-readable names into the binary addresses
used by the Internet Protocol
d Example
– Computer www.cs.purdue.edu
– Has the IP address 128.10.19.20

DNS Terminology
d Names are hierarchical
d Each name divided into segments by period character, which
is read “dot”
d Most significant segment is on the right
d Rightmost segment known as a top-level domain (TLD)
d Client program known as a resolver
– Used by web browser, email, etc

Top-Level Domains
2222222222222222222222222222222222222222222222222222
Domain Name Assigned To
2222222222222222222222222222222222222222222222222222
aero Air transport industry
2222222222222222222222222222222222222222222222222222
arpa Infrastructure domain
2222222222222222222222222222222222222222222222222222
asia For or about Asia
2222222222222222222222222222222222222222222222222222
biz Businesses
2222222222222222222222222222222222222222222222222222
com Commercial organizations
2222222222222222222222222222222222222222222222222222
coop Cooperative associations
2222222222222222222222222222222222222222222222222222
edu Educational institutions
2222222222222222222222222222222222222222222222222222
gov United States government
2222222222222222222222222222222222222222222222222222
info Information
2222222222222222222222222222222222222222222222222222
int International treaty organizations
2222222222222222222222222222222222222222222222222222
jobs Human resource managers
2222222222222222222222222222222222222222222222222222
mil United States military
2222222222222222222222222222222222222222222222222222
mobi Mobile content providers
2222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Top-Level Domains
(continued)
2
2222222222222222222222222222222222222222222222222
Domain Name Assigned To
2
2222222222222222222222222222222222222222222222222
museum Museums
2
2222222222222222222222222222222222222222222222222
name Individuals
2
2222222222222222222222222222222222222222222222222
net Major network support centers
2
2222222222222222222222222222222222222222222222222
org Non-commercial organizations
2
2222222222222222222222222222222222222222222222222
pro Credentialed professionals
2
2222222222222222222222222222222222222222222222222
travel Travel and tourism
2
2222222222222222222222222222222222222222222222222
xxx Adult entertainment (porn)
2
2222222222222222222222222222222222222222222222222
country code A sovereign nation
2
2222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d In 2014, ICANN decided to allow many new TLDs

Domain Registration
d Organization
– Applies under a specific top-level domain
– Can choose an internal hierarchy
– Assigns each computer a name
d Geographic registration is possible
cnri.reston.va.us
d Some countries impose conventions
– Universities in Great Britain register under
ac.uk

Domains With Most Hosts (July 2013)
2
22222222222222222222222222222222222222222222
Domain Hosts Explanation
2
22222222222222222222222222222222222222222222
net 366592151 Networks
com 163634309 Commercial
jp 74461142 Japan
de 34904481 Germany
br 33691951 Brazil
it 26136473 Italy
cn 19976554 China
mx 17658991 Mexico
fr 17437386 France
au 16900586 Australia
ru 15122103 Russian Federation
nl 14011944 Netherlands
pl 14011944 Poland
ar 13335042 Argentina
edu 12251571 Educational
ca 9004861 Canada
uk 8116718 United Kingdom
in 7429638 India
tr 7146979 Turkey
tw 6429021 Taiwan
2
22222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d See domain survey at www . isc . org for details

Host Names and Services Offered
d Many organizations choose a host name to match the service
a computer offers
mail.foobar.com
ftp.foobar.com
www.foobar.com
d Although convenient for humans, a host name does not
specify which servers are running (e,g., a computer named
mail could run a web server)

DNS Servers
d Names divided into a hierarchy of servers
d Multiple groupings possible
d Hypothetical example
com
foobar
candy soap
peanut almond walnut
(a)
root server
server for
foobar.com
server for
candy.foobar.com
com
foobar
candy soap
peanut almond walnut
(b)
root server
server for
walnut.candy.foobar.com
server for
foobar.com

Name Resolution And Caching
d Resolver
– Acts as a client
– Is configured with address of local DNS server
– Contacts local server first
– Socket library resolver is gethostbyname
d Caching
– Follows locality of reference principle
– Each DNS server caches results
– Cached item never kept when stale

DNS Server Algorithm Part 1
Given:
A request message from a DNS name resolver
Provide:
A response message that contains the address
Method:
Extract the name, N, from the request
if ( server is an authority for N ) {
Form and send an authoritative response
to the requester;
else if ( answer for N is in the cache ) {
Form and send a nonauthoritative response
to the requester;
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

DNS Server Algorithm Part 2
else { /* Need to look up an answer */
if ( authority server for N is known ) {
Send request to authority server;
} else {
Send request to root server;
}
Receive response and place in cache;
Form and send a response to the requester;
}
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1 11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2222222222222222222222222222222222222222222222222222222222

Summary
d Applications provide all Internet services
d Internet offers connection-oriented stream communication or
connectionless message communication
d Most applications follow client-server approach
– Server starts first and awaits client
– Client contacts server
d Socket API is a de facto standard
d Application-layer protocol can define
– Data and message formats (representation)
– Rules for message exchange (transfer)

Summary
(continued)
d Applications reviewed include
– Web (URL, HTML, HTTP)
– Email (SMTP, RFC2822, MIME)
– File transfer (FTP)
– Remote login and remote desktop (TELNET)
– Domain Name System (DNS)

MODULE III
Foundations Of Data Communications
And The Physical Layer

Topics
d Motivation and model
d Information sources and signals
d Transmission media
d Reliability and channel coding
d Transmission modes
d Modulation and demodulation
d Multiplexing and demultiplexing (channelization)

What Is Data Communications?
d Broad field of study
d Usually associated with the Physical Layer
d Touches on
– Physics
– Mathematics
– Engineering
d Includes
– Transmission of signals
– Encoding data
– Modulation and multiplexing

Motivation
d Find ways to transmit analog and digital information
– Using natural phenomena (e.g., electromagnetic
radiation)
– Allow multiple senders to share a transmission medium
d Data communications provides
– A conceptual framework
– Mathematical basis

Key Concept
Although we tend to think of analog and digital communication
separately, ultimately, all communication uses the same
physical phenomena, usually electromagnetic energy.
d Differences lie in the way the physical phenomena are used
– Analog: use all values in a continuous range
– Digital: restrict use to a fixed set of values, usually two
d Data communications covers both analog and digital

Conceptual Framework For Data Communications
Physical Channel
(noise & interference)
Modulator
Multiplexor
Channel Encoder
Encryptor (Scrambler)
Source Encoder
Information Source 1
Channel Encoder
Encryptor (Scrambler)
Source Encoder
Information Source N
Demodulator
Demultiplexor
Channel Decoder
Decryptor (Unscrambler)
Source Decoder
Destination 1
Channel Decoder
Decryptor (Unscrambler)
Source Decoder
Destination N
. . .
. . .

Information Sources
And Signals

Sources Of Information
d An input signal can arise from
– Transducer such as a microphone
– Receiver such as an Ethernet interface
d We use the term signal processing to describe the
recognition and transformation of signals

Sine Waves

Sine Waves
d Fundamental because sine waves characterize many natural
phenomena
d Examples
– Audible tones
– Radio waves
– Light energy

Fourier Analysis
d Multiple sine waves can be added together
– Result is known as a composite wave
– Corresponds to combining multiple signals (e.g., playing
two musical tones at the same time)
d Mathematician named Fourier discovered how to decompose
an arbitrary composite wave into individual sine waves
d Fourier analysis provides the mathematical basis for signal
processing
d Bad news: according to Fourier, a digital wave decomposes
into an infinite set of sine waves

Sine Wave Characteristics
d Three important characteristics are used in networks:
frequency, amplitude, and phase
0
1 sec
2 sec
1
-1
t
(a) Original sine wave: sin(2πt)

0 0
1 sec
0.5 sec
2 sec
1
-1
t
2 sec
1
-1
t
(a) Original sine wave: sin(2πt) (b) Higher frequency: sin(2π2t)

0 0
0
1 sec
1 sec
0.5 sec
2 sec
1
-1
t
2 sec
1
-1
t
2 sec
1
-1
t
(c) Lower amplitude: 0.4 sin(2πt)

0 0
0 0
1 sec
1 sec 1 sec
0.5 sec
2 sec
1
-1
t
2 sec
1
-1
t
2 sec
1
-1
t
2 sec
1
-1
t
(c) Lower amplitude: 0.4 sin(2πt) (d) New phase: sin(2πt+1.5π)

Definition Of Analog Bandwidth
d Decompose a signal into a set of sine waves and take the
difference between the highest and lowest frequency
d Easy to compute from a frequency domain plot
d Example signal with bandwidth of 4 Kilohertz (KHz):
1
0
1 2 3 4 5 6
frequency (in KHz)
amplitude
bandwidth

Digital Signals And Signal Levels
d A digital signal level can represent multiple bits
d Example
0
+5
time
amplitude
1
0
1 1
0 0 0
1
two levels with a
single bit per level
8 bits sent

d Example
0
+5
time
amplitude
-5
-2
+2
+5
time
amplitude
1
0
1 1
0 0 0
1 11
10
00 00
01
11
01
10
two levels with a
four levels with
two bits per level
8 bits sent 16 bits sent

d Example
0
+5
time
amplitude
-5
-2
+2
+5
time
amplitude
1
0
1 1
0 0 0
1 11
10
00 00
01
11
01
10
two levels with a
four levels with
two bits per level
8 bits sent 16 bits sent
d Baud rate is number of times signal changes per second;
data rate in bits per second = baud ×
J
Q
log2( levels )
J
P

Converting Digital To Analog
d Approximate digital signal with a composite of sine waves:
t
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.................
.
.
.
.
.
.
.
.
.
.
.
(a) digital signal (b) sin(2πt/2)
(c) sin(2πt/2)+α sin(2π3t/2) (d) sin(2πt/2)+α sin(2π3t/2)+β sin(2π5t/2)
d Mathematically, the bandwidth of a digital signal is infinite

Converting Analog To Digital
d Three steps taken during conversion
quantization
sampling encoding
PCM encoder
analog
signal
digital
data
d Example sampling using eight levels
time

Converting Analog To Digital
d Three steps taken during conversion
quantization
sampling encoding
PCM encoder
analog
signal
digital
data
d Example sampling using eight levels
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
.....................................................................................................................................................
0
1
2
3
4
5
6
7
time
quanta

Sampling Rate And Nyquist Theorem
d How many samples should be taken per second?

d Mathematician named Nyquist discovered the answer:
sampling rate = 2 × fmax
where fmax is highest frequency in the composite signal

d Mathematician named Nyquist discovered the answer:
sampling rate = 2 × fmax
where fmax is highest frequency in the composite signal
d Example: to capture audio frequencies up to 4000 Hertz, a
digital telephone system samples at 8000 samples per
second
d Amount of data generated by a single digitized voice call:
data rate = 8000
second
samples
3
3333333 × 8
sample
bits
3
333333 = 64,000
second
bits
3
333333

Nonlinear Encoding
d Linear sampling does not work well for voice
d Researchers created nonlinear sampling that modify
dynamic range to reproduce sounds to which the human ear
is sensitive
d Mu-law (µ-law)
– Used in North America and Japan
– More dynamic range, but more sensitive to noise
d A-law
– Used in Europe
– Less sensitive to noise, but less dynamic range

Synchronization Errors And Line Coding
d Synchronization error occurs when receiver and sender
disagree about bit boundaries (clocks differ)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 0 1 1 0 1 0
1 0 0 0 1 1 0 1 1 0
sent
received
d Line coding techniques prevent synchronization errors

Example Line Coding: Manchester Encoding
d Used with Ethernet
d Synchronizes receiver with sender (transition represents bit)
d Example of (a) Manchester Encoding, and (b) differential
Manchester Encoding:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 1 0 0 1 1 1 0
0 1 0 0 1 1 1 0
(a)
(b)

A Taxonomy Of Transmission Media
Twisted Pair
Coaxial Cable
Optical Fiber
InfraRed
Laser
Terrestrial
Radio
Satellite
Electrical
Electromagnetic
(Radio)
Light
Energy Types
d Is anything omitted?

Some Really Bad News
d In the real world, entropy rules
d Transmission is plagued with problems

Loss, Interference, And Electrical Noise
d Problems in the electrical and electromagnetic worlds
– Resistance (leads to loss)
– Capacitance (leads to distortion)
– Inductance (leads to interference)
d Random electromagnetic radiation is called noise
– Can be generated by specific sources such as electric
motor
– Background radiation is an inescapable feature of the
universe

Examples
d When electrical signals propagate down a wire,
electromagnetic energy is radiated (i.e., the wire acts like an
antenna)
d When electromagnetic radiation encounters metal, a small
electrical current is induced that can interfere with signals
being carried on the wire
d When an electrical pulse is sent down an unterminated wire,
reflection comes back
d When a signal passes across the connection between two
wires, reflection and loss occur
d Note: a network diagnostic tool uses reflection to find the
distance to the point where a cable has been cut

How Can We Reduce The Effect Of Noise
On Copper Wiring
d Several techniques have been invented
– Unshielded Twisted Pair (UTP)
– Coaxial cable
– Shielded Twisted Pair (STP)
d All are used in computer networks

How Twisted Pair Helps

+5 +5 +5 +5
+3 +3 +3 +3
difference +8
source of radiation
d In an untwisted pair of wires, more current is generated in
first wire the interference hits

+5 +5 +5 +5
+3 +3 +3 +3
difference +8
source of radiation
d In an untwisted pair of wires, more current is generated in
first wire the interference hits
+5 +5 +5 +5
+3 +3 +3 +3
difference 0
source of radiation
d Twisting exposes each wire equally

Coaxial Cable And Shielding
d Better protection: wrap a metal shield around the wire
outer plastic covering
braided metal shield
plastic insulation
inner wire for signal
d Shielding can be added to twisted pair
– Around entire cable containing many pairs
– Around each pair as well as around cable
d Shielding determines maximum data rate

Wiring Standards And Data Rates
2
2222222222222222222222222222222222222222222222222222222222222222222222
Category Description Data Rate
(in Mbps)
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 1 Unshielded twisted pair used for telephones < 0.1
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 2 Unshielded twisted pair used for T1 data 2
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 3 Improved CAT2 used for computer networks 10
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 4 Improved CAT3 used for Token Ring networks 20
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 5 Unshielded twisted pair used for networks 100
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 5E Extended CAT5 for more noise immunity 125
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 6 Unshielded twisted pair tested for 200 Mbps 200
2
2222222222222222222222222222222222222222222222222222222222222222222222
CAT 7 Shielded twisted pair with a foil shield 600
around the entire cable plus a shield around
each twisted pair
2
2222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d What common data rate is missing from the list?

Media Using Light Energy
d InfraRED transmission (short range and low data rate)
d Point-to-point lasers (useful between buildings)
d Optical fiber (high data rate and long distance)

Media Using Light Energy
d InfraRED transmission (short range and low data rate)
d Point-to-point lasers (useful between buildings)
d Optical fiber (high data rate and long distance)
d Why light stays in a fiber:
α α
(a) (b) (c)
Refraction Absorption Reflection
critical
angle
low
density
high
density
θ

Electromagnetic Spectrum And Properties
100 102 104 106 108 1010 1012 1014 1016 1018 1020 1022 1024
Radio & TV
Low
frequencies
Micro-
wave InfraRed UV X ray Gamma
ray
1 KHz 1 MHz 1 GHz 1 THz visible light
2
222222222222222222222222222222222222222222222222222222222222222222222
Classification Range Type Of Propagation
2
222222222222222222222222222222222222222222222222222222222222222222222
Low
< 2 MHz
Wave follows earth’s curvature, but
Frequency can be blocked by unlevel terrain
2
222222222222222222222222222222222222222222222222222222222222222222222
Medium
2 to 30 MHz
Wave can reflect from layers of the
Frequency atmosphere, especially the ionosphere
2
222222222222222222222222222222222222222222222222222222222222222222222
High
> 30 MHz
Wave travels in a direct line, and will
Frequency be blocked by obstructions
2
222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1

Satellite Communication
d Three types of communication satellites
2222222222222222222222222222222222222222222222222222222222222222222222
Orbit Type Description
2222222222222222222222222222222222222222222222222222222222222222222222
Low Has the advantage of low delay, but the disadvantage
Earth Orbit that from an observer’s point of view on the earth,
( LEO ) the satellite appears to move across the sky
2222222222222222222222222222222222222222222222222222222222222222222222
Medium An elliptical (rather than circular) orbit primarily
Earth Orbit used to provide communication at the North and
( MEO ) South Poles
2222222222222222222222222222222222222222222222222222222222222222222222
Geostationary Has the advantage that the satellite remains at a fixed
Earth Orbit position with respect to a location on the earth’s
( GEO ) surface, but the disadvantage of being farther away
2222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

GEO Satellites
d Figure below shows the earth’s atmosphere drawn to scale
d Where would a GEO satellite be in the figure?
Earth
atmosphere

GEO Satellites
(continued)
d Distance to GEO satellite is 35,785 km or 22,236 miles
d Approximately 3 times earth’s diameter or one-tenth of the
distance to the moon

GEO Satellites
(continued)
d In other words: the satellite is far off the page

GEO Satellites
(continued)
d In other words: the satellite is far off the page
d A consequence for networking: a long round-trip time, even
at the speed of light:
Round trip time =
3 × 108 meters/sec
2 × 35.8 × 106meters
3
3333333333333333333 = 0.238 sec

Measures Of Transmission Media
d Propagation delay - time required for a signal to traverse a
medium
d Channel capacity - maximum data rate

Channel Capacity
d Nyquist’s Theorem gives theoretical bound on maximum
data rate for hardware bandwidth B and K signal levels
D = 2 B log2K
d Mathematical result known as Shannon’s Theorem gives the
maximum channel capacity, C, in the presence of noise
C = B log2( 1 + S/N)
d Quantity S / N is known as the signal-to-noise ratio

Assessment
d Nyquist’s Theorem gives us hope: using more signal levels
can increase the data rate
d Shannon’s Theorem is sobering: electrical noise in the
universe limits the effective channel capacity of any
practical communication system

Reliability And Channel Coding

Sources Of Errors And Types
d Error sources: interference, distortion, and attenuation
d Resulting error types:
2
22222222222222222222222222222222222222222222222222222222222222222222222
Type Of Error Description
2
22222222222222222222222222222222222222222222222222222222222222222222222
Single Bit Error A single bit in a block of bits is changed and
all other bits in the block are unchanged (often
results from very short-duration interference)
2
22222222222222222222222222222222222222222222222222222222222222222222222
Burst Error Multiple bits in a block of bits are changed
(often results from longer-duration interference)
2
22222222222222222222222222222222222222222222222222222222222222222222222
Erasure (Ambiguity) The signal that arrives at a receiver is ambiguous
(does not clearly correspond to either a logical 1
or a logical 0; can result from distortion
or interference)
2
22222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Channel coding used to detect and correct errors

Concept Of Forward Error Correction (FEC)
encoder
add extra bits
for protection
output codeword
transmission over channel
decoder
check and
optionally correct
receive codeword
ORIGINAL MESSAGE ORIGINAL MESSAGE
Discard
d Examples:
– Single parity bit
– Row And Column (RAC)
– Cyclic Redundancy Check (CRC)

Example: Row And Column Code
d To send 12 bits, arrange the bits in a matrix, compute a
parity for each row and column, and send 20 bits
1 0 1 1 1
0 0 1 0 1
1 0 1 0 0
0 0 1 1 0
parity for
each column
parity for
each row
bits from
dataword

Example: Row And Column Code
d To send 12 bits, arrange the bits in a matrix, compute a
parity for each row and column, and send 20 bits
1 0 1 1 1
0 0 1 0 1
1 0 1 0 0
0 0 1 1 0
parity for
each column
parity for
each row
bits from
dataword
d Receiver computes same parity for the 12 bits and compares
to the parity bits received
1 0 1 1 1
0 1 1 0 1
1 0 1 0 0
0 0 1 1 0
single bit
changed during
transmission
locations where
calculated parity
bits disagree,
indicating the
row and column
of the error

Hamming Distance
d Used to assess code’s resistance to errors
d Defined to be number of bit changes to transform bit string
S1 into bit string S2
d Can be computed as number of 1 bits in the exclusive or of
S1 and S2
d To assess code’s strength, compute Hamming distance
among all possible pairs of codewords, and take the
minimum
d If minimum Hamming distance is n, an error that changes
fewer than n bits will be detected

Internet Checksum Computation
Given:
A message, M, of arbitrary length
Compute:
A 16-bit 1s complement checksum, C
Method:
Pad M to an exact multiple of 16 bits;
Set a 32-bit checksum integer, C, to zero;
for ( each 16-bit group in M ) {
Treat the 16 bits as an integer and add to C;
}
Extract high-order 16 bits of C and add to C;
Checksum is inverse of the low-order 16 bits;
If the checksum is zero, substitute all 1s;
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

Cyclic Redundancy Code (CRC)
d Used with Ethernet and other high-speed networks
d Properties:
Arbitrary Length
Message
Excellent Error
Detection
Fast Hardware
Implementation
As with a checksum, the size of a dataword is not fixed,
which means a CRC can be applied to an arbitrary length
message
Because the value computed depends on the sequence
of bits in a message, a CRC provides excellent error
detection capability
Despite its sophisticated mathematical basis, a CRC
computation can be carried out extremely fast by
hardware

Explanation Of CRC
d Mathematicians explain CRC computation as the remainder
from polynomial division
d Theoretical computer scientists explain CRC as the
remainder from a division of binary numbers
d Cryptographers explain CRC as an operation in a Galois
field of order 2
d Computer programmers explain CRC as an algorithm that
iterates through a message and uses table lookup
d Hardware architects explain CRC computation as a small
hardware pipeline unit that uses exclusive or

Question
d Can you explain the following?
– Fact 1: it is possible to write a function that computes
the 32-bit CRC used with Ethernet
– Fact 2: commercial Ethernet products use hardware
instead of software to compute a CRC

Terminology
d Serial - one bit at a time
d Parallel - multiple bits at a time

Terminology
d Serial - one bit at a time
d Parallel - multiple bits at a time
d Taxonomy of transmission methods:
Isochronous
Synchronous
Asynchronous
Serial
Parallel
Transmission Mode

Serial Ordering Of Bits And Bytes
d Both sides must agree on order in which bits are transmitted
d Two approaches known as big-endian and little-endian
d Example: Ethernet uses byte big-endian and bit little-endian
order
byte 1 byte 2 byte 3 byte 4
x x x x
x x x x
x x x x
x x x x
x x x x
x x x x
x x x x
x x x x
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

Asynchronous And Synchronous Transmission
d Asynchronous: line idle when not in use; data starts at
arbitrary time
0
+15
-15
voltage
time
arbitrary
idle start 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 1 0 stop
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
idle
arbitrary
d Synchronous: each bit slot used
0
+15
-15
voltage
time
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 1 0 1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0
receiver must know how
to group bits into bytes

Illustration Of Simplex And Duplex Modes
send receive
receive
send
send
receive
receive
send
send
receive
(a) simplex
(b) full-duplex
(c) half-duplex

Illustration Of Amplitude Modulation

carrier

carrier
signal

carrier
signal
modulated carrier

Illustration Of Frequency Modulation
signal
modulated carrier

Shift Keying
d Like modulation except signal is digital
0 1 0 0 1 1 0
carrier
digital signal
carrier with amplitude shift keying

A Challenge
Write a computer program that takes as input a series of points
defining a signal and produces plots of sine waves that show
amplitude and frequency modulation as in the previous
diagrams

Other Modulation Topics
d Phase shift modulation
d Increasing bits per second by combining amplitude and
phase shift (QAM techniques)
d Constellation diagrams to represent combinations
d Modems (modulator / demodulator)

Multiplexing And Demultiplexing
(Channelization)

Concept Of Multiplexing And Types
sender 1
sender 2
sender N
receiver 1
receiver 2
receiver N
.
.
.
.
.
.
multiplexor demultiplexor
shared medium

Concept Of Multiplexing And Types
sender 1
sender 2
sender N
receiver 1
receiver 2
receiver N
.
.
.
.
.
.
shared medium
d Types:
– Frequency division multiplexing
– Wavelength division multiplexing
– Time division multiplexing
– Code division multiplexing

Frequency Division Multiplexing (FDM)
d Used in broadcast radio and cable TV
sender 1
sender 2
sender N
receiver 1
receiver 2
receiver N
channel 1
channel 2
channel N
.
.
.
.
.
.
.
.
.
d Demultiplexing implemented with sets of filters
filter 1
filter 2
filter N
.
.
.
demultiplexor
frequencies
for all
channels
each output has
frequencies for
one channel

FDM In Practice
d Each channel assigned a range of frequencies
2
22222222222222222222222222222222
Channel Frequencies Used
2
22222222222222222222222222222222
1 100 KHz - 300 KHz
2 320 KHz - 520 KHz
3 540 KHz - 740 KHz
4 760 KHz - 960 KHz
5 980 KHz - 1180 KHz
6 1200 KHz - 1400 KHz
2
22222222222222222222222222222222
11
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
d A guard band separates adjacent channels
0 200 400 600 800 1000 1200 1400
1 2 3 4 5 6 KHz
guard band

Wavelength Division Multiplexing (WDM)
d Form of FDM used with light (i.e., on an optical fiber)
d Separate frequencies called colors or lambdas
d Prisms used to separate frequencies
λ1
λ2
λk
λ1
λ2
λk
optical fiber
carrying a beam of light
prism
d Current technology is Dense WDM (DWDM); an individual
channel can provide 10 Gbps

Time Division Multiplexing
d Senders take turns transmitting
sender 1
sender 2
sender N
receiver 1
receiver 2
receiver N
.
.
.
.
.
.
1
2
3
. . .
N
1
2
3
. . .
data flow
d Synchronous TDM
– Each sender assigned a slot (typically round-robin)
– Used by the telephone company
d Statistical TDM
– Sender only transmits when ready (e.g., Ethernet)

Code Division Multiplexing
d Mathematical form of multiplexing used with cell phones
d Algorithm
– Each sender/receiver pair is assigned a unique number
called a chip sequence
– Senders multiply the data value by their chip sequence
(orthogonal vector spaces)
– Transmitted value is a sum of all senders
– Each receiver multiplies incoming value by its chip
sequence to extract data
d Advantage over statistical TDM: lower delay when network
loaded

Hierarchical Multiplexing
d Hierarchies used with FDM and TDM to combine multiple
lower-capacity channels
d Example of TDM hierarchy used by the phone system
24 DS-0 digital phone
channels (64 Kbps each)
channels (1.544 Mbps each)
channel (274.176 Mbps total)

Inverse Multiplexing
d Divides data from a single channel into several lower-speed
channels
d Used when high-speed channel is unavailable or too
expensive
d Some ISPs use inverse multiplexing to combine several 10
Gbps channels into a higher-speed channel
multiple low-speed connections
single high-speed
input
single high-speed
output

Summary
d Data communications deals with the Physical Layer and
data transmission
d Concepts include
– Signals and conversion between digital and analog
– Transmission media
– Reliability and channel coding
– Modulation and demodulation
– Multiplexing and demultiplexing

MODULE IV
Computer Network Technologies:
Access, Wired And Wireless LANs,
Extensions, Bridging, And
Layer 2 Switching

Topics
d Access technologies
d Interconnection technologies
d Local area network packets, frames, and topologies
d Media access mechanisms and the IEEE MAC sub-layer
d Wired LAN technologies (Ethernet and 802.3)
d Wireless Networking Technologies
d LAN Extensions
d Switches and switched networks

Definition Of Access
d Used in the “last mile” between a provider and a subscriber
d Informally classified as either narrowband or broadband
d May not be the bottleneck
d Many are asymmetric with higher data rate downstream
provider’s
facility
subscriber’s
location
downstream
upstream
d Note: party that is downstream pays a fee for service

Access Technology Types
d Narrowband (less than 128 Kbps)
– Dialup
– Integrated Services Digital Network (ISDN)
– Is disappearing
d Broadband (more than 128 Kbps)
– Digital Subscriber Line (DSL)
– Cable modems
– Wireless (e.g., Wi-Fi and 4G)

Digital Subscriber Line (DSL) Technologies
d Use frequency-division multiplexing to share local loop
between data and POTS
d Head-end equipment is DSL Access Multiplexor (DSLAM)
d Asymmetric Digital Subscriber Line (ADSL)
– 255 downstream carrier frequencies, 31 upstream
– Maximum downstream data rate is 8.45 Mbps
– Adaptive selection of carrier frequencies
0 4 26 138 1100
KHz
POTS upstream downstream

Cable Modem Technology
d Sends data over CATV coaxial cable system
d Standard is DOCSIS (Data-Over-Cable Service Interface
Specification)
d Head-end equipment known as Cable Modem Termination
System (CMTS)
d Version 1.x uses frequency-division multiplexing
d Maximum downstream data rate is 52 Mbps
d Bandwidth shared among multiple subscribers

Specification)
System (CMTS)
– Each subscriber receives
N
1
33 of the bandwidth

Specification)
System (CMTS)
– Each subscriber receives
N
1
33 of the bandwidth
– Cable company chooses N

Other Access Technologies
d Hybrid systems include optical fiber plus copper
– Fiber To The Curb (FTTC)
– Fiber To The Building (FTTB)
– Fiber To The Premises (FTTP)
– Fiber To The Home (FTTH)
d Key question: how much capacity is needed at each point
downstream?
d Answer: it depends on whether endpoints have traffic in
common
– Broadcasts are shared
– Individual communications are not

Other Access Technologies
(continued)
d Wireless
– Wi-Fi
– WIMAX
– Satellite
– 3G and 4G cellular services
d Leased point-to-point circuits (e.g., T1 or fractional T1)

Interconnections At The Core Of The Internet
d Typically needed by large ISPs
d Circuits leased from common carriers (phone companies)
d Terminated with a Data Service Unit / Channel Service Unit
(DSU/ CSU)
d Upstream interface aggregates many lower-speed access
connections
d Key idea: data rates based on voice
– Basic data rate: single digital voice channel (64 Kbps)
– Higher data rate circuits created from multiples of voice
channels
d SONET encoding and framing used

Example Data Rates Of Leased Circuits
222222222222222222222222222222222222222222222222222222222222
Name Bit Rate Voice Circuits Location
222222222222222222222222222222222222222222222222222222222222
basic rate 0.064 Mbps 1
222222222222222222222222222222222222222222222222222222222222
T1 1.544 Mbps 24 North America
222222222222222222222222222222222222222222222222222222222222
222222222222222222222222222222222222222222222222222222222222
222222222222222222222222222222222222222222222222222222222222
E1 2.048 Mbps 30 Europe
222222222222222222222222222222222222222222222222222222222222
222222222222222222222222222222222222222222222222222222222222
222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d T-standards used in North America
d E-standards used in Europe
d Note: T prefix specifies encoding as well as data rate; data
rate alone is given by Digital Signal Level (DS) standards

High Capacity Data Circuits
2
22222222222222222222222222222222222222222222222222222222222222222
Copper Name Optical Name Bit Rate Voice Circuits
2
22222222222222222222222222222222222222222222222222222222222222222
STS-1 OC-1 51.840 Mbps 810
2
22222222222222222222222222222222222222222222222222222222222222222
STS-3 OC-3 155.520 Mbps 2430
2
22222222222222222222222222222222222222222222222222222222222222222
STS-12 OC-12 622.080 Mbps 9720
2
22222222222222222222222222222222222222222222222222222222222222222
STS-24 OC-24 1,244.160 Mbps 19440
2
22222222222222222222222222222222222222222222222222222222222222222
STS-48 OC-48 2,488.320 Mbps 38880
2
22222222222222222222222222222222222222222222222222222222222222222
STS-192 OC-192 9,953.280 Mbps 155520
2
22222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
d STS standards specify copper interface
d OC standards specify optical fiber interface
d Suffix C on OC-standards means single channel

Local Area Networks:
(Packets, Frames, Topologies)

Networks
d Distinct from physical communication systems
d Attach multiple endpoints
d Two broad categories
– Circuit switched
– Packet switched

Circuit Switched Networks
d Provide point-to-point communication between pairs of
endpoints
d Establish path between sender and receiver
d Separate steps for circuit creation, use, and termination
d Performance equivalent to an isolated physical path
d Circuit can be
– Permanent/ provisioned (left in place for long periods)
– Switched (created on demand)
d Concept: user leases piece of underlying infrastructure for a
time period

Packet Switched Networks
d Form the basis for the Internet
d Multiplex communication over shared media
d All data divided into packets (maximum size fixed)
d After sending one packet, sender allows others a chance to
transmit before sending a second packet
d Arbitrary, asynchronous communication
d No set-up required before communication begins
d Performance varies due to statistical multiplexing
d Concept: underlying infrastructure is shared among users

Illustration Of Circuit And Packet Switching
d Circuit switching
circuit-switched network

d Circuit switching provides 1-to-1 dedicated connections

d Packet switching
1
2
1
2
3
. . .
packet-switched network

d Packet switching provides statistical TDM sharing
1
2
1
2
3
. . .
packet-switched network

Categories Of Packet Switched Networks
2222222222222222222222222222222222222222222222222222222222222222222222
Name Expansion Description
2222222222222222222222222222222222222222222222222222222222222222222222
LAN Local Area Network Least expensive; spans a single
room or a single building
2222222222222222222222222222222222222222222222222222222222222222222222
MAN Metropolitan Area Network Medium expense; spans a major
city or a metroplex
2222222222222222222222222222222222222222222222222222222222222222222222
WAN Wide Area Network Most expensive; spans sites in
multiple cities
2222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Categories Of Packet Switched Networks
2222222222222222222222222222222222222222222222222222222222222222222222
2222222222222222222222222222222222222222222222222222222222222222222222
LAN Local Area Network Least expensive; spans a single
room or a single building
2222222222222222222222222222222222222222222222222222222222222222222222
MAN Metropolitan Area Network Medium expense; spans a major
city or a metroplex
2222222222222222222222222222222222222222222222222222222222222222222222
WAN Wide Area Network Most expensive; spans sites in
multiple cities
2222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Everyone loves names that end in “AN”
2222222222222222222222222222222222222222222222222222222222222222222222222
2222222222222222222222222222222222222222222222222222222222222222222222222
PAN Personal Area Network Spans the area around an individual
used for earphones
2222222222222222222222222222222222222222222222222222222222222222222222222
SAN Storage Area Network Spans the distance between a disk
farm and processors in a data center
2222222222222222222222222222222222222222222222222222222222222222222222222
CAN Chip Area Network Spans a single chip and connects
processor, memories, etc.
2222222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Standards Bodies And Their Bias
d Standards bodies and academic departments each emphasize
certain layers of a protocol stack, leading to the following
views
Application
Transport
Internet
Data Link
Physical
APPLICATION
TRANSPORT
INTERNET
DATA LINK
PHYSICAL
textbooks W3C IETF IEEE

IEEE 802 Model And Standards
d IEEE (Institute of Electrical and Electronics Engineers)
– Professional society of engineers
– Standardizes vendor-independent technologies
d Project 802
– LAN/ MAN standards committee
– Organized in 1980
– Focuses on layer 1 and layer 2 standards
– Divides layer 2 into two sublayers
* Logical Link Control (LLC)
* Media Access Control (MAC)

Example IEEE Standards
2
2222222222222222222222222222222222222222222222222222222222
ID Topic
2
2222222222222222222222222222222222222222222222222222222222
802.1 Higher layer LAN protocols
2
2222222222222222222222222222222222222222222222222222222222
802.2 Logical link control
2
2222222222222222222222222222222222222222222222222222222222
802.3 Ethernet
2
2222222222222222222222222222222222222222222222222222222222
802.4 Token bus (disbanded)
2
2222222222222222222222222222222222222222222222222222222222
802.5 Token Ring
2
2222222222222222222222222222222222222222222222222222222222
802.6 Metropolitan Area Networks (disbanded)
2
2222222222222222222222222222222222222222222222222222222222
802.7 Broadband LAN using Coaxial Cable (disbanded)
2
2222222222222222222222222222222222222222222222222222222222
802.9 Integrated Services LAN (disbanded)
2
2222222222222222222222222222222222222222222222222222222222
802.10 Interoperable LAN Security (disbanded)
2
2222222222222222222222222222222222222222222222222222222222
802.11 Wireless LAN (Wi-Fi)
2
2222222222222222222222222222222222222222222222222222222222
802.12 Demand priority
2
2222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

More Example IEEE Standards
222222222222222222222222222222222222222222222222
ID Topic
222222222222222222222222222222222222222222222222
802.13 Category 6 - 10Gb LAN
222222222222222222222222222222222222222222222222
802.14 Cable modems (disbanded)
222222222222222222222222222222222222222222222222
802.15 Wireless PAN
802.15.1 (Bluetooth)
802.15.4 (ZigBee)
222222222222222222222222222222222222222222222222
802.16 Broadband Wireless Access
802.16e (Mobile) Broadband Wireless
222222222222222222222222222222222222222222222222
802.17 Resilient packet ring
222222222222222222222222222222222222222222222222
802.18 Radio Regulatory TAG
222222222222222222222222222222222222222222222222
802.19 Coexistence TAG
222222222222222222222222222222222222222222222222
802.20 Mobile Broadband Wireless Access
222222222222222222222222222222222222222222222222
802.21 Media Independent Handoff
222222222222222222222222222222222222222222222222
802.22 Wireless Regional Area Network
222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Standards Define
d Network topology (shape)
d Endpoint addressing scheme
d Frame (packet) format
d Media access mechanism
d Physical layer aspects and wiring

Illustration Of The Four LAN Topologies
Bus Ring
Mesh
Star
d Each topology has advantages and disadvantages

Endpoint Addressing Scheme
d Each station on a LAN is assigned a unique address
d Each packet specifies a destination address
d LAN hardware uses the address in a packet to determine
which station(s) receive a copy

IEEE Standard For Addressing
d Formal name: IEEE Media Access Control address (MAC
address)
d Informally called an Ethernet address
d Each address is 48 bits long
d Assigned to Network Interface Card (NIC) when device
manufactured
d Divided into subfields
– 3-byte Organizationally Unique ID (OUI)
– 3-byte Network Interface Controller (NIC)

Illustration Of Fields In An IEEE 48-Bit Address
Network Interface Controller
(NIC) specific
Organizationally Unique
Identifier (OUI)
8
7
6
5
4
3
2
1
3 bytes 3 bytes
bits of most significant byte
0 → unicast, 1→ multicast
0 → global, 1→ local
d Address types
2
22222222222222222222222222222222222222222222222222222222222222222222
Address Type Meaning And Packet Delivery
2
22222222222222222222222222222222222222222222222222222222222222222222
unicast Destination is a single computer; only that computer
should receive a copy of the packet
2
22222222222222222222222222222222222222222222222222222222222222222222
broadcast Destination is all computers on a network; they
should each receive a copy of the packet
2
22222222222222222222222222222222222222222222222222222222222222222222
multicast A subset of the computers on a network should
receive a copy of the packet
2
22222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Algorithm For Processing An Incoming Packet
Purpose:
Handle a packet that has arrived over a LAN
Method:
Extract destination address, D, from the packet;
if ( D matches “my address” ) {
accept and process the packet;
} else if ( D matches the broadcast address ) {
} else if ( D matches one of the multicast addresses for a
multicast group of which I am a member ) {
} else {
ignore the packet;
}
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
222222222222222222222222222222222222222222222222222222222222

Frame Format
d Layer 2 packet is called a frame
d General layout of a frame
HEADER PAYLOAD
optional prelude optional postlude
d Header usually has fixed fields
d Each technology imposes a maximum payload size
d Note: we will see specific frame formats later

Framing And Serial Communications Systems

d Consider sending packets over a leased circuit

d Circuit hardware either provides a stream of bits or a stream
of bytes (characters)
d We will consider hardware that provides a byte stream
– No frame boundaries
– Any 8-bit value can appear in the data

d Circuit hardware either provides a stream of bits or a stream
of bytes (characters)
d We will consider hardware that provides a byte stream
– No frame boundaries
– Any 8-bit value can appear in the data
d How can we send packets over such a system?
d Answer: sender and receiver must agree on framing

Example Framing Used With A Leased Circuit
d Use SOH and EOT characters to mark the start and end of a
frame
SOH EOT
HEADER PAYLOAD
6 bytes arbitrary bytes

Example Framing Used With A Leased Circuit
d Use SOH and EOT characters to mark the start and end of a
frame
SOH EOT
HEADER PAYLOAD
6 bytes arbitrary bytes
d Use byte stuffing within the payload
2
2222222222222222222222222222222222
Byte In Payload Sequence Sent
2
2222222222222222222222222222222222
SOH ESC A
2
2222222222222222222222222222222222
EOT ESC B
2
2222222222222222222222222222222222
ESC ESC C
2
2222222222222222222222222222222222
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1

Illustration Of Byte Stuffing
ESC SOH EOT ESC
ESC C ESC A ESC B ESC C
original data
stuffed data
d Internet uses SLIP or PPP (standards) for transmission over
serial circuits
d Bit stuffing techniques are also available for systems that
transfer a stream of bits

Media Access Mechanisms
(IEEE MAC Sublayer)

MAC Protocols
d Control access to shared medium
d Two types of channel allocation
– Static
– Dynamic
d General principle:
Static channel allocation suffices when the set of
communicating entities is known in advance and does not
change; most networks require a form of dynamic channel
allocation.

Taxonomy Of Media Access Mechanisms
Reservation
Polling
Token passing
ALOHA
CSMA / CD
CSMA / CA
FDMA
TDMA
CDMA
Controlled Access
Protocols
Random Access
Protocols
Channelization
Protocols
Multi-Access
Protocols

Channelization Protocols
d Employ and extend basic multiplexing techniques
d May be static or dynamic
d Three basic types
222222222222222222222222222222222222222222222
Protocol Expansion
222222222222222222222222222222222222222222222
FDMA Frequency Division Multi-Access
222222222222222222222222222222222222222222222
TDMA Time Division Multi-Access
222222222222222222222222222222222222222222222
CDMA Code Division Multi-Access
222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1

Controlled Access Protocols
d Three principal forms
2
2222222222222222222222222222222222222222222222222222222222222222
Type Description
2
2222222222222222222222222222222222222222222222222222222222222222
Polling Centralized controller repeatedly polls stations
and allows each to transmit one packet
2
2222222222222222222222222222222222222222222222222222222222222222
Reservation Stations submit a request for the next round of
data transmission
2
2222222222222222222222222222222222222222222222222222222222222222
Token Passing Stations circulate a token; each time it receives
the token, a station transmits one packet
2
2222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d All three have been used in practice

Algorithm For Polled Access
Purpose:
Control transmission of packets through polling
Method:
Controller repeats forever {
Select a station, S, and send a polling message to S;
Wait for S to respond by sending a packet or passing;
}
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

Algorithm For Reservation-Based Access
d Often used with satellite systems
d Stations inform a controller if they have data to send
Purpose:
Control transmission of packets through reservation
Method:
Controller repeats forever {
Form a list of stations that have a packet to send;
Allow each station on the list to transmit;
}
11
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

Algorithm For Token Passing Access
d Special packet known as a token passed among senders
d Station sends one packet each time token arrives
Purpose:
Control transmission of packets through token passing
Method:
Each computer on the network repeats {
Wait for the token to arrive;
Transmit a packet if one is waiting to be sent;
Send the token to the next station;
}
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

Example Random Access Protocols
22222222222222222222222222222222222222222222222222222222222222222222
Type Description
22222222222222222222222222222222222222222222222222222222222222222222
ALOHA Historic protocol used in an early radio network in
Hawaii; popular in textbooks and easy to analyze,
but not used in real networks
22222222222222222222222222222222222222222222222222222222222222222222
CSMA / CD Carrier Sense Multi-Access with Collision Detection
The basis for the original Ethernet, and the most widely
used random access protocol
22222222222222222222222222222222222222222222222222222222222222222222
CSMA / CA Carrier Sense Multi-Access with Collision Avoidance
The basis for Wi-Fi wireless networks
22222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1

Aloha
d Used in early network in Hawaii (ALOHAnet)
d Two carrier frequencies, inbound and outbound
d Central transmitter rebroadcast each incoming packet
central
transmitter
outbound frequency
inbound frequency
outlying station
d If inbound packets collide, each sender waits a random time
and retransmits
d Channel utilization under 20%

CSMA / CD
d Used in original Ethernet (1973)
d Provides access to shared medium
d Principle features
– Carrier Sense (CS)
– Multiple Access (MA)
– Collision Detection (CD)
d Uses binary exponential backoff

CSMA / CD Algorithm
Method:
When a packet is ready, perform CS (wait for access);
Delay for the interpacket gap;
Set variable x to the standard backoff range, d ;
Attempt to transmit the packet and perform CD;
While (collision occurred during transmission) {
Choose q to be a random delay between 0 and x ;
Delay for q microseconds;
Double x in case needed for the next round;
Attempt to retransmit the packet and perform CD;
}
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

CSMA / CA
d Alternative to CSMA / CD
d Used in wireless networks (Wi-Fi)
d Needed because signals have limited distance, δ
d Example: computer 1 cannot receive transmission when
computers 2 and 3 communicate
computer 1 computer 2 computer 3
δ δ
d All computers in range of computers 2 and 3 must be
informed that a transmission will occur

Illustration Of CSMA / CA
1: ready to send (RTS)
2: clear to send (CTS)
2: clear to send (CTS)
3: packet transmission
computer 1 computer 2 computer 3
d Communicating pair exchange RTS and CTS before packet
transmission
d Any computer less than δ away from either computer 2 or 3
hears at least one of the RTS / CTS messages

Wired LAN technologies
(Ethernet and 802.3)

Wired LAN Technologies

d Explosion of technologies and products during 1980s

d Consolidation during the 1990s

d Consolidation during the 1990s
d Currently: one de facto wired LAN standard
Ethernet

Ethernet Technology
d Invented at Xerox PARC in 1973
d Standardized by Digital, Intel, and Xerox (DIX) in 1978
d Frame has a 14-byte header followed by payload of 46 to
1500 bytes
d Frame format and addressing have survived virtually
unchanged
header 46 - 1500 bytes of payload
6-byte
destination address
6-byte
source address
2-byte
type header details
4-byte CRC

Ethernet Address Filtering
d Recall: station accepts a copy of the frame if destination
address matches
– The station’s unicast address
– The broadcast address (all 1s)
– A multicast address to which station is listening
d Other frames are ignored
d Promiscuous mode allows a station to receive all frames
regardless of address
– Basis of protocol analyzer software such as Wireshark

Question
If one is looking at the bits of an Ethernet frame as the frame is
transmitted across a wire, which bit specifies whether the frame
has been sent to a unicast destination address? Hint: look at the
48-bit MAC address format, the Ethernet header format and the
byte and bit ordering (Module 2).

Frame Type Field
d 2-octet field in frame header
d Set by sender to identify contents of frame
d Used by receiver to determine how to process the frame
d Values are standardized
d Examples:
– Type 0x0800 used for IPv4 datagram
– Type 0x86DD used for IPv6 datagram
– Type 0x0806 used for ARP

Illustration Of Frame Demultiplexing
IPv4
module
IPv6
module
0800 86DD
frame
arrives
demultiplexing
d Performed when frame arrives
d Usually handled by protocol software
d Frame type field examined and frame passed to appropriate
protocol module; unrecognized types are discarded

IEEE’s Version Of Ethernet
d Standardized in 1983 as IEEE standard 802.3
d Not widely adopted
d Header type field reinterpreted as a frame length
d Eight bytes of payload occupied by LLC / SNAP header
header new
hdr. 46 - 1492 bytes of payload
48-bit destination
address
48-bit source
address
16-bit
length
IEEE LLC / SNAP Header
24-bit
LLC
24-bit
OUI
16-bit
type
4-byte CRC

Ethernet Wiring
d Evolved through three generations
– Thicknet
– Thinnet
– Twisted pair
d Illustrate a range of possible network wiring schemes

Illustration Of Thicknet Wiring
thick Ethernet cable
computer with NIC
AUI cable
transceiver
terminator
d Heavy coaxial cable typically in the ceiling
d Each computer attached to the cable

Illustration Of Thinnet Wiring
computer with NIC
terminator
Thinnet cable
d Flexible coaxial cable
d Connections run point-to-point among computers
d Disadvantage: user can disconnect the network

Illustration Of Twisted Pair Ethernet Wiring
computer with NIC
hub
twisted pair wiring
d Unshielded or shielded twisted pairs using RJ45 connectors
d Multiple pairs allows full-duplex operation
d Each computer connects to central hub
d Topology is physical star, but logical bus
d Hub is known as “bus in a box”

Evolution Of Twisted Pair Ethernet Technologies
d Several variants of twisted pair Ethernet have been created
d Variants differ in data rate and wiring required
2
2222222222222222222222222222222222222222222222222222222
Designation Name Data Rate Cable Used
2
2222222222222222222222222222222222222222222222222222222
Twisted Pair
10BaseT
Ethernet
10 Mbps Category 5
2
2222222222222222222222222222222222222222222222222222222
Fast
100BaseT
Ethernet
100 Mbps Category 5E
2
2222222222222222222222222222222222222222222222222222222
Gigabit
1000BaseT
Ethernet
1 Gbps Category 6
2
2222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Wireless Networking
Technologies

Wireless Networks
d Many types exist
d Technologies differ in
– Distance spanned
– Data rates
– Physical characteristics of electromagnetic energy
* Ability to permeate obstructions like walls
* Susceptibility to interference
– Isolated channel vs. shared channel

A Taxonomy Of Wireless Networks
d We use a basic taxonomy to help classify wireless
technologies
Local Area
Networks (LANs)
Metropolitan Area
Networks (MANs)
Wide Area
Networks (WANs)
Personal Area
Networks (PANs)
Wireless Networks
d Note: the terminology is qualitative because some
technologies span multiple categories

Personal Area Network (PAN)
d Terminology used primarily with wireless networks
d Spans short distance
d Dedicated to a single user (not shared)

Personal Area Network (PAN)
d Terminology used primarily with wireless networks
d Spans short distance
d Dedicated to a single user (not shared)
d Example PAN technologies
2
22222222222222222222222222222222222222222222222222222222222222222222
Type Purpose
2
22222222222222222222222222222222222222222222222222222222222222222222
Communication over a short distance between a
Bluetooth small peripheral device such as a headset or mouse
and a system such as a cell phone or a computer
2
22222222222222222222222222222222222222222222222222222222222222222222
Line-of-sight communication between a small device,
InfraRed often a hand-held controller, and a nearby system such
as a computer or entertainment center
2
22222222222222222222222222222222222222222222222222222222222222222222
Communication over distances about as large as a
ZigBee residence, which allows electrical appliances to connect
to the Smart Grid
2
22222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

ISM Wireless Bands
d ISM stands for Industrial, Scientific, and Medical
d Region of the electromagnetic spectrum available for use
without license
d Used for wireless LANs and PANs (e.g., cordless phones)
d Three separate bands
902
MHz
928
MHz
2.4
GHz
2.484
GHz
5.725
GHz
5.850
GHz
26 MHz
bandwidth
83.6 MHz
bandwidth 125 MHz
bandwidth
Unlicensed does not mean unregulated.

Wireless LANs And Wi-Fi
d Variety of wireless LANs have been created
d Vendors moved to open standards in 1990s, with IEEE
providing most of the standards under 802.11
d In 1999, vendors formed Wi-Fi Alliance

Wireless LANs And Wi-Fi
d Variety of wireless LANs have been created
d Vendors moved to open standards in 1990s, with IEEE
providing most of the standards under 802.11
d In 1999, vendors formed Wi-Fi Alliance
d Example IEEE wireless standards
22222222222222222222222222222222222222222222222222222222222222222222222
IEEE Frequency Data Modulation Multiplexing
Standard Band Rate Technique Technique
22222222222222222222222222222222222222222222222222222222222222222222222
original
2.4 GHz 1 or 2 Mbps FSK DSSS
222222222222222222222222222222222222222222222222222222222222
802.11 2.4 GHz 1 or 2 Mbps FSK FHSS
222222222222222222222222222222222222222222222222222222222222
InfraRed 1 or 2 Mbps PPM – none –
22222222222222222222222222222222222222222222222222222222222222222222222
802.11b 2.4 GHz 5.5 and 11 Mbps PSK DSSS
22222222222222222222222222222222222222222222222222222222222222222222222
802.11g 2.4 GHz 22 and 54 Mbps various OFDM
22222222222222222222222222222222222222222222222222222222222222222222222
802.11n 2.4 GHz 54 to 600 Mbps various OFDM
22222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Spread Spectrum Transmission
d Uses multiple frequencies for a single channel
d Can increase performance or provide immunity to noise

Spread Spectrum Transmission
d Uses multiple frequencies for a single channel
d Can increase performance or provide immunity to noise
d Major spread spectrum techniques
2
2222222222222222222222222222222222222222222222222222222222222222222
2
2222222222222222222222222222222222222222222222222222222222222222222
Direct Similar to CDMA where a sender multiplies
DSSS Sequence the outgoing data by a sequence to form
Spread multiple frequencies and the receiver
Spectrum multiplies by the same sequence to decode
2
2222222222222222222222222222222222222222222222222222222222222222222
Frequency A sender uses a sequence of frequencies
FHSS Hopping to transmit data, and a receiver uses the
Spread same sequence of frequencies to extract
Spectrum data
2
2222222222222222222222222222222222222222222222222222222222222222222
Orthogonal A frequency division multiplexing scheme
OFDM Frequency where the transmission band is divided
Division into many carriers in such a way that
Multiplexing the carriers do not interfere
2
2222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

More IEEE Wireless LAN Standards
2
222222222222222222222222222222222222222222222222222222222222222222
Standard Purpose
2
222222222222222222222222222222222222222222222222222222222222222222
802.11e
Improved quality of service, such as a guarantee of
low jitter
2
222222222222222222222222222222222222222222222222222222222222222222
802.11h
Like 802.11a, but adds control of spectrum and power
(primarily intended for use in Europe)
2
222222222222222222222222222222222222222222222222222222222222222222
802.11i
Enhanced security, including Advanced Encryption
Standard; the full version is known as WPA2
2
222222222222222222222222222222222222222222222222222222222222222222
802.11k
Will provide radio resource management, including
transmission power
2
222222222222222222222222222222222222222222222222222222222222222222
802.11n
Data rate over 100 Mbps to handle multimedia (video)
applications (may be 500 Mbps)
2
222222222222222222222222222222222222222222222222222222222222222222
802.11p
Dedicated Short-Range Communication (DSRC) among
vehicles on a highway and vehicle-to-roadside
2
222222222222222222222222222222222222222222222222222222222222222222
802.11r
Improved ability to roam among access points without
losing connectivity
2
222222222222222222222222222222222222222222222222222222222222222222
802.11s
Proposed for a mesh network in which a set of nodes
automatically form a network and pass packets
2
222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Wireless LAN Architecture
d IEEE defines two possible modes for wireless LAN
communication

communication
d Infrastructure mode
– Wireless devices communicate through an access point
(AP)
– APs connect to each other and (usually) the Internet
– Typical uses: corporate wireless LAN, Internet cafe

communication
d Infrastructure mode
– Wireless devices communicate through an access point
(AP)
– APs connect to each other and (usually) the Internet
– Typical uses: corporate wireless LAN, Internet cafe
d Ad hoc mode
– Direct communication among wireless devices
– Forwarding possible
– Seldom used

Illustration Of Infrastructure Mode Wireless LAN
d Basic Service Set (BSS) for an AP is defined as set of devices
that can hear the AP
d APs interconnect through wired network
BSS #1 BSS #2 BSS #3
AP AP AP
access point
interconnect such as a switch
wireless
computer
range of
access point

Practical Considerations And Association
d In practice BSSs can overlap (given wireless device can hear
more than one AP)
BSS #1 BSS #2 BSS #3
switch
AP AP AP
router
to Internet
computer
in range
of two APs
d To solve the problem each device associates with one AP at
any time

Practical Considerations: Wi-Fi Channels

d 11 channels defined for North America in 2.4 GHz range

d Bad news: 22 MHz bandwidth means channels overlap
1: 2.412
2: 2.417
3: 2.422
4: 2.427
5: 2.432
6: 2.437
7: 2.442
8: 2.447
9: 2.452
10: 2.457
11: 2.462
22 MHz

d Bad news: 22 MHz bandwidth means channels overlap
1: 2.412
2: 2.417
3: 2.422
4: 2.427
5: 2.432
6: 2.437
7: 2.442
8: 2.447
9: 2.452
10: 2.457
11: 2.462
22 MHz
d Good news: channels 1, 6, and 11 can operate simultaneously
with no interference

Addresses In 802.11 Frame Format
d 802.11 frame is not the same as an Ethernet frame
d Each 802.11 frame includes four MAC addresses
– Source (e.g., wireless device)
– Destination AP (associated AP)
– Router along the path to the Internet
– Extra address for ad hoc mode
CTL DUR Address 1
(destination)
Address 2
(source)
Address 3
(dest. 2) SEQ Address 4 Payload
(0 to 2312 bytes) CRC
AP or wireless
computer’s MAC
sender’s
MAC address
router’s
MAC address
used in
ad hoc mode

Coordination Among Access Points
d Coordinated approach
– Initial design
– Similar to cellular telephone
– APs communicate to achieve smooth handoff
d Uncoordinated approach
– Later alternative
– APs do not communicate
– Wireless device changes association when communication
with an AP lost
– Lower overall cost

CSMA/ CA Protocol (Review)
d Alternative to CSMA/CD used in wireless LANs
d Allows stations within range of communicating pair to know
when communication starts
d Requires exchange of Ready-To-Send (RTS) and Clear-To-
Send (CTS) messages
d Delay associated with each message to ensure protocol is
efficient and correct

CSMA/ CA Protocol Details
d SIFS — Short Inter-Frame Space of 10 µsec
d DIFS — Distributed Inter-Frame Space of 50 µsec
d Slot Time of 20 µsec
time time
DIFS
RTS
SIFS
CTS
SIFS
data
SIFS
ACK

Wireless MAN Technology (WiMax)
d WiMax standard, IEEE 802.16, provides two types
– Fixed (802.16-2004) — endpoint does not move
– Mobile (802.16e-2005) — endpoint moves
d Uses 2
2222222222222222222222222222222222222222222222222222
Access
– Last-mile alternative to DSL or cable modems
– High-speed interconnection for nomadic users
– Unified data and telecommunications access
– As a backup for a site’s Internet connection
2
2222222222222222222222222222222222222222222222222222
Interconnect
– Backhaul from Wi-Fi access points to a provider
– Private connections among sites of a company
– Connection between small and large ISPs
2
2222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Illustration Of WiMax Uses
d Fixed type of WiMax used for high-capacity backhaul
requires Line-Of-Sight (LOS)
service
provider
LOS
backhaul
NLOS access
Wi-Fi region
wired connection

Standards For Wireless PANs
d Used in industrial as well as consumer products
d Remote control protocols optimized for short commands
(do not need high data rate)
2
222222222222222222222222222222222222222222222222222222222222222222222
Standard Purpose
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.1a Bluetooth technology (1 Mbps; 2.4 GHz)
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.2 Coexistence among PANs (noninterference)
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.3 High rate PAN (55 Mbps; 2.4 GHz)
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.3a Ultra Wideband (UWB) high rate PAN (110 Mbps; 2.4 GHz)
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.4 ZigBee technology – low data rate PAN for remote control
2
222222222222222222222222222222222222222222222222222222222222222222222
802.15.4a Alternative low data rate PAN that uses low power
2
222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Other Short-Distance Wireless Technologies

d Infrared Data Association (IrDA)
– Family of standards (data rate of 2.4 Kbps to 16 Mbps)
– Range of several meters
– Directional transmission with cone covering 30 degrees
– Generally low power consumption

d Infrared Data Association (IrDA)
– Family of standards (data rate of 2.4 Kbps to 16 Mbps)
– Range of several meters
– Directional transmission with cone covering 30 degrees
– Generally low power consumption
d Radio Frequency IDentification (RFID) tags
– Over 140 RFID standards exist
– Passive RFID tags draw power from reader’s signal
– Active RFID tags contain a multi-year battery
– Frequencies from less than 100 MHz to 868-954 MHz

Wireless WAN Technologies
d Cellular communication systems
d Satellite communication systems

Cellular Telephones And Data Networking
d There are more cell phones in the world than computers
d The smart phone is now the network interface of choice in
emerging countries
d Cell phone providers have switched to the Internet protocols

Cellular Telephones And Data Networking
d There are more cell phones in the world than computers
d The smart phone is now the network interface of choice in
emerging countries
d Cell phone providers have switched to the Internet protocols
all

Current Cellular System Architecture
d Cell has a tower that connects to mobile switching system
d Each mobile switching system connects to PSTN or Internet
. . . . . .
Public Switched Telephone Network plus a connection to the Internet
wired connection
Mobile
Switching
Centers cell
d Handoff decision made by infrastructure

Theoretical And Actual Cells
theoretical

Theoretical And Actual Cells
actual
theoretical
d Problems include: overlap and gaps

Cell Size And Expected Cell Phone Density
d Textbook diagrams show equal-size cells
d In practice, cell size related to expected number of cell
phones
d Smaller cells used in high-population areas
d Larger cells used in rural areas

Frequency Assignment

d Goal: minimize interference

d Principle
Interference can be minimized if an adjacent pair of cells do not
use the same frequency.

d Principle
d Method: devise an assignment of frequencies such that two
adjacent cells are not assigned the same frequency

d Principle
d Technique: create a pattern that can be repeated

d Principle
d Technique: create a pattern that can be repeated
d Known as a cluster approach

Example Clusters That Are Used
3-cell 4-cell 7-cell 12-cell
d Each cell in cluster assigned a unique frequency
d When replicated, clusters cover 2-dimensional surface
d Mathematically, the concept is tiling the plane

Illustration Of Cluster Replication
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
A
B
C
D
E
F
G
d No pair of adjacent cells assigned the same frequency

Four Generations Of Cellular Networks
d 1G used analog (1970s - 1980s)
d 2G and 2.5G use digital signals for voice (1990s-)
d 3G and 3.5G also include data transfer at rates of 400 Kbps
through 2 Mbps (2000s-)
d 4G offers higher data rates and support for real-time
multimedia such as television (2008-)

Cellular Technologies
d Many competing standards
d European Conference Of Postal and Telecommunications
Administrators chose a TDMA technology known as Global
System for Mobile Communications (GSM) for Europe
d In US, each carrier created its own standards
– Motorola created iDEN using TDMA
– Others adopted IS-95A, which uses CDMA
d Japan chose PDC, which uses TDMA

Summary Of 2G Wireless Standards
2
2222222222222222222222222222222222222222222
Approach Standard Generation
2
2222222222222222222222222222222222222222222
GSM 2G
2
2222222222222222222222222222222
GPRS 2.5G
2
2222222222222222222222222222222
GSM EDGE (EGPRS) 2.5G
2
2222222222222222222222222222222
EDGE Evolution 2.5G
2
2222222222222222222222222222222
HSCSD 2.5G
2
2222222222222222222222222222222222222222222
CDMA
IS-95A 2G
2
2222222222222222222222222222222
IS-95B 2.5G
2
2222222222222222222222222222222222222222222
TDMA
iDEN 2G
2
2222222222222222222222222222222
IS-136 2G
2
2222222222222222222222222222222
PDC 2G
2
2222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Note: 2.5G standards extend 2G standards by adding some
features of 3G

Third Generation Standards
d 2G standards were consolidated and extended:
222222222222222222222222222222222222222222222222222222
Approach Standard Successor To
222222222222222222222222222222222222222222222222222222
WCDMA
UMTS IS-136, IS-95A, EDGE, PDC
222222222222
HSDPA UMTS
222222222222222222222222222222222222222222222222222222
1xRTT IS-95B
2222222222222222222222222222222222222222
CDMA 2000 EVDO 1xRTT
2222222222222222222222222222222222222222
EVDV 1xRTT
222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d EVDO and EVDV data transfer standards evolved at
approximately the same time to deliver data at 2.4 Mbps or
3.1 Mbps
d HSDPA can achieve 14 Mbps

Fourth Generation Standards
d Initially, the ITU insisted on high performance before using
the term 4G
d Eventually, the ITU allowed intermediate technologies “to be
advertised” as 4G
2222222222222222222222222222222222222222222222222222222222222222
Classification Standard
2222222222222222222222222222222222222222222222222222222222222222
Can be advertised as 4G HSPA+, HTC Evo 4G, LTE, WiMAX
2222222222222222222222222222222222222222222222222222222222222222
Adheres to IMT-Advanced LTE Advanced, WiMAX Advanced
2222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
11
1
1
1
1
1
11
1
1
1
1
1

Review Of Satellite Types
d Low Earth Orbit (LEO)
– Appears to move across the sky
– Requires a cluster of 66 satellites to cover the earth
surface
d Medium Earth Orbit (MEO)
– Covers the poles
– Seldom used for general communication
d Geostationary Earth Orbit (GEO)
– Appears to remain stationary in the sky
– Requires only three satellites to cover the earth’s surface

GEO Coverage Of The Earth’s surface
d In the best case, only three satellites needed
EARTH
satellites
satellite
coverage
(footprint)
d Surface area covered known as footprint
d Ratio of distance to earth’s diameter approximately to scale

VSAT Satellite Technology
d Stands for Very Small Aperture Terminal
d Parabolic antenna focuses incoming signal
cross section
of dish antenna
receiver
incoming energy
d Example use: connect a company’s retail stores

Frequency Bands Used With VSAT Technology
d Multiple bands available
d Each band has disadvantages
222222222222222222222222222222222222222222222222222222222222222222222
Band Frequency Footprint Signal Strength Effect Of Rain
222222222222222222222222222222222222222222222222222222222222222222222
C Band 3 - 7 GHz Large Low Medium
222222222222222222222222222222222222222222222222222222222222222222222
Ku 10 - 18 GHz Medium Medium Moderate
222222222222222222222222222222222222222222222222222222222222222222222
Ka 18 - 31 GHz Small High Severe
222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1

Global Positioning System (GPS)

d 24 satellites

d 24 satellites
d Arranged in 6 orbital planes

d 24 satellites
d Civilian version has accuracy between 20 and 2 meters

d 24 satellites
d Relevance to data networking

d 24 satellites
d Relevance to data networking
– Provides accurate time
– Can be used to synchronize remote points in a data
network (needed by some protocols)

Software Defined Radio
d Also known as a software programmable radio
d New approach emerging from research
d Exciting possibilities
d Replaces fixed radio components with mechanism that can be
controlled by a programmable processor
d Can make better use of spectrum
d Potential downside: user might choose parameters that
interfere with police or emergency vehicles

Features Controlled In A Software Radio
2
2222222222222222222222222222222222222222222222222222222222222222222222
Feature Description
2
2222222222222222222222222222222222222222222222222222222222222222222222
Frequency The exact set of frequencies used at a given time
2
2222222222222222222222222222222222222222222222222222222222222222222222
Power The amount of power the transmitter emits
2
2222222222222222222222222222222222222222222222222222222222222222222222
Modulation The signal and channel coding and modulation
2
2222222222222222222222222222222222222222222222222222222222222222222222
Multiplexing Any combination of CDMA, TDMA, FDMA and others
2
2222222222222222222222222222222222222222222222222222222222222222222222
Signal Direction Antennas can be tuned for a specific direction
2
2222222222222222222222222222222222222222222222222222222222222222222222
MAC Protocol All aspects of framing and MAC addressing
2
2222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Enabling technologies
– Tunable analog filters to select frequencies and control
power
– Multiple antenna management to select direction

Multiple Antenna Management
d Needed because
– No single antenna handles all frequencies
– Directional signals important in focusing communication
d Multiple-Input Multiple-Output (MIMO) technology can aim
transmission or reception

Network Design Tradeoffs

d Network technology engineered for
– Maximum data rate
– Cost

– Cost
d LAN technologies maximize data rate and minimize cost

– Cost
d LAN technologies maximize data rate and minimize cost
d General principle
A maximum length specification is a fundamental part of LAN
technology; LAN hardware will not work correctly over wires
that exceed the bound.

Technologies That Extend LANs
d Variety of techniques have been invented to extend LANs
d Three key extension technologies
– Fiber modems
– Repeaters
– Bridges

Fiber Modems
d Communicate over an optical fiber
d Can span long distance
d Provide standard network interface (e.g., Ethernet)
d Can be used to extend connection between computer and
network
d Illustration of an extended network connection
Ethernet hub
or switch
connection
from computer
fiber modem fiber modem
optical fiber

Repeaters
d Operate at layer 1 (do not understand packets)
d Repeat and amplify signals
d Low cost
d Example use: extended infrared sensor on a cable box
repeater
remote
sensor
Cable box
connection to
cable box extended connection
d Disadvantage: amplifies and repeats noise

Switches and
Switched Networks

Bridge
d Originally sold as stand-alone device to extend two LAN
segments
d Operates at layer 2
d Can connect two or more segments
d Listens in promiscuous mode on each segment and sends
copy of each frame to other segments
d Does not copy noise, collisions, or frames that are incorrectly
formed
d Makes connected segments appear to be a single, large LAN
d Uses source MAC address in frames to learn computer
locations automatically, and uses destination MAC address to
filter frames

Illustration Of A Bridge Learning
A B C X Y Z
hub 1 hub 2
bridge
LAN segment
Event Segment 1 Segment 2 Frame Sent

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
Bridge boots – – –

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
A sends to B A – Both Segments

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
B sends to A A, B – Segment 1 only

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
X broadcasts A, B X Both Segments

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
Y sends to A A, B X, Y Both Segments

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
Y sends to X A, B X, Y Segment 2 only

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
C sends to Z A, B, C X, Y Both Segments

A B C X Y Z
hub 1 hub 2
bridge
LAN segment
C sends to Z A, B, C X, Y Both Segments
Z sends to X A, B, C X, Y, Z Segment 2 only

General Principle
Because a bridge permits simultaneous activity on attached
segments, a pair of computers on one segment can communicate
at the same time as a pair of computers on another segment.
d Each segment forms a separate collision domain

A Problem With Bridges

d A bridge always forwards broadcast and multicast frames

d Consider four bridges used to connect four LAN segments in
a loop
hub 1 hub 2 hub 3 hub 4
Bridge 1 Bridge 2 Bridge 3
Bridge 4

a loop
Bridge 4
d What happens if a computer attached to one of the segments
sends a broadcast frame?

a loop
Bridge 4
d What happens if a computer attached to one of the segments
sends a broadcast frame?
Copies of the frame cycle around the bridges forever!

Distributed Spanning Tree
d Prevents a packet from circulating around a cycle of bridges
d Initial protocol developed by Perlman at Digital Equipment
Corporation
d Executed by each bridge when the bridge boots
d Allows bridges to break a forwarding cycle
d Name Spanning Tree Protocol (STP) applies to basic
protocol
d Many variants have been created with extended names

How STP Works
d Executed at startup
d Distributed algorithm
– Each bridge runs it independently
– No central coordination
d Algorithm guaranteed to converge quickly
d No data packets forwarded until STP finishes

Steps Taken By STP
d Bridges exchange a series of STP messages (frames) that are
used to
– Elect a root bridge
– Select a shortest path to the root
d Each bridge disables forwarding broadcast or multicast
except along the selected path
d Result is a tree

Bridging Is Alive And Well
d Stand-alone bridge devices are seldom used
d Bridge technology is now incorporated into other devices
– DSL modems
– Cable modems
– Wi-Fi “repeaters”
– Satellite systems

Layer 2 Switch
d Physically similar to a layer 2 hub
– Network device
– Connects multiple computers
– Computers appear to be attached to a LAN segment
d Logically similar to a set of bridged networks
– Switch understands packets, not just signals
– No contention, and no need for CSMA / CD
– Ports operate in parallel
– Switch can include services that examine packets

Logical Function Of A Switch
switch
computers
simulated
bridge
simulated Ethernet segment
port on
the switch
d Switch offers same advantage as bridged networks: multiple
transfers can occur simultaneously

Actual Switch Architecture
switch
computers
fabric
(interconnect)
interfaces
d Switching fabric used for high throughput

Thought Problem
Suppose a computer is unplugged from a port on a Layer 2
switch and plugged into another port. Suppose the computer
does not send any packets. Will the computer continue to
receive unicast frames that are sent to it? Why or why not?

Virtual Local Area Network (VLAN) Switch
d Physically
– Similar to a conventional Layer 2 switch
– Has ports to which computer can connect
d Logically
– Manager can configure one or more broadcast domains
– Each port assigned to one broadcast domain
d Frame sent to broadcast or multicast address only propagated
to ports in the same broadcast domain

Networking Technologies:
Past And Present

A Wide Variety of Networking Technologies

d LAN technologies
– Token ring (esp., IBM Token Ring)
– FDDI/ CDDI

d LAN technologies
– Token ring (esp., IBM Token Ring)
– FDDI/ CDDI
d WAN technologies
– X.25
– Frame Relay
– ATM
– ISDN
– MPLS
d See Chapter 19 for a longer list

Asynchronous Transfer Mode (ATM)
d Created by phone companies in 1990s
d Intended as replacement for the Internet
d Paradigm was connection-oriented
d Used small cells (53 octets)
d Network guaranteed per-connection Quality of Service (QoS)
– Throughput
– Bound on delay
– Bound on jitter

Asynchronous Transfer Mode (ATM)
(continued)
d QoS in ATM
– Specified for each transfer (i.e., each TCP connection)
– Required setup time
– Meant each switch maintained state
– Was difficult/impossible to enforce at high speed
d Despite the failure of ATM, proponents still argue that
Internet needs QoS

Summary
d Packet switching divides data into small packets
d Each packet (frame) specifies destination
d Access technologies are used in the last mile
d Media access can be controlled, random, or channelized
d IEEE specifies Local Area Network standards
d Topologies used with LANs: bus, star, ring, and mesh
d Ethernet is the de facto standard for wired LANs
d Current Ethernets use twisted pair wiring

Summary
d Wireless networks include PANs, LANs, and WANs,
d Cellular telephones are using packet technology
d Satellite can deliver data through a dish antenna
d Software-defined radio adds flexibility to wireless devices
d LAN extensions include repeaters and bridges
d Once stand-alone devices, bridges are now incorporated into
other devices
d Layer 2 switch acts like bridged networks

MODULE V
Internetworking:
Concepts, Addressing, Architecture,
Protocols, Datagram Processing,
Transport-Layer Protocols, And
End-To-End Services

Topics
d Internet concept and architecture
d Internet addressing
d Internet Protocol packets (datagrams)
d Datagram forwarding
d Address resolution
d Error reporting mechanism
d Configuration
d Network address translation

Topics
(continued)
d Transport layer protocol characteristics and techniques
d Message transport with the User Datagram Protocol (UDP)
d Stream transport with the Transmission Control Protocol
(TCP)
d Routing algorithms and protocols
d Internet multicast and multicast routing

Internet Concept
And Internet Architecture

What Is The Internet?

d Users see it as services and applications
– Web and e-commerce
– Email, texting, instant messenger
– Social networking and blogs
– Music and video download (and upload)
– Voice and video teleconferencing

d Users see it as services and applications
– Web and e-commerce
– Email, texting, instant messenger
– Social networking and blogs
– Music and video download (and upload)
– Voice and video teleconferencing
d Networking professionals see it as infrastructure
– Platform on which above services run
– Grows rapidly

Growth Of The Internet
1981 1985 1990 1995 2000 2005 2010
0M
100M
200M
300M
400M
500M
600M
700M
800M
900M
1000M
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
......
......
......
......
........
..........
.........
......
............
..........
............
..........
.........
.........
d Plot shows number of computers on the Internet each year

Growth Of The Internet (log scale)
1981 1985 1990 1995 2000 2005 2010
102
103
104
105
106
107
108
109
1010
. . . . .
.......
......
......
. . . . .
.......
..........
........
........
......
......
......
......
.......
......
......
......
......
......
......
......
. . . . .
......
......
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
d Plot shows number of computers on the Internet each year

Actual Size Of The Internet
d Previous plots are somewhat misleading
– Derived by walking the Domain Name System
– Only report hosts with IP addresses
d Since around 2000, many Internet devices
– Do not have a fixed IP address
– Connect behind a NAT box (e.g., wireless router)
d Actual size is difficult to measure

Internet Architecture And Design

d If one were to design a global communication system from
scratch
– How should it be organized?
– Which technology or technologies should be used?

d If one were to design a global communication system from
scratch
– How should it be organized?
– Which technology or technologies should be used?
d The challenges
– Which applications should it support?
– Which network technologies should it use
* PANs / LANs / MANs / WANs
* Wired / wireless
* Terrestrial / satellite

(continued)
d Key principles
– Internet is designed to accommodate extant services plus
new services that will be invented
– Internet is designed to accommodate any network
technology, allowing each technology to be used where
appropriate

Internet Philosophy
d Infrastructure
– Provides a packet communication service
– Treats all attached endpoints as equal (any endpoint can
send a packet to any other endpoint)
– Does not restrict or dictate packet contents
– Does not restrict or dictate underlying network
technologies
d Attached endpoints
– Run applications that use the network to communicate
with applications on other endpoints
– Control all content and provide all services

Advantages Of The Internet Philosophy
d Accommodates heterogeneous underlying networks
d Accommodates arbitrary applications and services
d Separates communication from services

Internet
d Follows a network of networks approach
d Allows arbitrary networks to be included
d Uses IP routers to interconnect individual networks
d Permits each router to connect two or more networks
routers
networks

Internet Architecture: Logical View
host
computers
d Computers attached to Internet known as host computers
d To a host, Internet appears to be one giant network

Internet Architecture: Physical View
net 2
net 4
net 5
net 3
net 1
router
physical net
host
computers
d Network of heterogeneous networks connected by routers
d Each host attaches to a network

Before We Discuss
Internet Addressing

The Situation
d Internet addressing is defined by the Internet Protocol (IP)
d IP is changing
– Current version is 4 (IPv4)
– New version is 6 (IPv6)

History Of The Internet Protocol
d IP separated from TCP in 1978
d Version 1-3 discarded quickly; version 4 was the first
version used by researchers
d By early 1990s, a movement started that clamored for a new
version of IP because the 32-bit address space would run out
“soon”
d In 1993, the IETF received proposals, and formed a working
group to find a compromise
d By 1995, a new version had been proposed and documents
written

Background Of The New Version Of IP
d Various groups offered opinions about the features

– Cable companies wanted support for broadcast delivery

– Telephone companies argued that everyone would soon
be using a connection-oriented network technology
(ATM)

(ATM)
– Several groups wanted mobility

(ATM)
– The military pushed for better security

(ATM)
– The military pushed for better security
d A compromise was reached: IP version 6 includes all the
above

The Uphill Battle To Change IPv4
d IP is difficult to change because
– IP lies at the heart of the Internet protocols
– Version 4 of IP has a proven track record
The success of the current version of IP is incredible — the
protocol has accommodated changes in hardware technologies,
heterogeneous networks, and extremely large scale.

The Hourglass Model
Appl1 Appl2 Appln
Net1 Net2 Netm
IP
. . .
. . .
d IP lies in the middle — changing it means changing all
hosts and routers in the Internet

Our Approach
d In the current Internet, both IPv4 and IPv6 are relevant and
important
d Throughout the course, we will
– Discuss general concepts
– See how IPv4 and IPv6 implement the concepts

Addressing In The Internet

d Can we use MAC addresses across an internet?

d No: heterogeneity means
– Multiple types of MAC addresses
– MAC address meaningful on one network not
meaningful on another

d No: heterogeneity means
– Multiple types of MAC addresses
– MAC address meaningful on one network not
meaningful on another
d Solution
– Create new addressing scheme that is independent of
MAC addresses

The Two Forms Of Addresses
d Identity
– Unique number assigned to each endpoint
– Analogous to Ethernet address
d Locator
– Endpoint address encodes location information, such as
* Geographic location
* Location relative to a service provider
* Computer on a given physical network

Two Principles To Keep In Mind
Both identify and locator forms have advantages in
some situations; no form is best in all cases
Addressing is inherently linked to routing; the
choice of an addressing scheme affects the cost of
computing and maintaining routes

The IPv4 Addressing Scheme
d Unique number is assigned to each Internet host
d 32-bit binary value known as IPv4 address
d Virtual address, not derived from MAC address
d Divided into two parts
– Prefix identifies physical network (locator)
– Suffix identifies a host on the network (identity)

Dotted Decimal Notation (IPv4)
d Convenient for humans
d Divides IPv4 address into octets of eight bits each
d Represents each octet in decimal separated by dots

Dotted Decimal Notation (IPv4)
d Convenient for humans
d Divides IPv4 address into octets of eight bits each
d Represents each octet in decimal separated by dots
d Examples
22222222222222222222222222222222222222222222222222222222222222222222222222
32-bi t Binary Number Equivalent Dot ted Decimal
22222222222222222222222222222222222222222222222222222222222222222222222222
10000001 00110100 00000110 00000000 129 . 52 . 6 . 0
22222222222222222222222222222222222222222222222222222222222222222222222222
11000000 00000101 00110000 00000011 192 . 5 . 48 . 3
22222222222222222222222222222222222222222222222222222222222222222222222222
00001010 00000010 00000000 00100101 10 . 2 . 0 . 37
22222222222222222222222222222222222222222222222222222222222222222222222222
10000000 00001010 00000010 00000011 128 . 10 . 2 . 3
22222222222222222222222222222222222222222222222222222222222222222222222222
10000000 10000000 11111111 00000000 128 . 128 . 255 . 0
2222222222222222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1

Division Between Prefix And Suffix
d Original scheme (classful addressing)
– Each address divided on octet (8-bit) boundary
– Division could be computed from the address
d Current scheme (classless addressing)
– Formal name Classless Inter-Domain Routing (CIDR)
– Division permitted at arbitrary bit position
– Boundary must be specified external to the address

Classful Addressing
d Now historic
d Explains IPv4 multicast range
0 prefix suffix
1 0 prefix suffix
1 1 0 prefix suffix
1 1 1 0 multicast address
1 1 1 1 reserved (not assigned)
Class A
Class B
Class C
Class D
Class E
0 1 2 3 4 8 16 24 31
Bits

Address Mask
d Required with classless addressing
d Associated with a network
d Specifies division of addresses into network prefix and host
suffix for that network
d 32-bit binary value
– 1-bits correspond to prefix
– 0-bits correspond to suffix
d Example mask that specifies six bits of prefix
11111100 00000000 00000000 00000000

CIDR Notation
d Used by humans to enter address mask
d Avoids dotted decimal errors
d Follows address with slash and integer X, where X is the
number of prefix bits
d Example
– In dotted decimal, a 26-bit mask is
255 . 255 . 255 . 192
– CIDR merely writes
/26

Table Of CIDR And Dotted Decimal Equivalences
Length (CIDR) Address Mask Notes
/ 0 0 0 0 0
. . .
/ 1 128 0 0 0
. . .
/ 2 192 0 0 0
. . .
/ 3 224 0 0 0
. . .
/ 4 240 0 0 0
. . .
/ 5 248 0 0 0
. . .
/ 6 252 0 0 0
. . .
/ 7 254 0 0 0
. . .
/ 8 255 0 0 0
. . .
/ 9 255 128 0 0
. . .
/ 10 255 192 0 0
. . .
/ 11 255 224 0 0
. . .
/ 12 255 240 0 0
. . .
/ 13 255 248 0 0
. . .
/ 14 255 252 0 0
. . .
/ 15 255 254 0 0
. . .
/ 16 255 255 0 0
. . .
All 0s (equivalent to no mask)
1-octet boundary
2-octet boundary

Table Of CIDR And Dotted Decimal Equivalences
Length (CIDR) Address Mask Notes
/ 17 255 255 128 0
. . .
/ 18 255 255 192 0
. . .
/ 19 255 255 224 0
. . .
/ 20 255 255 240 0
. . .
/ 21 255 255 248 0
. . .
/ 22 255 255 252 0
. . .
/ 23 255 255 254 0
. . .
/ 24 255 255 255 0
. . .
/ 25 255 255 255 128
. . .
/ 26 255 255 255 192
. . .
/ 27 255 255 255 224
. . .
/ 28 255 255 255 240
. . .
/ 29 255 255 255 248
. . .
/ 30 255 255 255 252
. . .
/ 31 255 255 255 254
. . .
/ 32 255 255 255 255
. . .
3-octet boundary
All 1s (host specific mask)

Why CIDR Is Useful
d ISPs assign IP addresses
d Corporate customer with N computers needs N addresses
d CIDR permits ISP to round to nearest power of two
d Example
– Assume ISP owns address block 128.211.0.0/ 16
– Customer has 12 computers
– ISP assigns 4 bits of suffix to customer
– Mask used is /28
– Example: customer is assigned 128.211.0.16/ 28
– Each computer at customer site has unique final 4 bits

Example Of A /28 Address Block
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 0 0 0 0 0 0 1 1 0 1 0 0 1 1 1
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 28 31
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0
0 28 31
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 0 0 0 0 0 0 1 1 0 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1
0 28 31
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1 0 0 0 0 0 0 0 1 1 0 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0
0 28 31
Network Prefix 128.211.0.16 / 28
Address Mask 255.255.255.240
Lowest Host Address 128.211.0.17
Highest Host Address 128.211.0.30

Special IPv4 Addresses
d Some address forms are reserved
22222222222222222222222222222222222222222222222222222222222222222222
Prefix Suffix Type Of Address Purpose
22222222222222222222222222222222222222222222222222222222222222222222
all-0s all-0s this computer used during bootstrap
22222222222222222222222222222222222222222222222222222222222222222222
network all-0s network identifies a network
22222222222222222222222222222222222222222222222222222222222222222222
network all-1s directed broadcast broadcast on specified net
22222222222222222222222222222222222222222222222222222222222222222222
all-1s all-1s limited broadcast broadcast on local net
22222222222222222222222222222222222222222222222222222222222222222222
127 / 8 any loopback testing
22222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
d Loopback address ( 127.0.0.1 ) used for testing
– Packets never leave the local host
d Addresses 240.0.0.0/ 8 and above are multicast

Host Address Count
d For a given network prefix, the all-0s and all-1s suffixes
have special meaning
d Consequence: if a suffix has N bits, 2
N
− 2 hosts can be
present

IP Addressing Principle
An IP address does not identify a specific computer. Instead,
each IP address identifies a connection between a computer and
a network.

IP Addressing Principle
An IP address does not identify a specific computer. Instead,
each IP address identifies a connection between a computer and
a network.
d Consequence
A router or a host with multiple network connections must be
assigned one IP address for each connection.
d Note: host with multiple network connections is called a
multi-homed host

Illustration Of IPv4 Address Assignment
Wired Ethernet 131.108.0.0 / 16
Wi-Fi Net
223.240.129.0 / 24
WAN 78.0.0.0 / 8
223.240.129.2
131.108.99.5
78.0.0.17
223.240.129.17
router 1
router 2
d Each network assigned a unique prefix
d Each host on a network assigned a unique suffix

IPv6 Host Addresses
d Like IPv4
– Binary value
– Divided into locator prefix and unique ID suffix
– Identifies a connection to a network
d Unlike IPv4
– 128 bits long
– Suffix can be derived from MAC address
– 3-level address hierarchy

The IPv6 3-Level Hierarchy
GLOBAL PREFIX SUBNET INTERFACE (COMPUTER)
K bits 64–K bits 64 bits
d Prefix size chosen by ISP
d Subnet area allows organization to have multiple networks

IPv6 Address Types
22222222222222222222222222222222222222222222222222222222222222222
Type Purpose
22222222222222222222222222222222222222222222222222222222222222222
unicast The address corresponds to a single computer. A
datagram sent to the address is routed along a
shortest path to the computer.
22222222222222222222222222222222222222222222222222222222222222222
multicast The address corresponds to a set of computers, and
membership in the set can change at any time. IPv6
delivers one copy of the datagram to each member of
the set.
22222222222222222222222222222222222222222222222222222222222222222
anycast The address corresponds to a set of computers that
share a common prefix. A datagram sent to the
address is delivered to exactly one of the computers
(e.g., the computer closest to the sender).
22222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Colon Hex Notation
d Syntactic form used by humans to enter addresses
d Replacement for IPv4’s dotted decimal
d Expresses groups of 16 bits in hexadecimal separated by
colons
d Example:
105.220.136.100.255.255.255.255.0.0.18.
128.140.10.255.255
becomes
69DC:8864:FFFF:FFFF:0:1280:8C0A:FFFF

Colon Compression
d Many IPv6 addresses contain long strings of zeroes
d Successive zeros can be replaced by two colons
d Example
FF0C:0:0:0:0:0:0:B1
can be written:
FF0C::B1

Two Major Reasons To Adopt IPv6

d More addresses

d More addresses
– Eventually, IPv4 addresses will be depleted

d More addresses
– IPv6 provides more addresses than we will ever need
340,282,366,920,938,463,463,374,607,431,768,211,456

d More addresses
340,282,366,920,938,463,463,374,607,431,768,211,456
– 1024 addresses per square meter of the Earth’s surface!

d More addresses
340,282,366,920,938,463,463,374,607,431,768,211,456
d Hype and excitement

d More addresses
340,282,366,920,938,463,463,374,607,431,768,211,456
– Researchers view IPv6 as an opportunity to be part of
the action

d More addresses
340,282,366,920,938,463,463,374,607,431,768,211,456
– Researchers view IPv6 as an opportunity to be part of
the action
– Industries view IPv6 as an opportunity for revenue
enhancement

IPv6 And Children

IPv6 And Children
Child Of Famous Parent Internet Protocol version 6
3333333333333333333333333333333333333333333333333333333333333333333333
d Greatness is anticipated and
expectations run high
d Child’s success is often
compared to the parent’s
d To achieve acclaim, the child
must outperform the parent
d We say that the child grows up
“in the shadow” of the parent

IPv6 And Children
3333333333333333333333333333333333333333333333333333333333333333333333
d Bad news: guiding genetic
principle is known as the
“tendency toward the mean”

IPv6 And Children
3333333333333333333333333333333333333333333333333333333333333333333333
3333333333333333333333333333333333333333333333333333333333333333333333
d IPv6’s success is often
compared to IPv4’s
d To achieve acclaim, IPv6
must outperform IPv4
d IPv6 has been growing up
“in the shadow” of IPv4

IPv6 And Children
3333333333333333333333333333333333333333333333333333333333333333333333
3333333333333333333333333333333333333333333333333333333333333333333333
d IPv6’s success is often
compared to IPv4’s
d To achieve acclaim, IPv6
must outperform IPv4
d IPv6 has been growing up
“in the shadow” of IPv4
d Bad news: guiding engineering
“second-system syndrome”

Internet Protocol Packets
(IP datagrams)

Internet Packets
Because it includes incompatible networks, the Internet cannot
adopt a particular hardware packet format. To accommodate
heterogeneity, the Internet Protocol defines a hardware-
independent packet format.

IP Datagram
d Virtual packet format used in the Internet
d Same general layout as a network frame
Header Data Area (known as a payload area)
d Format of header determined by protocol version (IPv4 or
IPv6)
d Size of payload determined by application
– Maximum payload is almost 64K octets
– Typical datagram size is 1500 octets

IPv4 Datagram Header
d Most header fields have fixed size and position
d Header specifies source, destination, and content type
0 4 8 16 19 24 31
VERS H. LEN SERVICE TYPE TOTAL LENGTH
IDENTIFICATION FLAGS FRAGMENT OFFSET
TIME TO LIVE TYPE HEADER CHECKSUM
SOURCE IP ADDRESS
DESTINATION IP ADDRESS
IP OPTIONS (MAY BE OMITTED) PADDING
BEGINNING OF PAYLOAD (DATA BEING SENT)
.
.
.

A Few Details
d SOURCE IP ADDRESS field gives the IPv4 address of the
original source
d DESTINATION IP ADDRESS field gives the IPv4 address of
the ultimate destination
d Intermediate router addresses do not appear in header
d Header size
– Almost no Internet datagrams contain options
– Therefore header length is usually 20 octets

IPv6 Header Arrangement
d Multiple headers used: base plus zero or more extension(s)
Base
Header
Extension
Header 1
. . . Extension
Header N
Payload (Data)
optional
d The figure is not to scale: extension headers and/or the
payload can be much larger than the base header

IPv6 Base Header Format
0 4 12 16 24 31
VERS TRAFFIC CLASS FLOW LABEL
PAYLOAD LENGTH NEXT HEADER HOP LIMIT
SOURCE ADDRESS
DESTINATION ADDRESS
d Flow Label field allows datagram to be associated with a
flow

Identifying Headers
d Each header contains a NEXT HEADER field
d Value specifies the type of the next item
d Each layer 4 protocol (UDP, TCP, etc) is also assigned a
type

Example Use Of Next Header Field
d Illustration of headers when a datagram contains a base
header and transport protocol
Base Header
NEXT=TCP
TCP Data
d Illustration of headers when a datagram also contains an
optional route header
Base Header
NEXT=ROUTE
Route Header
NEXT=TCP
TCP Data

The Size Of An Extension Header
d Fixed length headers
– Size is specified in the standards document
– Protocol software contains size constant
d Variable length headers
– Size is determined by sender
– Header contains an explicit length field
0 8 16 31
NEXT HEADER HEADER LEN
ONE OR MORE OPTIONS

Consequences For Packet Processing
d Consider a host or router that receives an IPv6 datagram
d The datagram contains a set of extension headers
d Each extension header can contain an explicit length field
d To parse the datagram, IP software must iterate through
headers
d Conclusion: processing IPv6 can entail extra overhead

Internet Communication Paradigm
d Each datagram handled independently
d Datagram formed on source computer
d Source sends datagram to nearest router
d Router forwards datagram to next router along path to
destination
d Final router delivers datagram to destination
d Datagram passes across a single physical network at each
step

Datagram Forwarding
d Performed by initial host and each router along path
d Selects next hop for the datagram as either
– Next router along the path
– Ultimate destination
d Uses a forwarding table with one entry per network
d Important point: size of forwarding table proportional to
number of networks in the Internet

Forwarding Table Entry
d Uses IP addresses only (no MAC addresses)
d Contains
– Destination network IP prefix
– Address mask for the destination network
– IP address of next hop

Illustration Of An IPv4 Forwarding Table
30.0.0.7
40.0.0.7
40.0.0.8
128.1.0.8
128.1.0.9
192.4.10.9
30.0.0.0 / 8 40.0.0.0 / 8 128.1.0.0/16 192.4.10.0/24
Destination Mask Next Hop
30.0.0.0 255.0.0.0 40.0.0.7
40.0.0.0 255.0.0.0 deliver direct
128.1.0.0 255.255.0.0 deliver direct
192.4.10.0 255.255.255.0 128.1.0.9
(a)
(b)
router R1 router R2 router R3
d In practice, table usually contains a default entry

Prefix Extraction
d Forwarding paradigm
– Use network prefix when forwarding
– Use host when delivering
d Conceptual forwarding step
– Compare destination in each forwarding table entry with
datagram’s destination address, D
– During comparison, only examine network prefix
d Note: mask in forwarding table makes comparison efficient
if ( (Mask[i] & D) == Destination[i] ) forward to NextHop[i];

Longest Prefix Match
d Classless addressing means forwarding table entries can be
ambiguous
d Example: consider destination 128.10.2.3 and a table that
includes the following two entries:
128.10.0.0 / 16 next hop A
128.10.2.0 / 24 next hop B
d The destination matches both of them!
d Solution: select the match that has the longest prefix (in the
example, take next hop B)
d Known as longest prefix match

Datagram Encapsulation
d Needed because underlying network hardware does not
understand datagrams
d Entire datagram travels in payload area of frame
Frame Header Frame Payload
IP Header IP Payload
d Frame header contains MAC address of next hop
d Frame only used for trip across one network: when frame
arrives at next hop, datagram is extracted and frame is
discarded
d Datagram remains intact end-to-end

Illustration Of Encapsulation
Router 1
Router 2
Source host
Destination host
Net 1
Net 2
Net 3
datagram
datagram
datagram
datagram
datagram
datagram
datagram
Frame Hdr 1
Frame Hrd 2
Frame Hdr 3

Semantics Of Internet Communication
d IP uses best effort delivery semantics

d IP attempts to deliver each datagram, but specifies that a
datagram can be
– Lost
– Duplicated
– Delayed
– Delivered out-of-order
– Delivered with bits scrambled

datagram can be
– Lost
– Duplicated
– Delayed
d Motivation: accommodate any underlying network

datagram can be
– Lost
– Duplicated
– Delayed
d Motivation: accommodate any underlying network
d Note: in practice, IP works and it works well

MTU And Network Heterogeneity
d Each network technology specifies a Maximum Transfer
Unit (MTU) that is the largest amount of data that can be
sent in a packet
d Example: Ethernet MTU is 1500 octets
d Datagram can be as large as the network MTU
d Consider a 1500-octet datagram set from H1 to H2 in the
following network
R
H1 H2
Net 1 (MTU=1500) Net 2 (MTU=1000)
d Datagram can reach router R, but cannot traverse Net 2

Datagram Fragmentation
d Technique for accommodating heterogeneous MTUs
d Needed if datagram exceeds MTU
d Original datagram divided into smaller datagrams called
fragments
d Header of fragment derived from original datagram header
d Each fragment is forwarded independently
d IPv4 allows routers to perform fragmentation
d IPv6 requires sending host to perform fragmentation
d Important principle for both IPv4 and IPv6:
The ultimate destination reassembles fragments.

The General Idea Of Fragmentation
d Divide the payload into a series of datagrams
IP Header original datagram payload
IP Hdr 1 IP Hdr 2 IP Hdr 3
payload 1 pay. 3
payload 2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
d Note: the tail fragment may be smaller than the others

IPv4 Fragmentation Details
d Datagram header contains fixed fields that control
fragmentation
d A bit in FLAGS field specifies whether given datagram is a
fragment or complete datagram
d An additional FLAGS bit specifies whether the fragment
carries the tail of the original datagram
d OFFSET field specifies where the payload belongs in the
original datagram

IPv6 Fragmentation Details
d Always performed by the original source, never by routers
d Rule: no header changes are allowed as an IPv6 datagram
traverses the Internet
d Consequences
– Source must discover path MTU
– Separate extension header contains fragmentation
information (same items as IPv4)
d Fragmentable part of datagram may include some extension
headers

Illustration of IPv6 Fragmentation
Unfragmentable
Part
Fragmentable Part
(can include headers as well as data)
Unfragmentable
Part
Frag. 1
Header
P1
Unfragmentable
Part
Frag. 2
Header
P2
Unfragmentable
Part
Frag. 3
Header
P3
(a)
(b)
(c)
(d)
P1
P2
P3
d A datagram (a) divided into fragments (b through d)

Collecting Fragments
d Destination collects incoming fragments
d IDENTIFICATION field used to group related fragments
d OFFSET field allows receiver to recreate the original
payload
d LAST FRAGMENT bit allows receiver to know when all
fragments have arrived
d If a fragment fails to arrive within a timeout period, entire
datagram is discarded
d Note: if an IPv4 fragment is divided into subfragments,
reassembly does not require reassembling subfragments

Thought Problem
d Suppose
– A vendor sells a network security appliance that fits
between a computer and an Ethernet switch
– The appliance encrypts each IP datagram that the
computer sends
– Encryption adds only three bytes of extra data to the
payload
d Measurements show that throughput decreases dramatically
whenever the appliance is enabled
d Explain the lower throughput

Review Of Datagram Transmission

d Host or router has datagram to send

d IP uses longest-prefix match to look up datagram’s
destination address in forwarding table and obtains
– Network over which to send (in case there is more than
one network connection)

d IP encapsulates datagram in frame (entire datagram placed
in payload area of frame)

d Is the resulting frame ready to send to the next hop?

d Is the resulting frame ready to send to the next hop?
No!

Hardware And Protocol Addressing

d Underlying network hardware
– Only understands MAC addresses
– Requires each outgoing frame to contain the MAC
address of the next hop

d IP forwarding
– Deals only with (abstract) IP addresses
– Computes the IP address of the next hop

d IP forwarding
– Deals only with (abstract) IP addresses
– Computes the IP address of the next hop
d Conclusion
The IP address of the next hop must be translated to a MAC
address before a frame can be sent.

Address Resolution
d Translates IP address to equivalent MAC address that the
hardware understands
d IP address is said to be resolved
d Restricted to a single physical network at a time
d Example: consider computer X sending to computer Y
X
A
B
C
Y
D
R1 R2
d A MAC address is needed at each hop

An Example With MAC Addresses
X Y
R1 R2
128.10.0.0 /16 10.0.0.0 /8 192.168.0.0 /16
MAC: 3A-12-C9
IP: 128.10.0.1
MAC: 04-CF-47
IP: 192.168.0.1
MAC: 59-61-33
IP: 128.10.0.100
MAC: 97-27-D3
IP: 10.0.0.100
MAC: 8E-1A-7F
IP: 10.0.0.200
MAC: 54-DB-31
IP: 192.168.0.200
Sender NEXT-HOP SRC MAC DST MAC SRC IP DST IP
2
222222222222222222222222222222222222222222222222222222222222222222222222
X 128.10.0.100 3A-12-C9 59-61-33 128.10.0.1 192.168.0.1
R1 10.0.0.200 97-27-D3 8E-1A-7F 128.10.0.1 192.168.0.1
R2 192.168.0.1 54-DB-31 04-CF-47 128.10.0.1 192.168.0.1
d How can a host or router find the MAC address of the next
hop?

Address Resolution Protocol (ARP)
d Designed for IPv4 over Ethernet
d Used by two computers on the same physical network
d Allows a computer to find the MAC address of another
computer
d Operates at layer 2
d Uses network to exchange messages
d Computer seeking an address sends request to which another
replies

Example Of ARP Exchange
d Assume
– Four computers attached to an Ethernet
– Computer B has a datagram to send
d Computer B
– Uses forwarding table to find next-hop address IC
– Broadcasts an ARP request: “I’m looking for a computer
with IP address IC”
d Computer C
– Receives the request and replies; “I’m the computer with
IP address IC”

Illustration Of The ARP message Exchange
Ethernet switch Ethernet switch
W X Y Z W X Y Z
(a) (a)
d Request is broadcast to all computers
d Only the intended recipient replies
d Reply is sent unicast

ARP Message Format

ARP Message Format
d Sufficiently general to permit
– Arbitrary high-level protocol address
– Arbitrary hardware address

ARP Message Format
d In practice, only used with IP and 48-bit Ethernet addresses

ARP Message Format
d In practice, only used with IP and 48-bit Ethernet addresses
0 8 16 24 31
HARDWARE ADDRESS TYPE PROTOCOL ADDRESS TYPE
HADDR LEN PADDR LEN OPERATION
SENDER HADDR (first 4 octets)
SENDER HADDR (last 2 octets) SENDER PADDR (first 2 octets)
SENDER PADDR (last 2 octets) TARGET HADDR (first 2 octets)
TARGET HADDR (last 4 octets)
TARGET PADDR (all 4 octets)

ARP Encapsulation
d ARP message is placed in payload area of hardware frame
d When used with Ethernet, type is 0x0806
d Source and destination MAC addresses must be added to
frame header before sending
FRAME
HEADER FRAME PAYLOAD CRC
ARP MESSAGE

ARP Algorithm And Caching
Given:
An incoming ARP request or response
Purpose:
Process the message and update the ARP cache
Method:
Extract sender’s IP address, I, and MAC address, M
If ( address I is already in the ARP cache ) {
Replace corresponding MAC address with M;
}
if ( message is a request and target is “me” ) {
Add sender’s entry to the ARP cache providing
no entry exists;
Generate and send a response;
}
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2222222222222222222222222222222222222222222222222222222222

Boundary Between Protocol And MAC Addressing
d ARP isolates hardware addresses, allowing layers above to
use only IP
Network Interface
Physical
address resolution
Application
Transport
Internet
conceptual
boundary
IP addresses
used
MAC addresses
used

Thought Problem
d ARP is sometimes cited as a security weakness
d If someone gains access to a given network, how can they
exploit ARP to intercept packets?

Address Binding With IPv6
d IPv6 does not use ARP
d Instead, IPv6 defines a new address binding mechanism
known as IPv6 Neighbor Discovery (IPv6-ND)
d IPv6-ND
– Maintains a neighbor cache
– Keeps the cache up-to-date at all times
d IPv6-ND operation
– Sends a multicast request to find neighbors and populate
the cache
– Polls neighbors periodically, even if no datagrams are
being sent to the neighbor

IP Error Detection And Reporting
d Recall that IP allows datagrams to be
– Lost
– Duplicated
– Delayed

– Lost
– Duplicated
– Delayed
d Why is error reporting needed?

– Lost
– Duplicated
– Delayed
d Why is error reporting needed?
d Answer: best-effort does not mean “careless” — the design
is intended to tolerate errors in the underlying networks, not
to introduce them
d IP reports problems when they are detected

General Error Detection
d A variety of basic error detection mechanisms exist
d Examples
– Parity bits and other forward error codes can detect
transmission errors
– A CRC can detect an incorrect frame
– The IP header checksum can detect an incorrect
datagram header
– IP’s TTL (hop limit) can detect a routing loop
– A reassembly timer can detect lost fragments
d Only some types of errors can be reported

Internet Control Message Protocol (ICMP)
d Required and integral part of IP
d Reports errors back to the original source
d Uses IP to carry messages
d Defines many types of messages, each with a specific format
and contents
d Includes information messages as well as error reports
d ICMPv4 and ICMPv6 share many messages

Example ICMP Messages
22222222222222222222222222222222222222222222222222222222222222222222
Number Type Purpose
22222222222222222222222222222222222222222222222222222222222222222222
0 Echo Reply Used by the ping program
22222222222222222222222222222222222222222222222222222222222222222222
3 Dest. Unreachable Datagram could not be delivered
22222222222222222222222222222222222222222222222222222222222222222222
5 Redirect Host must change a route
22222222222222222222222222222222222222222222222222222222222222222222
8 Echo Request Used by the ping program
22222222222222222222222222222222222222222222222222222222222222222222
11 Time Exceeded TTL expired or fragments timed out
22222222222222222222222222222222222222222222222222222222222222222222
12 Parameter Problem IP header is incorrect
22222222222222222222222222222222222222222222222222222222222222222222
30 Traceroute Used by the traceroute program
22222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Most heavily-used ICMP messages are 8 and 0, which are
sent and received by the ping program

ICMP Encapsulation
d Two levels of encapsulation
– ICMP message encapsulated in an IP datagram
– IP datagram encapsulated in a network frame
ICMP Payload
ICMP Hdr

Example Of An ICMP Error Report
d Host S creates a datagram for destination D
d S sets the TTL to 255 and sends the datagram
d Datagram reaches a loop in the middle of the Internet
d Datagram circulates around the loop until the TTL reaches
zero
d Router that decrements the TTL to zero
– Sends a type 11 ICMP message to S
– Discards the datagram that caused the problem

Protocol Configuration
d Many items must be set before protocols can be used
– IP address of each network interface
– Address mask for each network
– Initial values in the forwarding table
d Process is known as protocol configuration
d Usually occurs when operating system boots
d Two basic approaches
– Manual
– Automatic

Manual Configuration
d Used for IP routers or host that has a permanent IP address
d Manager
– Enters configuration once
– Specifies that the configuration be saved in non-volatile
storage
– Interfaces include Command Line Interface (CLI) and
web
d OS
– Fetches values from non-volatile storage whenever the
device boots

Automatic Configuration
d Used primarily for hosts
d Initially created for diskless workstations
d Basic idea
– Use network to obtain configuration information
– Configure protocol software, and then start to run
applications
d A seeming paradox
Automatic configuration requires a computer to
be able to use a network before the computer’s
protocol parameters have been configured.

Ways To Solve The Paradox
d Use layer 2 protocols to obtain layer 3 parameters, then use
layer 3 to obtain higher layers
– Historic approach
– Relied on Ethernet broadcast
– One computer on a network responded to requests
d Use layer 3 to obtain all parameters
– Current approach
– Relies on IP broadcast (IPv4) or multicast (IPv6)
– Means routers can forward requests to a remote server

Dynamic Host Configuration Protocol (DHCP)
d The standard protocol for automatic configuration
d Popular in private enterprises as well as with service
providers
d Host broadcasts/multicasts a request and receives a reply
d Single message exchange allows a host to obtain
– An IP address and address mask to use
– The IP address of a default router
– The address of a DNS server
– A DNS name
– The location of an image to boot (optional)

DHCP Message Format
d Same message format used for requests and responses
0 8 16 24 31
OP HTYPE HLEN HOPS
TRANSACTION IDENTIFIER
SECONDS ELAPSED FLAGS
CLIENT IP ADDRESS
YOUR IP ADDRESS
SERVER IP ADDRESS
ROUTER IP ADDRESS
CLIENT HARDWARE ADDRESS (16 OCTETS)
.
.
.
SERVER HOST NAME (64 OCTETS)
.
.
.
BOOT FILE NAME (128 OCTETS)
.
.
.
OPTIONS (VARIABLE)
.
.
.

DHCP Protocol
d Significant features of the protocol
– Recovers from loss or duplication
– Avoids synchronized flooding of requests after a power-
failure and restart
– Host discovers DHCP server once and caches server
address for future interaction
d Derived from BOOTstrap Protocol (BOOTP), but adds
dynamic address assignment

Address Lease Paradigm
d DHCP server
– Owns a set of IP addresses
– Chooses an address from the set when a request arrives
– Issues a lease for the address for specified time, T
d Client
– Obtains an address and starts a timer for T time units
– Uses the address to communicate
– When the timer expires, requests the server renew the
lease
– Either receives a renewal and restarts timer or stops
using the address

Thought Problem
d Consider how addresses are assigned
d An ISP using DHCP can choose which IP address to assign
to a customer at a given time
d There are two approaches
– The ISP can remember which address was previously
assigned to each customer and use the same address
– The ISP can assign addresses at random, meaning the
customer will not retain the same address
d Many ISPs try to change the address frequently
d Why?

IPv6 Configuration
d DHCPv6 has been defined, but...
d IPv6 prefers a new procedure known as
IPv6 autoconfiguration
d General idea: host can generate an address without using a
server
d Motivation: allow two hosts to communicate without further
infrastructure

Steps In IPv6 Autoconfiguration
d Obtain a network prefix
– Convention is to use a /64 prefix
– Globally-valid prefix can be obtained from a router
– Local-scope prefix created if no router available
d Generate a unique suffix
d Verify that no one else on the network is using the resulting
address

IPv6 Autoconfiguration in Practice
d Need a unique host suffix
d For /64 network, a 64-bit host suffix is needed
d Recommended approach
– Start with MAC address (globally unique, but only 48
bits)
– Create a 64-bit value
d IEEE standard EUI-64 specifies how 48 bits of an IEEE
MAC address are placed in a 64-bit host suffix

Network Address Translation
(NAT)

NAT Motivation
d IPv4 was running out of addresses
d ISPs only want to limit a customer to one IP address at any
time, but customers want multiple devices to be online
d Engineers invented Network Address Translation (NAT) as a
way to solve both problems

NAT Operation
d Conceptually, NAT device is located between computers at
a site and the rest of the Internet
d Site
– Only needs one globally-valid IP address
– Can have multiple local hosts using the Internet
d Local host has full Internet access
d Service is transparent
– No change in protocols on local hosts
– No change in protocols on Internet servers

Conceptual Organization Of NAT
Internet
Internet site with
multiple computers
NAT Device
from the Internet,
site appears to
be a single host
d NAT is said to be in-line
d From the Internet, site appears to be a single computer
d From within the site, each computer appears to have an
independent connection to the Internet

Addresses Used by NAT
d NAT device runs a DHCP server to hand out IP addresses to
computers at the site
d Addresses assigned are IPv6 link-local or IPv4 private
2
2222222222222222222222222222222222222222222222222
Block Description
2
2222222222222222222222222222222222222222222222222
10.0.0.0/8 Class A private address block
2
2222222222222222222222222222222222222222222222222
169.254.0.0/16 Class B private address block
2
2222222222222222222222222222222222222222222222222
172.16.0.0/12 16 contiguous Class B blocks
2
2222222222222222222222222222222222222222222222222
192.168.0.0/16 256 contiguous Class C blocks
2
2222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
– NAT translates source and/or destination addresses in
datagrams that pass between the site and the Internet

NAT Variants
d Basic NAT
– Only translates IP addresses
– Seldom used in practice
d NAPT
– Translates IP address and transport-layer port numbers
– Most widely-used type of NAT
d Twice NAT
– Works with DNS server
– Provides NAPT plus ability to accept incoming
communication

Example Of Basic NAT
d Suppose
– NAT box has globally-valid IP address of 128.210.24.6
– Computer at a site has private address 192.168.0.1
– Computer contacts Internet site 198.133.219.25
d Resulting translation is:
NAT
to the
Internet
SRC = 192.168.0.1
DST = 198.133.219.25
SRC = 128.210.24.6
DST = 198.133.219.25
SRC = 198.133.219.25
DST = 128.210.24.6
SRC = 198.133.219.25
DST = 192.168.0.1
valid address
128.210.24.6
host at site with
private address
192.168.0.1

Implementation Of NAT
d NAT device keeps an internal translation table
d Table stores translations for both outgoing and incoming
datagrams
d Values filled in automatically when computer at site first
sends datagram to the Internet
d Translation table for previous example
2
22222222222222222222222222222222222222222222222222222222222222
Direction Field Old Value New Value
2
22222222222222222222222222222222222222222222222222222222222222
out
IP Source 192.168.0.1 128.210.24.6
222222222222222222222222222222222222222222222222222
IP Destination 198.133.219.25 -- no change --
2
22222222222222222222222222222222222222222222222222222222222222
in
IP Source 198.133.219.25 -- no change --
222222222222222222222222222222222222222222222222222
IP Destination 128.210.24.6 192.168.0.1
2
22222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1

Transport-Layer NAT (NAPT)
d Handles TCP, UDP, and ICMP
d Translates TCP/ UDP protocol port numbers as well as IP
addresses
d Permits multiple computers at a site to contact the same
Internet service simultaneously without interference
d Examples:
– Two computers at a site download songs from iTunes
– Three computers at a site contact Google simultaneously

Example Of NAPT Translation
d Suppose
– Computers at site have private addresses assigned from
private address block 192.168 / 16
– Two computers at the site each contact TCP port 30000
on computer 128 210.19.20
d NAPT chooses a new port number for each and translates
22222222222222222222222222222222222222222222222222222222222222222222
Dir. Fields Old Value New Value
22222222222222222222222222222222222222222222222222222222222222222222
out IP SRC:TCP SRC 192.168.0.1:30000 128.10.24.6:40001
22222222222222222222222222222222222222222222222222222222222222222222
out IP SRC:TCP SRC 192.168.0.2 :30000 128.10.24.6 :40002
22222222222222222222222222222222222222222222222222222222222222222222
in IP DEST:TCP DEST 128.10.24.6 :40001 192.168.0.1 :30000
22222222222222222222222222222222222222222222222222222222222222222222
in IP DEST:TCP DEST 128.10.24.6 :40002 192.168.0.2 :30000
22222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1

NAT In Practice
d Many consumer products have NAT built in
d Examples:
– Cable and DSL modems
– Wireless routers
d Note that most wireless routers provide both wired and
wireless network connections; they provide NAT on all
connections
Internet connection
DSL or cable modem
wireless router

Transport Layer Protocols:
Characteristics And Techniques

What Should A Network Provide?

d One possibility: network centric
– Network offers all services, such as email, web, etc
– Host accesses services
– Network authenticates user, handles reliability
– Know as customer-provider communication

d One possibility: network centric
– Network offers all services, such as email, web, etc
– Host accesses services
– Network authenticates user, handles reliability
– Know as customer-provider communication
d Another possibility: network provides communication
– Network only transfers packets
– Applications handle everything else, including
reliability, flow control, and authentication
– Known as end-to-end communication

End-To-End Principle
d Fundamental concept in the Internet
d Network provides best-effort packet transport
d Endpoints
– Control communication
– Provide all reliability
d Consequence
Some of the most complex protocols in the Internet protocol
suite run in hosts rather than in routers.

Transport Layer
d Layer between applications and IP
Application
Transport
Internet
Network Interface
Physical
LAYER 1
LAYER 2
LAYER 3
LAYER 4
LAYER 5
d Allows multiple applications on a given host to
communicate with applications on other hosts
d Uses IP to carry messages

Problems A Transport Protocol Can Handle
d Accommodate speed mismatch between sender and receiver
d Detect and recover from datagram loss
d Eliminate duplicate packets
d Guarantee that messages arrive in order
d Respond to congestion in the Internet
d Prevent delayed packets from being misinterpreted
d Verify that data was not corrupted during transit
d Ensure that each party has agreed to communicate
d Note: a given transport protocol may not handle all
problems

Techniques Transport Protocols Use
d Application demultiplexing
– Sender places a value in each outgoing packet that
identifies an application on the receiving host
– Receiver uses the value to determine which application
should receive the packet

d Application demultiplexing
– Sender places a value in each outgoing packet that
identifies an application on the receiving host
– Receiver uses the value to determine which application
should receive the packet
d Flow-control mechanisms
– Receiver informs sender of acceptable data rate
– Sender limits rate to prevent overrunning the receiver

(continued)
d Congestion control mechanisms
– Receiver or network informs sender about congestion in
the network
– Sender reduces data rate (packet rate) until congestion
subsides

(continued)
d Congestion control mechanisms
– Receiver or network informs sender about congestion in
the network
– Sender reduces data rate (packet rate) until congestion
subsides
d Sequence numbers
– Sender places a sequence number in each packet
– Receiver uses the sequence numbers to ensure no
packets are missing and that packets are delivered in the
correct order

(continued)
d Positive acknowledgement with retransmission
– Receiver sends acknowledgement to inform sender when
a packet arrives
– Sender retransmits packet if acknowledgement fails to
arrive within a specified time

(continued)
d Positive acknowledgement with retransmission
– Receiver sends acknowledgement to inform sender when
a packet arrives
– Sender retransmits packet if acknowledgement fails to
arrive within a specified time
d Sliding window
– Instead of transmitting a packet and waiting for an
acknowledgement, a sender transmits K packets and
each time an acknowledgement arrives, transmits another

Transport Protocols Used In The Internet
d Two primary transport protocols used in the Internet
– User Datagram Protocol (UDP)
– Transmission Control Protocol (TCP)
d Choice determined by application protocol
– Many applications specify the use of a single transport
(e.g., email transfer uses TCP)
– Some applications allow the use of either (e.g., DNS
queries can be sent via UDP or TCP)
d Recall: each transport protocol has some surprising
characteristics

Message Transport With
The User Datagram Protocol

User Datagram Protocol (UDP)
d Used
– During startup
– For VoIP and some video applications
d Accounts for less than 10% of Internet traffic
d Blocked by some ISPs

UDP Characteristics
d End-to-end
d Connectionless communication
d Message-oriented interface
d Best-effort semantics
d Arbitrary interaction
d Operating system independence
d No congestion or flow control

End-To-End Communication
d UDP provides communication among applications
d Sending UDP
– Accepts outgoing message from application
– Places message in a User Datagram
– Encapsulates User Datagram in an IP datagram and
sends
d Receiving UDP
– Accepts incoming User Datagram from IP
– Extracts message and delivers to receiving application
d Note: message is unchanged by the network

Connectionless Communication
d An application using UDP can
– Send a message to any receiver (universal)
– Send at any time (asynchronous)
– Stop sending at any time (unterminated)
d That is, a sender does not
– Inform the network before sending (i.e., does not
establish a communication channel)
– Inform the other endpoint before sending
– Inform the network or other endpoint that no more
messages will be sent

Message-Oriented Interface
d UDP
– Accepts and delivers messages (blocks of data)
– Does not require all messages to be the same size, but
does define a maximum message size
– Places each outgoing User Datagram in a single IP
datagram for transmission
– Always delivers a complete message to receiving
application
d Sending application must divide outgoing data into
messages; UDP sends what it is given (or reports an error if
the message is too large)

UDP Message Size
d UDP allows up to 64K octet messages
d As a practical limit, the size of a User Datagram is limited
by payload area in IP datagram
d Maximum IP payload is 64K octets minus size of IP header
d Therefore, the maximum UDP payload is 64K octets minus
size of IP and UDP headers (usually 64K octets minus 28)
d Application can choose any message size up to the
maximum UDP payload

Large And Small Messages
d What happens if an application sends a 10K octet message?

d The message fits into an IP datagram, but...

d The message fits into an IP datagram, but... network frames
have a smaller MTU (typically 1500 octets)

d So, the result of sending a large message is

IP Fragmentation!

IP Fragmentation!
d What happens if an application chooses a small message
size, such as 20 octets?

IP Fragmentation!
d What happens if an application chooses a small message
size, such as 20 octets?
Inefficiency!

Choosing An Optimal Message Size

d What size messages should an application send?

d Optimal UDP message size is S = M – H
– M is the path MTU (i.e., minimum MTU on the path)
– H is the size of IP and UDP headers

d Optimal UDP message size is S = M – H
– M is the path MTU (i.e., minimum MTU on the path)
– H is the size of IP and UDP headers
d Finding M requires an application to
– Violate layering and obtain forwarding information from
IP
– Note: for IPv4, only the local MTU is known
d Bottom line: it may be difficult/ impossible for an
application to compute S

UDP Semantics
d UDP uses IP for delivery

UDP Semantics
d UDP uses IP for delivery and offers the same semantics!

UDP Semantics
d UDP packet can be
– Lost
– Duplicated
– Delayed
– Delivered out of order
– Delivered with data bits altered

UDP Semantics
d UDP packet can be
– Lost
– Duplicated
– Delayed
d Note 1: UDP does not introduce such errors; the errors arise
from the underlying networks

UDP Semantics
d UDP packet can be
– Lost
– Duplicated
– Delayed
d Note 1: UDP does not introduce such errors; the errors arise
from the underlying networks
d Note 2: UDP does include an optional checksum to protect
the data (but the checksum may be disabled)

Using Best-Effort Semantics
d Questions
– Do best-effort semantics make any sense for
applications?
– Why would a programmer choose UDP?

Using Best-Effort Semantics
d Questions
– Do best-effort semantics make any sense for
applications?
– Why would a programmer choose UDP?
d Answers
– Retransmitting a lost message does not make sense for
real-time audio and video applications because a
retransmitted packet arrives too late to be used
– Additional real-time protocols can be added to UDP to
handle out-of-order delivery (we will cover later in the
course)

Arbitrary Interaction
d UDP permits arbitrary interaction among applications
1-to-1
1-to-many
Many-to-1
Many-to-many
d Application programmer chooses interaction type
d Ability to send a single message to multiple recipients can
be valuable

Efficient Implementation Of Interaction
d Key point: UDP can use IP broadcast or multicast to deliver
messages
d Provides efficient delivery to a set of hosts
d Example: UDP packet sent to IPv4 destination address
255.255.255.255 is delivered to all hosts on the local
network (IPv6 has an all nodes multicast address)
d No need for sender to transmit individual copies
d Allows application to find a server without knowing the
computer on which the server runs
d Broadcast is a significant advantage of UDP over TCP for
some applications

Operating System Independence
d Goal is to allow applications on heterogeneous computers to
interact
d Must avoid OS-specific identifiers, such as
– Process IDs
– Task names
d Instead, create application identifiers that are not derived
from any OS

UDP Application Identifiers
d 16-bit integer known as UDP protocol port number
d Each application using UDP must obtain a port number
d Sending UDP
– Places a port number in UDP header to identify
destination application on receiving host
– Also includes port number of sending application
d Receiving UDP
– Uses value in header to select appropriate application
UDP protocol port numbers are universal across all computers,
and do not depend on the operating system.

Identifying An Application
d Both sending and receiving applications need a port number
d Assignment of port numbers depends on the type of
application
d Application that offers a standardized service (server)
– Uses a well-known port number for the service
– Value is less than 1024
– Example: TFTP service uses UDP port 69
d Other applications (client)
– Request a port number from the local operating system
– Value is greater than 49151

Steps Taken To Contact A Service
d Request an unused local port number from the local
operating system
d Obtain the IP address of the local computer from the
operating system
d Look up the port number of the service to be contacted
d Obtain the domain name of a computer that runs the service
and map to an IP address
d Form a UDP datagram with a source port field set to the
local port number and the destination port field set to the
port number of the service
d Request that the UDP datagram be encapsulated in an IP
datagram and sent using the source and destination IP
addresses obtained above

Examples Of Well-Known UDP Ports
222222222222222222222222222222222222222222222222222222
Port Number Description
222222222222222222222222222222222222222222222222222222
0 Reserved (never assigned)
7 Echo
9 Discard
11 Active Users
13 Daytime
15 Network Status Program
17 Quote of the Day
19 Character Generator
37 Time
42 Host Name Server
43 Who Is
53 Domain Name Server
67 BOOTP or DHCP Server
68 BOOTP or DHCP Client
69 Trivial File Transfer
88 Kerberos Security Service
111 Sun Remote Procedure Call
123 Network Time Protocol
161 Simple Network Management Protocol
162 SNMP Traps
514 System Log
222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

UDP Datagram Format
d Extremely thin layer
d User Datagram is divided into header and payload
d Header contains only 8 octets:
0 16 31
UDP SOURCE PORT UDP DESTINATION PORT
UDP MESSAGE LENGTH UDP CHECKSUM
PAYLOAD (MESSAGE DATA)
. . .
d Question: why is length needed?

UDP Checksum

UDP Checksum
d 16-bit 1s-complement checksum

UDP Checksum
d Covers entire UDP packet, including data (recall: IP does
not checksum the payload)

UDP Checksum
d Is optional: value of zero means sender did not compute a
checksum

UDP Checksum
d Is optional: value of zero means sender did not compute a
checksum
d Includes extra pseudo header that contains IP addresses
d Example of IPv4 pseudo header:
0 16 31
IP SOURCE ADDRESS
IP DESTINATION ADDRESS
ZERO PROTO UDP LENGTH

Purpose Of A Pseudo Header
d Receiver can verify that message arrived at correct computer
as well as correct application on that computer
d Consequence for NAT: if it changes the IP source or
destination address, NAT must recompute UDP checksum
d Note: pseudo headers provide another example of layering
violations

UDP Encapsulation
d User Datagram travels in IP datagram
d Two levels of encapsulation occur
UDP Payload
UDP Hdr
d Note: the message the application places in the UDP
Payload field may also have header and payload fields

Transmission Control Protocol
(Stream Transport)

Transmission Control Protocol (TCP)
d The primary transport-layer protocol used in the Internet
d Accounts for about 90% of all Internet traffic (some
estimates are higher)
d Provides reliability
d Appeals to programmers

TCP Characteristics
d End-to-end communication
d Connection-oriented paradigm
d Point-to-point connections
d Complete reliability
d Full-duplex communication
d Stream interface
d Reliable connection startup
d Graceful connection shutdown

End-To-End Communication
d TCP provides communication among pairs of applications
d Allows an application on one host to communicate with an
application on another host
d Permits multiple applications on a given computer to
communicate simultaneously without interference
d Uses protocol port numbers to distinguish among
applications
d Note: TCP ports are completely independent of UDP ports

End-To-End Principle And Transport Protocols

d Transport protocols operate in end systems, and view the
underlying Internet as a virtual network

TCP TCP
IP IP
IP
net iface. net iface.
net iface.
appl. appl.
net 1 net 2
Host A Host B
router
The Internet
communication system
as viewed by TCP

TCP TCP
IP IP
IP
net iface.
appl. appl.
net 1 net 2
Host A Host B
router
as viewed by TCP
d IP does not read or interpret TCP packets

TCP TCP
IP IP
IP
net iface.
appl. appl.
net 1 net 2
Host A Host B
router
as viewed by TCP
d IP does not read or interpret TCP packets
d When forwarding datagrams, router only processes layers 1
through 3

TCP Protocol Port Numbers
d 16-bit integers used to identify applications
d Each application needs a port number
d TCP well-known port assignments are independent of UDP
assignments
d However, to help humans, the same value chosen if service
available via either transport
d Examples
– Both UDP and TCP assign port 53 to the Domain Name
System (DNS)
– Both UDP and TCP assign port 7 to the echo service

Protocol Ports, The Four-Tuple, And Flows
d Key concept: because a TCP connection corresponds to a
pair of endpoints, the connection is identified by four items
– IP source address
– TCP source port
– IP destination address
– TCP destination port
d Commonly called the four-tuple
d Explains how an application such as a web server can
communicate with multiple clients at the same time
d Interestingly, more than four values must be extracted from
a frame to identify a TCP flow

TCP’s Connection-Oriented Paradigm
d Analogous to a telephone call
d Pair of applications must
– Establish a TCP connection before communicating
– Terminate the connection when finished
d Important insights
– A TCP connection is virtual because only the two
endpoints know a connection is in place
– TCP does not have keep-alive messages: no packets are
exchanged unless applications are sending data

Limited Interaction
d A TCP connection only provides communication between a
pair of applications
d Known as a point-to-point communication
d TCP connection does not support
– Reception from an arbitrary set senders
– Multi-point connections with more than two endpoints
– Broadcast or multicast delivery

The TCP Reliability Guarantee
d TCP provides full reliability
d Compensates for
– Loss
– Duplication
– Delivery out of order
d Does so without overloading the underlying networks and
routers
d TCP makes the following guarantee
Data will be delivered or sender will (eventually) be notified.

TCP Reliability
d Uses timeout-and-retransmission
d Receiver returns an acknowledgement (ACK) to sender
when data arrives
d Sender waits for acknowledgement and retransmits data if
no acknowledgement arrives

Illustration Of TCP Retransmission
Events at Host 1 Events at Host 2
send message 1
receive message 1
send ack 1
receive ack 1
send message 2
receive message 2
send ack 2
receive ack 2
send message 3
retransmission timer expires
retransmit message 3
receive message 3
send ack 3
packet lost

Why TCP Retransmission Is Hard
d TCP designed for Internet
– Round-trip delays differ among connections
– Round-trip delays vary over time
d Waiting too long introduces unnecessary delay
d Not waiting long enough sends unnecessary copies
d Key to TCP’s success: adaptive retransmission

How Bad Is The Internet?
d In the old days: delays in seconds, high variability

d Now:

d Now: delays in seconds, high variability

d Now: delays in seconds, high variability
Example round-trip measurements from Ireland to California, 2009

Adaptive Retransmission
d Continually estimate round-trip time of each connection
d Set retransmission timer from round-trip estimate
d Illustration of timeout on two connections:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
est 1
est 2
timeout
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
est 1
est 2
timeout
.
.
.
packet lost
packet lost

Review Of Sliding Window
d Transport protocols use sliding window mechanism
d Idea is to send multiple packets before waiting for an
acknowledgment
d Window size is relatively small (tens of packets, not
millions)
d Motivation is to increase throughput

Illustration Of TCP’s Sliding Window
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
window
window
window
initial position
intermediate position
final position
already acknowledged
still unsent
window moves as
acknowledgements arrive

How Sliding Window Improves Data Rate
host 1 host 2
stop-and-go
send
packet
send
packet
send
packet
send
packet
send
ack
send
ack
send
ack
send
ack
done

How Sliding Window Improves Data Rate
host 1 host 2 host 1 host 2
stop-and-go sliding window
send
packet
send
packet
send
packet
send
packet
send
ack
send
ack
send
ack
send
ack
done
send
four
packets
send
four
acks
done
d Window size of K improves data rate by a factor of K

TCP Flow Control And TCP Window
d Flow control mechanism coordinates data being sent with
receiver’s speed
d Buffer size used instead of data rate
d Receiver tells sender size of initial buffer
d Each acknowledgement specifies space remaining in buffer
d Known as window advertisement

Illustration Of TCP Flow Control
Sender Events Receiver Events
advertise window=2500
send data octets 1-1000
send data octets 2001-2500 ack up to 1000, window=1500
ack up to 2000, window=500
receive ack for 1000
application reads 2000 octets
application reads 1000 octets
.
.
.

TCP Congestion Control And Slow Start
d TCP uses loss or changes in delay to infer congestion in the
network
d When congestion is detected, sending TCP temporarily
reduces the size of the window
d When a packet is lost, TCP temporarily reduces the effective
window to one half its current value
d Later, TCP slowly increases the window again
d Congestion avoidance also used when a connection starts
– Temporarily use a window size of one segment
– Double the window size when ACK arrives
– Known as slow start

Full-Duplex Communication
d TCP connection between A and B provides two independent
data streams, one from A to B and the other from B to A
d Each side
– Has a receive buffer
– Advertises a window size for incoming data
– Uses sequence numbers to number outgoing data bytes
– Implements timeout-and-retransmission for data it sends
d Application can choose to shut down communication in one
direction

Full-Duplex Communication
(continued)
d Each TCP packet contains fields for both forward and
reverse data streams
– Sequence number for data being sent in the forward
direction
– Acknowledgement number for data that has been
received

Stream Interface
d After connection is established, TCP accepts a
stream of data bytes from the sending application and
transfers them
d Sending application can choose amount of data to pass on
each request
d Surprise: TCP decides how to group bytes into packets
d Known as stream interface
d Consequence
Data may be passed to a receiving application in chunks that
differ from the chunks that the sending application generated.

Connection Startup And Shutdown
d Difficult problem
d Packets can be
– Lost
– Duplicated
– Delayed
d Either end can crash and reboot
d Need to know that both sides have agreed to start/ terminate
the connection

Reliable Connection Startup
d TCP guarantees reliable connection startup that avoids
replay problems
d Performed with 3-way handshake
send SYN
receive SYN
send SYN + ACK
receive SYN + ACK
send ACK
receive ACK
d Each side chooses starting sequence number at random

Graceful Connection Shutdown
d Analogous to 3-way handshake for startup
d Guarantees no ambiguity about connection termination
send FIN + ACK
receive FIN + ACK
send FIN + ACK
receive FIN + ACK
send ACK
receive ACK
.
.
.

TCP Segment Format
d TCP packet is called a segment
d Segment is encapsulated in IP for transmission
d Single format used for SYNs, FINs, ACKs, and data
0 4 10 16 24 31
SOURCE PORT DESTINATION PORT
SEQUENCE NUMBER
ACKNOWLEDGEMENT NUMBER
HLEN NOT USED CODE BITS WINDOW
CHECKSUM URGENT POINTER
OPTIONS (if any)
BEGINNING OF DATA
.
.
.

Routing Algorithms
And Routing Protocols

Historical Perspective
d Computing in the 1960s
– Mainframes
– Batch processing with punched cards
– Usually one computer per organization
d Computing in the 1970s
– Minicomputers
– A few computers per organization
– Dumb terminals

Traditional Wide Area Networks
d Developed during 1960s mainframe era
d Predate
– LANs
– PCs
d Basic motivation
– Interconnect mainframe at one site to mainframes at
other sites
– Allow resource sharing
d First to employ dynamic routing

Traditional WAN Architecture
d Dedicated device known as packet switch placed at each site
d Packet switch provides
– Local connections for host computer(s) at the site
– Long-distance connections to other sites
d Connection among sites
– Leased digital circuits
– Leased raw copper or fiber with customer supplying
modems

Packet Switch Used In Traditional WAN
d Special-purpose, stand-alone device
d Dedicated to packet forwarding
d Small computer with
– Processor
– Memory
– Program on stable storage
– I/ O interfaces

Conceptual View Of Traditional Packet Switch
packet switch
local
computers
processor
memory
I/O interfaces
for local computers
I/O interfaces
for remote sites
internal
interconnects
leased circuits
to remote
sites
d Memory needed to store packets

Store And Forward Paradigm
d Key paradigm used in packet switching
d Operation
– Interface hardware places each arriving packet in a
queue in memory
– Processor continually removes next packet from the
queue and forwards toward its destination
d Motivation: memory is a buffer that accommodates a short
burst of packets that arrive back-to-back
Important point: packet traffic tends to be bursty.

Example Of Traditional WAN Architecture
d Packet switch at each site connects to other sites
d Circuits accommodate traffic and desired robustness
packet
switch
at
site 1
packet
switch
at
site 2
packet
switch
at site 3
packet
switch
at site 4
each computer
connects to a
packet switch
digital circuits
between switches

Traditional WAN Addressing

d Hierarchical model analogous to Internet addressing

d Conceptual two-level hierarchy
( site, computer at the site )

d Conceptual two-level hierarchy
( site, computer at the site )
d In practice, one packet switch per site and K connections for
local computers means the address hierarchy is:
( packet switch, local connection on the switch )

Illustration Of Traditional WAN Addressing
1
2
3
4
5
6
1
2
3
4
5
6
packet
switch
1
packet
switch
2
address is [1,2]
address is [1,5]
address is [2,1]
address is [2,6]
d The two parts of an address are combined to form a single
binary number

Next-Hop Forwarding
d Analogous to IP datagram forwarding
d Each packet contains a destination address
d Forwarding uses only the packet switch portion of an
address; delivery uses the rest of the address
d If packet has reached the destination packet switch, deliver
to locally-connected computer
d Otherwise, forward to another packet switch that is closer to
the destination site

Algorithm For Packet Forwarding
Given:
An incoming packet arriving at a packet switch
Perform:
The next-hop forwarding step
Method:
Extract the destination address from the packet and
divide into packet switch, P, and computer, C;
if ( P is the same as “my” packet switch number ) {
Deliver the packet to local computer C;
} else {
Use P to select a next hop, and forward the packet
over the selected link to the next hop;
}
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2222222222222222222222222222222222222222222222222222222

WAN Forwarding Table
d Analogous to IP forwarding table
d Each entry in table refers to a switch, not an individual
computer
to reach send to
switch 1
switch 2
switch 3
interface 1
local delivery
interface 4
Example WAN with three packet switches Forwarding table for switch 2
[1,2]
[1,5]
[3,2]
[3,5]
[2,1] [2,6]
packet
switch
1
packet
switch
3
packet
switch 2
interface 1
interface 4

Modern WAN Architecture
d Uses IP technology
d Router at site has
– Local connections to networks at the site
– Long-distance connections to routers at other sites
d Typical use: connect all sites of an organization

Illustration Of Modern WAN Connections
LAN (e.g., Ethernet) Router
local computers
connections
to other sites
d Uses conventional IP router
d Typical remote connection is a leased data circuit
d Router can also provide connection to the Internet

Routing Algorithms
And Internet Routing

Constructing A Forwarding Table
d Static routing
– Used in Internet hosts
– Entries inserted when system boots and do not change
d Dynamic routing
– Used in packet switches and IP routers
– Initial entries inserted when system boots
– Routing software continually monitors network,
computes shortest paths, and updates forwarding table

Static Routing
d Used in most hosts
d Only K+1 entries in forwarding table if host has K network
connections
d K entries, one per network connection
– IP prefix for the network
– Address mask for the network
– Interface for the network
d Final entry: default route
– default IP router address
– Interface for the default router

Dynamic Routing
d Routing Software
– Runs on each packet switch or router
– Computes shortest paths and installs entries in local
forwarding table
d Models the network as a graph
1 2
3 4
1 2
3 4
Example WAN Equivalent graph
edges or links
nodes

Example Graph And Next-Hop Forwarding Tables
1 2
3 4
2
222222222222222 2
2222222222222222 2
2222222222222222 2
2222222222222222
to send to send to send to send
reach over reach over reach over reach over
2
222222222222222 2
2222222222222222 2
2222222222222222 2
2222222222222222
1 – 1 (2,3) 1 (3,1) 1 (4,3)
2
222222222222222 2
2222222222222222 2
2222222222222222 2
2222222222222222
2 (1,3) 2 – 2 (3,2) 2 (4,2)
2
222222222222222 2
2222222222222222 2
2222222222222222 2
2222222222222222
3 (1,3) 3 (2,3) 3 – 3 (4,3)
2
222222222222222 2
2222222222222222 2
2222222222222222 2
2222222222222222
4 (1,3) 4 (2,4) 4 (3,4) 4 –
11
1
1
1
1
1
1
1
1
1
1
2
22222222222222211
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
2
222222222222222211
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
2
222222222222222211
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
2
222222222222222211
1
1
1
1
1
1
1
1
1
1
node 1 node 2 node 3 node 4

Dynamic Routing
d Goals
– Consistent, optimal routes
– Automatic route change to accommodate failures
d Each node (packet switch or router) participates
d Routing software on a node exchanges information with
routing software on other nodes
d Distributed computation
d Two basic algorithms employed
– Distance-Vector (DV)
– Link-State Routing (LSR)

Distance-Vector (DV) Routing
d Approach used in many early routing protocols
d Also known as Bellman Ford
d Node
– Receives information from neighbors
– Combines information from all neighbors with local
information
– Sends copy of processed information to all neighbors

How DV Works
d A participant periodically sends route advertisement to each
neighbor
d Advertisement specifies reachable sites and distance to each
I can reach site X, and its distance from me is Y.
I can reach site Z, and its distance from me is W.
.
.
.
d Neighbor receives advertisement and updates its forwarding
table
d In next round, neighbors each send advertisements to their
neighbors

Distance-Vector Algorithm
d Used when advertisement arrives
d Examine each item in advertisement
– If neighbor can reach site X and I cannot, add an entry
to my forwarding table for X with the neighbor as the
next hop
– If I already have a route to X with the neighbor as the
next hop, replace the distance in the route with the
advertised distance
– If I have a route to X that is more expensive than going
through the neighbor, change the next hop to the
neighbor

Measuring The Distance Of A Route
d Possible measures
– Hops
– Delay
– Throughput
– Economic or administrative cost
d Many protocols use hops, but routing software often permits
a manager to assign administrative hop counts

Link-State Routing (LSR)
d Chief alternative to distance-vector
d Each node
– Sends link status information
– Computes shortest paths independently
– Does not rely on computation performed by others
d Name
– Formal name is Link-State or Link-Status Routing
– Also called Shortest Path First (SPF), a somewhat
misleading term derived from underlying algorithm

How LSR Works
d Each pair of directly-connected nodes periodically
– Tests connection between them
– Broadcasts one of the following messages:
The link between X and Y is up.
or
The link between X and Y is down.
d Each node
– Collects incoming broadcast messages and creates a
graph
– Uses Dijkstra’s SPF algorithm to compute a forwarding
table (see text for details and example)

Review Of Internet Forwarding
d Hosts
– Use static routing
– Entries placed in forwarding table when system boots
and remain unchanged
d Routers
– Use dynamic routing
– Initial entries placed in forwarding table when system
boots and routing software updates entries continually

Example Of Host Routing
Ethernet 128.10.0.0 / 16
Router R1
128.10.0.0
default
255.255.0.0
0.0.0.0
direct
128.10.0.100
Net Mask Next hop
hosts on a network
forwarding table in each host
router address
128.10.0.100 to rest of
Internet
d Next hop in default route is known as a default router

Why Dynamic Internet Routing Is Needed
d Router
– Only has direct connections to a few networks
– Must know how to forward datagram to arbitrary
destination
d Example
R1 R2
network 1 network 2
network 3
d Router R1 must learn about network 2 and R2 must learn
about network 1

Important Principle
No single routing protocol can be used across the entire
Internet because the overhead is too high.

Autonomous System Concept
d Internet divided into a set of routing domains
d Each routing domain is
– Known as an autonomous system (AS)
– Assigned a unique number
d Generally, an AS is a contiguous set of routers and networks
under one administrative authority
d No exact definition; think of a large ISP or a large
corporation
d AS gathers and summarizes routing information before
passing it to another AS

Two Types Of Internet Routing Protocols

d Interior Gateway Protocols (IGPs)
– Used within an autonomous system
– Choice of IGP is made by each AS
– Relatively easy to install and manage

d Interior Gateway Protocols (IGPs)
– Used within an autonomous system
– Choice of IGP is made by each AS
– Relatively easy to install and manage
d Exterior Gateway Protocols (EGPs)
– Used between autonomous systems
– More complex to install and configure
– Include policy constraints that control which information
is revealed

Illustration Of IGPs and EGPs
R1
R2
R3
R4
R5
R6
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...................................
...................................
Autonomous
System 1
Autonomous
System 2
EGP used
IGP1 used IGP2 used
d Because metrics used in each AS may differ, direct
comparison is impossible

Principle Of Route And Data Flow
d Data flows in opposite direction of routes
d Example: ISP1 advertises route to customer Q and receives
traffic for customer Q
R1 R2
ISP1 ISP2
Routes for customers
of ISP1
Data to customers
of ISP1

Border Gateway Protocol (BGP)
d Primary Exterior Gateway Protocol used in the Internet
d Used by Tier 1 ISPs at the center of the Internet
d Current version is 4 (BGP-4)
d Characteristics
– Provides routing among autonomous systems
– Includes provisions for policies
– Distinguishes transit routes from terminal routes
– Uses reliable transport (TCP)
– Sends path information

Illustration Of BGP Paths
d Modified Distance-Vector protocol
d Advertisement contains a path in place of a distance
d Path lists the autonomous systems to destination
d Example
To reach network X, I send along path Z, Y, W,...
d Path information means receiver can apply policies (e.g.,
receiver can choose to ignore all routes that pass through AS
number N)

Routing Information Protocol (RIP)
d Among the earliest Interior Gateway Protocols
d Characteristics
– Distance-Vector that uses hop-count metric
– Sent over UDP (unreliable transport)
– Advertises CIDR prefixes
– Includes facility for default route propagation
– Broadcast or multicast delivery
d Current version is 2 (RIP2)

RIP2 Packet Format
0 8 16 24 31
COMMAND (1-5) VERSION (2) MUST BE ZERO
FAMILY OF NET 1 ROUTE TAG FOR NET 1
IP ADDRESS OF NET 1
ADDRESS MASK FOR NET 1
NEXT HOP FOR NET 1
DISTANCE TO NET 1
FAMILY OF NET 2 ROUTE TAG FOR NET 2
IP ADDRESS OF NET 2
ADDRESS MASK FOR NET 2
NEXT HOP FOR NET 2
DISTANCE TO NET 2
. . .
d Note: routing protocols run at application layer (layer 5)

Open Shortest Path First Protocol (OSPF)
d Created by the IETF to be an open standard (reaction to
proprietary protocols)
d Characteristics
– Interior Gateway Protocol
– Advertises CIDR prefixes
– Authenticated message exchange
– Can import routes from BGP
– Link-state algorithm
– Provides for multi-access networks
– Divides large network into areas

Illustration Of An OSPF Graph
R1 R2
R3
R4
R5 R6
R1 R2
R3
R4
R5 R6
a network the OSPF graph
d Graph shows a link between each pair of routers even
though some connections cross a shared network

Intermediate System - Intermediate System (IS-IS)
d Originally part of DECNET V protocols
d Uses LSR approach
d Initially
– Considered somewhat over featured
– Not widely accepted in the Internet
– Overshadowed by OSPF
d Eventually
– OSPF became complex as features were added
– IS-IS started to gain acceptance

Where Intuition Fails
d Routing is not like water flowing through pipes or traffic on
highways
– Multi-path routing is difficult
– Capacity can go unused if not along shortest path
d Fewest hops may not always be best
– Compare two Ethernet hops and one satellite hop
d Routing around congestion is not straightforward, and does
not always yield a big improvement
– Can cause out-of-order packets (TCP reacts)
– Can result in route flapping

Loops And Convergence
d Routing loop
– Circular routes
– Can be caused if “good news” flows backward
d Slow convergence (count to infinity) problem arises
– Routes fail to converge after a change
– Can cause a routing loop to persist

How Good News Can Backwash
d A story with three routers and a network
A B C
network N

How Good News Can Backwash
d A story with three routers and a network
A B C
network N
d In practice, modern DV protocols employ heuristics that
– Eliminate backflow
– Lock down changes after a failure

Other Routing Problems
d Black hole
– Routing system sends packets for a set of destinations to
a location where they are silently discarded
– Can be caused if routing update packets are lost
d Route flapping (lack of convergence)
– Routes continue to oscillate
– Can be caused by equal-length paths

Routing Overhead
d Traffic from routing protocols is “overhead”
d Specific cases
– DV advertisements tend to be large
– LSR uses broadcast
d Fundamental tradeoff
– Decreasing frequency of routing exchanges lowers
overhead
– Increasing frequency of routing exchanges reduces the
time between a failure and rerouting around the failure

Internet Multicast
And Multicast Routing

IPv4 Multicast
d Defined early; informally called “Deering multicast”
d Provides Internet-wide multicast dissemination
d Uses IPv4 addresses 224.0.0.0 through 239.255.255.255 (the
original Class D address space)
d In theory, any host in the Internet can
– Join or leave any group at any time
– Send a datagram to any group at any time

IPv4 Multicast
d Defined early; informally called “Deering multicast”
d Provides Internet-wide multicast dissemination
d Uses IPv4 addresses 224.0.0.0 through 239.255.255.255 (the
original Class D address space)
d In theory, any host in the Internet can
– Join or leave any group at any time
– Send a datagram to any group at any time
Internet-wide multicast is not widely deployed

IPv6 Multicast
d Fundamental part of IPv6
d IPv6 prohibits broadcast, but defines multicast groups that
are equivalent
– All routers
– All nodes

Internet Group Multicast Protocol (IGMP)
d Allows a host to join or leave a multicast group
d Restricted to a single network (host talks to local router)
d When first host on a network joins a new group or last host
on a network leaves a group, router(s) on the network
change multicast routes accordingly

IP Multicast And Ethernet Delivery
d When sending IP multicast across Ethernet
– Can use Ethernet multicast capability
– IP multicast address is mapped to an Ethernet multicast
address
d Problem
– Most interface hardware limits the number of Ethernet
multicast addresses that can be used simultaneously
– Trick: use a few multicast addresses and allow software
to decide how a given packet should be processed

Multicast Routing Protocols
d Needed to propagate multicast routes throughout the Internet
d Goals
– Ensure all participants in a group receive packets sent to
the group
– Avoid flooding multicast across a network unless a host
is listening
d General approach
– Form a graph-theoretic tree for each multicast group
– Forward multicast along links of the tree
d Trick: send a request for group X toward the “center” of the
Internet until it reaches a router that knows about group X

Example Multicast Routing Protocols
d Many multicast routing protocols have been proposed
d A few examples
2222222222222222222222222222222222222222222222
Protocol Type
2222222222222222222222222222222222222222222222
DVMRP Configuration-and-Tunneling
2222222222222222222222222222222222222222222222
CBT Core-Based-Discovery
2222222222222222222222222222222222222222222222
PIM-SM Core-Based-Discovery
2222222222222222222222222222222222222222222222
PIM-DM Flood-And-Prune
2222222222222222222222222222222222222222222222
MOSPF Link-State (within an organization)
2222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1

Summary
d Internet
– Consists of a network of heterogeneous networks
– Separates communication from content and services
– Accommodates arbitrary network technologies and
applications
d IPv4 uses 32-bit addresses; IPv6 uses 128-bit addresses
d Internet packet is known as an IP datagram
d Datagram is encapsulated for transmission
d Fragmentation and reassembly accommodate heterogeneous
MTUs

Summary
(continued)
d IPv4 uses ARP for address resolution and IPv6 uses ND
d ICMP (Internet Control Message Protocol) reports errors
back to the original source
d Ping uses ICMP echo request and echo response
d DHCP allows automatic configuration
d NAT hides multiple computers behind a single address
d Internet follows the end-to-end principle

Summary
(continued)
d Transport protocols that provide end-to-end service run in
hosts
d Internet has two main transport protocols
– UDP provides unreliable, connectionless message
delivery
– TCP provides reliable, stream-oriented delivery
d Dynamic routing was created for WANs and is used in the
Internet
– Distance Vector
– Link State (also called SPF)

Summary
(continued)
d Internet is divided into Autonomous Systems
d EGPs used between Autonomous Systems
d IGPs used within an Autonomous System
d Internet routing protocols include
– Border Gateway Protocol (BGP)
– Routing Information Protocol (RIP)
– Open Shortest Path First (OSPF)
– Intermediate System-Intermediate System (IS-IS)
d Multicast routing protocols defined, but are not in wide use

MODULE VI
Other Topics

Topics
d Measuring network performance
d Quality of Service (QoS) and provisioning
d Multimedia and IP telephony
d Network security
d Traffic engineering and MPLS
d Network management (SNMP)

Why Measure Network Performance?
d Optimization
d Planning (anticipating future needs)
d Assessing and understanding traffic
– Trends in applications and network use
– Detecting anomalous traffic patterns
d Contract (SLA) enforcement
d Bragging rights
– IT staff in an organization
– Marketing department in an equipment vendor

Qualitative Terminology And Marketing

d Marketing seems to love qualitative terms
– High-speed
– Fast
– Powerful
– High bandwidth

d Marketing seems to love qualitative terms
– High-speed
– Fast
– Powerful
– High bandwidth
d Unfortunately
– Qualitative terminology is vague
– Networking technologies change rapidly

Qualitative Terminology That Faded
d A high-speed leased line
– Was once defined to run at 9.6 Kbps
d The Internet’s Very high-speed Backbone Network System
(VBNS)
– Used OC-12 links, that are no longer considered very
high speed
d Fast Ethernet
– Runs at 100 Mbps and is only one-tenth as fast as
Gigabit Ethernet technology
d Broadband
– Was once defined by the FCC to start at 128 Kbps

Quantitative Measures
d Quantifiable measurement is surprisingly difficult
d Routes and data rates can be asymmetric, making
measurements in one direction differ from measurements in
the other
d Inserting measurement probes can affect the performance of
the system being measured
d Conditions can change rapidly

Aggregate Traffic Analysis

d Short-term variation
– Packets tend to arrive in clumps called bursts
d Long-term variation
– Diurnal and annual patterns exist

d Short-term variation
– Packets tend to arrive in clumps called bursts
d Long-term variation
– Diurnal and annual patterns exist
d Interestingly, data traffic is unlike voice traffic
– Aggregate of voice telephone calls is smooth average
– Aggregate of data traffic is bursty

Self-Similarity
Unlike voice telephone traffic, data traffic is bursty. Data
traffic is said to be self-similar because aggregates of data
traffic exhibit a pattern of burstiness that is statistically similar
to the burstiness on a single link.
The point: data traffic is not easy to analyze

Practical Measures Of Network Performance
d Three primary quantitative measures
2222222222222222222222222222222222222222222222222222222222222222222
Measure Description
2222222222222222222222222222222222222222222222222222222222222222222
Latency (delay) The time required to transfer a bit across
a network from one end to another
2222222222222222222222222222222222222222222222222222222222222222222
Throughput (capacity) The amount of data that can be transferred
over a network per unit time
2222222222222222222222222222222222222222222222222222222222222222222
Jitter (variability) The changes in delay that occur and the
duration of the changes
2222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d We will see that the three are not completely independent

Latency Or Delay
d Time required for data to travel “across” a network
d Think of latency as the time required for a single bit to
traverse a network
d Depends on
– Physical properties of the universe (the speed of light)
– Traffic on the network

Latency And Perceived Response Time
d Users are interested in response time
d Several components of delay contribute to overall response
time a user perceives
2
222222222222222222222222222222222222222222222222222222222222222222
Type Explanation
2
222222222222222222222222222222222222222222222222222222222222222222
Access Delay The time needed to obtain access to a
transmission medium (e.g., a cable)
2
222222222222222222222222222222222222222222222222222222222222222222
Propagation Delay The time required for a signal to travel across
a transmission medium
2
222222222222222222222222222222222222222222222222222222222222222222
Switching Delay The time required to forward a packet
2
222222222222222222222222222222222222222222222222222222222222222222
Queuing Delay The time a packet spends in the memory of a
switch or router waiting to be selected for
transmission
2
222222222222222222222222222222222222222222222222222222222222222222
Server Delay The time required for a server to respond to a
request and send a response
2
222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Bottlenecks
d Any part of a communication system can be a bottleneck
that causes the most delay
d Examples
– Access delay: acquiring a wireless channel
– Propagation delay: a satellite transmission
– Switching delay: deep packet inspection
– Server delay: a news agency web site overloaded during
a crisis
– Queuing delay: packets arriving faster than they depart

Assessing Delay
d Make multiple measurements over an interval
d Report minimum, maximum, mean, and standard deviation
d Divide delay into constituent components if possible
d Choose small intervals to look for repeated patterns

Throughput
d Maximum amount of data a network can transport per unit
time
d Expressed as data rate in bits per second (e.g., 100 megabits
per second)
d Mistakenly cited as network “speed”, but really a measure
of network capacity
d Gives an upper-bound on performance, not a guarantee

Assessing Throughput
d Several possible measures
– Capacity of a single communication channel
– Capacity along a path through the network
– Aggregate capacity of all channels
– Capacity among pairs of ingress and egress points when
used simultaneously

The Concept Of Goodput
d Invented to provide meaningful assessment of network
performance
d Defined as the effective rate at which an application receives
data
d Can differ from throughput for any of the following reasons
– Application protocol overhead
– Channel coding overhead
– Packet header overhead
– Receiver buffer limitations
– Congestion avoidance mechanisms
– Packet retransmission

Assessing Goodput
d Measure data that arrives successfully, and compute the
amount of data per unit time
d Goodput measurements also include the overhead introduced
by
– Operating system
– Transport protocol
– Lower layer encodings and protocols
– Application protocol and implementation
d Note: although they use the term throughput, most
measurement tools report goodput

Jitter
d Another prominent measure of network performance
d Especially important in transmission of streaming audio and
video
d Measures variation in delay
d Example
– Suppose network has average delay D
– If each packet takes exactly D time units to traverse the
network, jitter is zero
– If packets alternate between delays of D+ε and D – ε,
average delay remains D, but jitter increases

Key Observation
In the Internet, congestion is the single most significant
cause of packet loss, high jitter, and long delays.

Handling Jitter
d Replace the Internet with an isochronous network
– Approach used in the original telephone network
– All parallel paths have exactly the same delay
d Change the Internet to reserve capacity
– Discussed later in the module
d Keep the current Internet design and add protocols that
compensate for jitter
– Basic technique is a jitter buffer
– Discussed later in the module

Understanding Throughput And Delay
d An analogy
– Think of a network as a road between two locations
– Propagation delay determines how long it takes a single
car to traverse the road
– Throughput determines how many cars can enter the
road per unit time
d Observe
– Adding a lane doubles the throughput (i.e., capacity), but
leaves the delay unchanged
– It is possible to have arbitrarily high throughput, even if
the delay is long (imagine a long road with hundreds of
lanes)

(continued)
d The analogy helps us understand network measures

(continued)
Propagation delay specifies the time a single bit remains in
transit in a network. Throughput, which specifies how many
bits can enter the network per unit time, measures network
capacity.

(continued)
Propagation delay specifies the time a single bit remains in
transit in a network. Throughput, which specifies how many
bits can enter the network per unit time, measures network
capacity.
d The key consequence is incorporated in an aphorism
You can always buy more throughput, but you cannot buy
lower delay.

Delay-Throughput Product
d Specifies the maximum amount of data “in flight”
Bits present in a network = D × T
where
– D is delay measured in seconds
– T is throughput measured in bits per second
d Specifies how many bits can be transmitted before the first
bit arrives at the receiver
d Often incorrectly labeled the delay-bandwidth product

Delay-Throughput Terminology And Examples
d Ethernet
– Although it has high throughput, the short delay limits
the delay-throughput product
d Satellite link
– Usually has a high delay-throughput product because
delay is long and throughput is high
d Informally, we use an analogy
– A network with a long delay is called a long pipe
– A network with high throughput is called a fat pipe
– A satellite is known as a long, fat pipe

Delay, Throughput,
and Utilization

Relationship Between Delay And Throughput
d In theory, delay and throughput are independent
d In practice, they are related
d Reason
– Throughput determines rate at which traffic can pass
across a communication link
– A switch or router queues packets until they can be sent
– If data arrives at a switch or router faster than it leaves,
queue length grows, which means increased delay
(congestion)

Illustration Of How Congestion Occurs
d Consider a router with three 1 Gbps connections, and
assume that traffic is arriving over two connections destined
for the third
router
input 1 (1 Gpbs)
input 2 (1 Gbps)
output (1 Gbps)
d If the capacity of the red link is doubled, all links can
experience more congestion, which increases delay

Utilization
d Measure of the current load on a network link
d Given as a percentage of capacity being used, and expressed
as a real value between 0.0 and 1.0
d Example: if a link capable of 1 Gbps has traffic of 500
Mbps, link utilization is 0.5
d Because utilization changes over time, it is reported over an
interval by giving
– Peak (i.e., maximum)
– Average (i.e., mean)

Utilization As Estimate Of Delay
d Packet traffic is bursty
d Key discovery: the effective queuing delay can be estimated
from the utilization as follows:
D =
(1 − U)
D0
3333333
d Where
– D0 is delay when the network is idle
– U is current utilization between 0 and 1

Delay As A Function Of Utilization
utilization
relative
delay
25% 50% 75% 100%
1
2
3
4
5
6
idle delay

Practical Interpretation Of Utilization
d Delay increases rapidly as utilization climbs
d When utilization reaches 50%, delay is double
d When utilization reaches 80%, delay is five times higher
than average

The 50-80 Rule
d Heuristic managers follow
– When utilization reaches 50%, plan an upgrade
– When utilization reaches 80%, an upgrade is overdue
d Note: alternative consists of partitioning a network (e.g.,
separating VLANs)

Line Speed And Packets Per Second
d Networking equipment is said to operate at line speed if the
equipment can handle a sequence of back-to-back packets
d Observe
– Per-packet overhead is often the bottleneck in equipment
– For a given data rate, equipment processes
* Fewer packets per second if packets are large
* More packets per second if packets are small
d Conclusion: line speed is meaningless without a
specification of packet size

Quality of Service (QoS)
and Provisioning

Quality of Service (QoS)
d Set of technologies that can be used to provide service
guarantees
– Bound on latency
– Guarantee on throughput
– Bound on jitter
d Marketing
– Tries to equate QoS and “quality”
– Implies that lack of QoS means lack of quality

QoS In The Internet
d Motivation
– Make it possible to run applications such as streaming
video with no interruptions
– Allow service providers to charge (much) more for
better service
d Three approaches have been proposed and studied
– Priority
– Fine-grain QoS
– Coarse-grain QoS

Priority Approach
d Each packet assigned a priority, and multiplexing selects
packets in priority order
d Popular among ISPs, and used by some corporations to give
voice and video traffic priority
d Advantages
– Easy to implement
– Can assign priority to a “customer” rather than to a
specific type of data
d Disadvantages
– No quantitative guarantees
– Can lead to starvation

Fine-grain QoS Approach
d Pursued by the IETF under the name Integrated Services
(IntServ) and adopted in ATM networks
d QoS parameters negotiated for each flow (e.g., each TCP
connection)
– Maximum delay
– Minimum throughput
– Maximum jitter
d Difficult/ impossible to implement
After many years of research and standards work, the fine-
grain approach to QoS has been relegated to a few special
cases.

QoS Terminology That Has Survived
d Derived from ATM
2
2222222222222222222222222222222222222222222222222222222222222222222222222
Abbreviation Expansion Meaning
2
2222222222222222222222222222222222222222222222222222222222222222222222222
Data enters the flow at a fixed rate,
CBR Constant Bit Rate such as data from a digitized voice
call entering at exactly 64 Kbps
2
2222222222222222222222222222222222222222222222222222222222222222222222222
Data enters the flow at a variable
VBR Variable Bit Rate rate within specified statistical
bounds
2
2222222222222222222222222222222222222222222222222222222222222222222222222
The flow agrees to use whatever
ABR Available Bit Rate data rate is available at a given
time
2
2222222222222222222222222222222222222222222222222222222222222222222222222
No bit rate is specified for the flow;
UBR Unspecified Bit Rate the application is satisfied with
best-effort service
2
2222222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
d Bounds specified statistically (e.g., average and peak
throughput and burst size)

Coarse-grain QoS Approach
d Current approach approved by the IETF under the name
Differentiated Services (DiffServ)
d Divides traffic into classes
d Service guaranteed for each class rather than per flow
d Easier to implement than fine-grain approach
d Usually implemented as a proportional guarantee rather than
absolute quantities
d Example policy
At least 10% of the underlying network
capacity is reserved for voice traffic

Steps A Router Takes To Implement QoS
router implementing QoS
Classification
and Policing
Forwarding
Computation
Output
Queuing
Traffic
Scheduling
packets
arrive
packets
leave
d Policing enforces rules on incoming traffic
d Forwarding can select among multiple paths (router may
have many output queues)
d Queuing may use Random Early Discard (RED)

Traffic Scheduling
d Algorithm used to select packets from queues
d Principal types
2
22222222222222222222222222222222222222222222222222222222222222222222222
Algorithm Description
2
22222222222222222222222222222222222222222222222222222222222222222222222
Leaky Bucket Allows a queue to send packets at a fixed rate by
incrementing a packet counter periodically and using
the counter to control transmission
2
22222222222222222222222222222222222222222222222222222222222222222222222
Token Bucket Allows a queue to send data at a fixed rate by
incrementing a byte counter periodically and using the
counter to control transmission
2
22222222222222222222222222222222222222222222222222222222222222222222222
Weighted Selects packets from a set of queues according to a
Round Robin set of weights that divide the capacity into fixed
percentages, assuming a uniform packet size
2
22222222222222222222222222222222222222222222222222222222222222222222222
Deficit A variant of the round-robin approach that accounts for
Round Robin bytes sent rather than packets transferred, and allows
a temporary deficit caused by a large packet
2
22222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Traffic Engineering
d An approach to networking that allows a manager to
establish and control routes through a network and assign
specific types of data to each
d Implies
– Non-standard forwarding mechanism
– All traffic of a given type sent along a specified path
d Most popular technology: MPLS

Multi-Protocol Label Switching (MPLS)
d Widely deployed among tier-1 ISPs
d Requires participating routers to have MPLS module
d MPLS tunnel created by configuring routers along a path
d Router may allow manager to assign a portion of link
capacity to each tunnel
– Term multi-protocol arises because an MPLS packet can
contain arbitrary content

Label Rewriting
d Concept came from ATM and is used in MPLS
d Each link in path has different integer label
d Router rewrites label in MPLS datagram before forwarding
to next hop
d Known as label switching
d Motivation: avoid global coordination and allow local
assignment of labels
d Disadvantages:
– No general protocol available to configure an MPLS
path
– Can be difficult to debug

How MPLS Works
d Datagram encapsulated in MPLS header by router at the
start of a tunnel
d MPLS datagram tagged with label of path over which it
must pass
d Each router along the path
– Uses label to make forwarding decision
– Replaces label with value used on next hop
d MPLS encapsulation removed when datagram reaches end
of tunnel

Illustration Of Label Rewriting
R1 R2 R3
R4
H1
H2
A
A B A
B
C A
B
A
B
C
A

Illustration Of Label Rewriting
R1 R2 R3
R4
H1
H2
A
A B A
B
C A
B
A
B
C
A
send to label 4
4 32 B 32 12 B
12 3 C
receive from
label 3
d Labels along the path are: 4, 32, 12, 3

A Few Definitions
d Multimedia combines two or more forms of information,
such as
– Photos and music
– Audio and video
d Real-time refers to information that must be presented in a
predetermined timed sequence, such as
– Audio
– Video
d An individual source provides one particular sequence of
real-time information

A Few Definitions
(continued)
d Playback refers to the output of real-time information for a
user (e.g., video display or audio output)
d Sample rate refers to the rate at which real-time information
has been converted to digital form (e.g., audio sampled 8000
times per second)
d Synchronization refers to the coordination of playback
information from multiple sources (e.g., a movie requires
synchronization between audio and video)

Real-Time Sample Rates
d Each source of real-time data can choose a sample rate and
encoding
d Examples
– A video stream might contain 30 frames per second,
with an encoding that uses compression
– An audio stream might contain 8000 audio samples per
second using a PCM encoding
d Important concept
Because each source of real-time information can choose a
sample rate, playback and synchronization must know the
sample rate and encoding that was selected.

Transfer Of Streamed Real-Time Data
d Source
– Samples information at regular intervals
– Generates data continuously
– Prepares data for transmission
d Ideal transmission channel
– Accepts input at rate source produces
– Delivers output at same rate as input

Quantitative Network Performance
Needed For Real-Time Streaming
d QoS type: Constant Bit Rate (CBR)
d Throughput sufficient to accommodate sender’s data rate
(known in advance)
d Latency within a specified bound, usually 200 msec
d Jitter of zero or near-zero

Buffering
d Especially important in a packet transmission system
d Combines multiple samples into a single transmission
d Advantage
– Increases transmission efficiency
d Disadvantage
– Introduces delay

Buffering Example
d Consider PCM audio
d One eight-bit audio sample taken every 125 µseconds
d Ethernet has 1500 octet payload
d Waiting to fill an entire frame takes
125 × 10−6 seconds/byte × 1500 bytes = 0.188 seconds
d Filling a packet incurs delay at the source

Buffering Compromise
d Choose buffer size according to application
d Example: send 128 audio samples in each packet
d Tradeoffs
– Packet size is larger than one sample per packet, but
generates more packets than absolutely necessary
– Header overhead is a smaller percentage of total bits
than with one sample per packet, but a greater
percentage than for larger packets
– Latency is better than with many samples per packet, but
not as good as with one sample per packet

Streaming Of Real-Time Data
Across The Internet
d Must handle
– Lost packets
– Duplicated packets
– Packets delivered out of order
– Variance in delay (jitter)
d Key facts
– Conventional retransmission is useless
– Jitter is unavoidable

Two Useful Techniques
d Timestamps
– Provided by sender
– Assigned to each piece of data
– Allow receiver to know when data should be played
– Use relative values to avoid need for clock
synchronization
d Jitter buffer
– Used by receiver
– Accommodates small variance in delay

Jitter Buffer
d Used by receiver to assemble incoming real-time data
d Timestamp on an item determines where item is placed in
the playback sequence
d General principle: ensure information will be available in
time to play without delay
d Trick: to compensate for maximum jitter of d, delay
playback for d time units
d Result: jitter buffer holds just enough data so playback can
proceed uninterrupted

Illustration Of A Jitter Buffer
display
jitter buffer
playback
process
connection
to Internet
packets arrive
in bursts
packets extracted
at a uniform rate
d
d During normal operation, playback can continue for d time
units while waiting for delayed packets

Real-Time Transport Protocol (RTP)
d Widely used for voice and video
d Despite the name, not really a transport protocol
d Does not contain a jitter buffer and does not control
playback
d Provides three basic mechanisms
– Sequence number on each packet that allows a receiver
to handle loss and out-of-order delivery
– Timestamp used for playback of the data
– Series of source identifiers that tell a receiver the
source(s) of the data

RTP Details
d Allows sender and receiver to choose sample rate and
encoding
d Specifies a header for each message transferred
d Uses UDP for transport
d Separates timestamp from packet sequence number
d Includes a marker bit that allows some frames to be marked
d Companion protocol allows receivers to inform sender about
transfer

Motivation For RTP Design
d Marking
– Permits differential encoding with a full frame followed
by incremental changes
– Example use: video I-frame followed by B-frames
d Separation of timestamp and packet sequence
– Means timestamps do not need to be linearly related to
packets
– Allows compression schemes that vary the rate at which
data is sent

RTP Header Format
0 1 3 8 16 31
VER P X CC M PAYTYPE SEQUENCE NUMBER
TIMESTAMP
SYNCHRONIZATION SOURCE IDENTIFIER
CONTRIBUTING SOURCE IDENTIFIER
. . .
d TIMESTAMP is interpreted by sender and receiver
d PAYTYPE specifies the payload type
d Initial SEQUENCE NUMBER chosen at random
d CONTRIBUTING SOURCE IDENTIFIERS allow sender to
mix streams from multiple sources

RTP Encapsulation
d Three levels of encapsulation
UDP Payload
UDP Hdr
RTP Payload
RTP Hdr
d Use of UDP permits sending one multicast instead of
multiple unicast copies

IP Telephony
d Known as Voice over IP (VoIP)
d Two groups have created standards
– International Telecommunications Union (ITU)
– Internet Engineering Task Force (IETF)
d Standards agree on two basics
– Audio encoded using Pulse Code Modulation (PCM)
– RTP used to transfer digitized audio
d Standards disagree on
– Signaling
– Public Switched Telephone Network (PSTN) interaction

Signaling
d Telco term for the process of establishing and terminating a
call
d Includes
– Mapping a phone number to a location
– Finding a route to the called party
– Recording information used for accounting and billing
– Handling functions such as call forwarding
d Standard call management facility for the traditional
telephone system is known as Signaling System 7 (SS7)

IETF Approach
d Known as Session Initiation Protocol (SIP)
d Domain Name System used to map a telephone number to
an IP address
d SIP signaling system
– User agent makes or terminates calls (e.g., an IP phone)
– Location server consults a database of users, services to
which they subscribe, and preferences
– Proxy server forwards requests and optimizes routing
– Redirect server handles tasks such as call forwarding
and 800-number connections
– Registrar server allows users to register for service

ITU Approach
d Standard is H.323
d Differs substantially from terminology used by SIP
d Terminal provides IP telephone functions and may also
include facilities for video and data transmission
d Gatekeeper provides location and signaling functions, and
establishes connections to the PSTN
d Gateway interconnects the IP phone system and PSTN, and
handles both signaling and media translation
d Multipoint Control Unit (MCU) provides services such as
multipoint conferencing

International Softswitch Consortium (ISC)
d Formed by vendors to consolidate terminology from
multiple standards and create a single conceptual model
d Defined a list of 10 functions that are sufficient to explain
all others
d Invented new terms for each function

Summary Of VoIP Protocols And Layering
Layer Call User User Support Routing Signal
Process. multimedia Data Transport
5
H.323 RTP T.120 RTCP ENUM SIGTRAN
Megaco RTSP TRIP
MGCP NTP
SIP SDP
4 TCP UDP TCP TCP SCTP
UDP UDP
3 IP, RSVP, and IGMP
d Each protocol can be complex
d H.323 is an umbrella

H.323
d Large set of protocols collected together
d Provides voice, video, and data transfer
d Summary of major protocols
Layer Signaling Registration Audio Video Data Security
5 H.225.0-Q.931 H.225.9-RAS G.711 H.261 T.120 H.235
H.250-Annex G H.263 H.323
H.245 G.722
H.250 G.723
G.728
RTP, RTCP
4 TCP, UDP UDP TCP TCP, UDP
3 IP, RSVP, and IGMP

Telephone Number Mapping And Routing
d Two standards proposed by IETF
– TRIP relies on location servers to exchange information
– ENUM (E.164 NUMbers) uses arpa top-level domain in
the Domain Name System
d ENUM example
– Phone number is 1-800-555-1234
– Domain name is constructed as the string
4.3.2.1.5.5.5.0.0.8.1.e164.arpa

Network Security
d Large subject with many aspects
d Major problems include
2
222222222222222222222222222222222222222222222222222222222222222222222222
Problem Description
2
222222222222222222222222222222222222222222222222222222222222222222222222
Phishing Masquerading as a well-known site such as a bank
to obtain a user’s personal information, typically an
account number and access code
2
222222222222222222222222222222222222222222222222222222222222222222222222
Misrepresentation Making false or exaggerated claims about goods or
services, or delivering fake or inferior products
2
222222222222222222222222222222222222222222222222222222222222222222222222
Scams Various forms of trickery intended to deceive naive
users into investing money or abetting a crime
2
222222222222222222222222222222222222222222222222222222222222222222222222
Denial of Service Intentionally blocking a particular Internet site to
prevent or hinder business activities and commerce
2
222222222222222222222222222222222222222222222222222222222222222222222222
Loss of Control An intruder gains control of a computer system
and uses the system to perpetrate a crime
2
222222222222222222222222222222222222222222222222222222222222222222222222
Loss of Data Loss of intellectual property or other valuable
proprietary business information
2
222222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Examples Of Techniques Attackers Use
2
222222222222222222222222222222222222222222222222222222222222222222222222
Technique Description
2
222222222222222222222222222222222222222222222222222222222222222222222222
Wiretapping Making a copy of packets
2
222222222222222222222222222222222222222222222222222222222222222222222222
Replay Sending packets captured from a previous session
2
222222222222222222222222222222222222222222222222222222222222222222222222
Buffer Overflow Overflowing a memory buffer to overwrite values
2
222222222222222222222222222222222222222222222222222222222222222222222222
Address Spoofing Faking the IP source address in a packet
2
222222222222222222222222222222222222222222222222222222222222222222222222
Name Spoofing Using a misspelling of a well-known name
2
222222222222222222222222222222222222222222222222222222222222222222222222
DoS and DDoS Flooding a site with packets to prevent access
2
222222222222222222222222222222222222222222222222222222222222222222222222
SYN Flood Sending a stream of random TCP SYN segments
2
222222222222222222222222222222222222222222222222222222222222222222222222
Key Breaking Guessing a decryption key or password
2
222222222222222222222222222222222222222222222222222222222222222222222222
Port Scanning Probing ports to find a vulnerable application
2
222222222222222222222222222222222222222222222222222222222222222222222222
Packet Interception Removing a packet from the Internet
2
222222222222222222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Indirect Attacks
d Attacker commandeers computers of unwitting users
d Bots running on commandeered computers launch attack
d Example: Distributed Denial of Service (DDoS)
Internet
attacker commandeers
multiple computers
and streams packets
to target
aggregate traffic
overwhelms server

Packet Interception
d Extreme vulnerability
d Can be exploited for many attacks
d Permits man-in-the-middle attacks
d Example attacks
source
server man-in-the-middle
can wiretap, replay, spoof,
break keys, scan ports, and
impersonate a server
can impersonate a host or
pass altered packets on to
any Internet destination

Security Policy
d No absolutely secure network exists
d Before security mechanisms are meaningful, organization
must define a security policy
– Data integrity (no unauthorized change)
– Data availability (no disruption of service)
– Data confidentiality (no unauthorized access)
– Privacy (no disclosure of sender’s identity)
– Accountability (record keeping and audit trail)
– Authorization (who is permitted to access information)

Authorization And Authentication
d Authorization is intertwined with authentication
– Authorization meaningless without authentication
– Must know identity of a requester
d There is no point in defining a security policy that cannot be
enforced

Enforcement Mechanisms
2
2222222222222222222222222222222222222222222222222222222222222222222222
Technique Purpose
2
2222222222222222222222222222222222222222222222222222222222222222222222
Hashing Data integrity
2
2222222222222222222222222222222222222222222222222222222222222222222222
Encryption Confidentiality
2
2222222222222222222222222222222222222222222222222222222222222222222222
Digital Signatures Message authentication
2
2222222222222222222222222222222222222222222222222222222222222222222222
Digital Certificates Sender authentication
2
2222222222222222222222222222222222222222222222222222222222222222222222
Firewalls Site integrity
2
2222222222222222222222222222222222222222222222222222222222222222222222
Intrusion Detection Systems Site integrity
2
2222222222222222222222222222222222222222222222222222222222222222222222
Deep Packet Inspection & Content Scanning Site integrity
2
2222222222222222222222222222222222222222222222222222222222222222222222
Virtual Private Networks (VPNs) Data confidentiality and
trusted access
2
2222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Hash
d Used to guarantee message arrives with no
– Changes
– Additions
d Sender and receiver share a key
d Sender uses key to compute a small value, H, called a
– Message Authentication Code (MAC)
– Hash of the message
d Sender transmits H with the message
d Receiver uses same key to compute hash of received
message and compares to H

Encryption
d Fundamental security technique
d Predates computers and computer networks
d Extensive mathematical analysis
d Definitions
– Plaintext: original, unencrypted message
– Cyphertext: message after encryption
– Encryption key: short bit string used for encryption
– Decryption key: short bit string used for decryption
d Note: in some schemes, the encryption and decryption keys
differ; in others, they are identical

Mathematics Of Encryption
d Encryption and decryption viewed as functions
d Encrypt takes key, K1, and plaintext message, M, as
arguments and produces cyphertext, C, as a result
C = encrypt ( K1 , M )
d Decrypt takes a key, K2, and cyphertext, C, as arguments,
and produces a plaintext message, M, as a result
M = decrypt ( K2 , C )
d Mathematically, decrypt is the inverse of encrypt
M = decrypt ( K2 , encrypt ( K1 , M ) )

Two Main Types Of Encryption
d Private or secret key encryption (symmetric)
– Encryption and decryption use same key
– Key is a shared secret
M = decrypt ( K , encrypt ( K , M ) )
d Public key encryption (asymmetric)
– Encryption and decryption use different keys
– Public key is widely disseminated
– Private key is known only to one party
– Knowing a user’s public key does not help one guess the
corresponding private key

Authentication With Digital Signatures
d Uses encryption (works well with public key methods)
d Allows receiver to verify the identity of the sender
d Example
– Bob sends message to Alice
* Uses his private key to encode message
* Includes specific information such as Alice’s name
and a date to avoid a replay attack
– Alice
* Uses Bob’s public key to decrypt message
* Knows that only Bob could have sent the message

Authentication With Digital Signatures
(continued)
d Can use additional level of encryption to guarantee
confidentiality
d Bob signs message and encrypts using Alice’s public key
X = encrypt ( alice_pub , encrypt ( bob_priv, M ) )
d Alice decrypts message with her private key, and then
authenticates the sender by decrypting with Bob’s public
key
M = decrypt ( bob_pub , decrypt ( alice_priv , X ) )

Key Distribution
d Everyone needs to obtain a copy of each user’s public key
d If an attacker distributes an incorrect key, the entire
encryption scheme is compromised
d Question: how can public keys be distributed in a way that
guarantees each copy is correct?
d Several solutions have been proposed; most rely on key
authority organizations that hand out public keys
d Message containing keys signed by well-known authority is
a digital certificate
d Note: knowing the public key of an authority makes it
possible to obtain other public keys securely

Firewall Technology
d Inserted between site and Internet
d Filters packets according to policy
d Controls both incoming and outgoing traffic
d General approach: prevent all communication unless
explicitly allowed by policy

Firewall Example
d Consider a site with three servers
web sever
( 192.5.48.1 )
email sever
( 192.5.48.2 )
DNS sever
( 192.5.48.3 )
Internet
firewall in switch
d Firewall only allows packets to/ from each server

Firewall Example
(continued)
d Example of firewall rules for the site:
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
Dir Frame Type IP Src IP Dest IP Type Src Port Dst Port
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
in 0800 * 192.5.48.1 TCP * 80
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
in 0800 * 192.5.48.2 TCP * 25
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
in 0800 * 192.5.48.3 TCP * 53
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
in 0800 * 192.5.48.3 UDP * 53
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
out 0800 192.5.48.1 * TCP 80 *
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
out 0800 192.5.48.2 * TCP 25 *
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
out 0800 192.5.48.3 * TCP 53 *
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
out 0800 192.5.48.3 * UDP 53 *
2
2222222222222222222222222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Other Network Security Systems
d Intrusion Detection System (IDS)
– Watches incoming packet stream
– Attempts to identify unusual activity
d Deep Packet Inspection (DPI)
– Looks beyond header into packet contents
– Requires significant processing
d File inspection systems
– Examine whole data file (e.g., email)
– Can detect more problems than systems that examine
individual packets

Virtual Private Network (VPN)
d Emulates a dedicated network connection
d Sends traffic across commodity Internet
d Uses encryption to guarantee confidentiality
d Technique known as tunneling
d Can be used
– Among sites of an organization
– Between individual and organization

Encryption And Tunneling Used In VPNs
d Three basic approaches used
– Payload encryption
– IP-in-IP tunneling
– IP-in-TCP tunneling
d Original data is encrypted in all three
d For additional security, pad datagram length

Illustration Of IP-in-IP Tunneling
Used For A Secure VPN
src = R1
dst = R2
Encrypted Datagram Encapsulated For Transmission
Encrypted Version Of Original Datagram
src = X
dst = Y
Original (Unencrypted) Payload
encrypt

Examples Of Security Technologies
d PGP (Pretty Good Privacy)
d SSH (Secure Shell)
d SSL (Secure Socket Layer)
d TLS (Transport Layer Security)
d HTTPS (HTTP Security)
d IPsec (IP security)
d RADIUS (Remote Authentication Dial-In User Service)
d WEP (Wired Equivalent Privacy)
d WPA (Wi-Fi Protected Access)

Terminology
d Network manager or network administrator is a person
responsible for network
– Planning
– Installation
– Operation
– Monitoring
d Network refers to intranet
– Owned and operated by a single organization
– Contains many managed items such as routers, switches,
servers, and hosts
– May span multiple sites

An Interesting Problem
d Many protocol mechanisms have been created to overcome
problems automatically
– Forward error correction
– Retransmission
– Routing protocols
d Consequence: protocols may hide problems from a manager!

The Industry Standard Model
d Derived from ITU recommendation M.3400
d Known by abbreviation, FCAPS
d Acronym refers to five aspects of management
2
2222222222222222222222222222222222222222222222222222222222
Abbreviation Meaning
2
2222222222222222222222222222222222222222222222222222222222
F Fault detection and correction
2
2222222222222222222222222222222222222222222222222222222222
C Configuration and operation
2
2222222222222222222222222222222222222222222222222222222222
A Accounting and billing
2
2222222222222222222222222222222222222222222222222222222222
P Performance assessment and optimization
2
2222222222222222222222222222222222222222222222222222222222
S Security assurance and protection
2
2222222222222222222222222222222222222222222222222222222222
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1

Fault Isolation And Root-Cause Analysis
d Users report high-level symptoms
– Example: I lost access to a shared file system
d Manager must relate symptoms to underlying cause
– Cable cut
– Power supply has failed or disk has crashed
– Software configuration changed (e.g., file system renamed
or moved)
– Security changed (e.g., password expired)

Network Element
d Generic term for a managed entity
– Physical device
– Service (e.g., DNS)
d Examples
222222222222222222222222222222222222222222222222222
Manageable Network Elements
222222222222222222222222222222222222222222222222222
Layer 2 Switch IP router
222222222222222222222222222222222222222222222222222
VLAN Switch Firewall
222222222222222222222222222222222222222222222222222
Wireless Access Point Digital Circuit (CSU/DSU)
222222222222222222222222222222222222222222222222222
Head-End DSL Modem DSLAM
222222222222222222222222222222222222222222222222222
DHCP Server DNS Server
222222222222222222222222222222222222222222222222222
Web Server Load Balancer
222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

Element Management System
d Management tool that can manage one element at a time
d Typically, supplied by vendor of the network element
d Limitation of element management systems
– When configuring MPLS tunnel across multiple routers,
element management system only allows manager to
configure one router at a time
– If routers sold by multiple vendors, each vendor may have
its own element management system
d Unfortunately, many networks only have element management

Types Of Network Management Tools
Physical Layer Testing Performance Monitoring
Reachability And Connectivity Flow Analysis
Packet Analysis Routing And Traffic Engineering
Network Discovery Configuration
Device Interrogation Security Enforcement
Event Monitoring Network Planning

How Should Management Systems Operate?
d Some possibilities
– Use a parallel physical network
– Use a parallel logical network
– Use a special link-layer protocol
– Use the same links, equipment, and protocols as data
d Surprise: modern network management follows the last
approach

Simple Network Management Protocol (SNMP)
d Internet standard
d Allows software in a manager’s computer (manager) to interact
with software that runs in an element (agent)
d Specifies format and meaning of messages exchanged
d Runs as an application protocol over TCP or UDP
d Uses fetch-store paradigm

SNMP Fetch-Store Paradigm
d Set of conceptual variables defined
d Each variable given a name
d Set of variables known as Management Information Base
(MIB)
d SNMP offers two basic operations
– GET to read the value of a variable
– PUT to store a value into a variable
d All management functions are defined as side-effects of GET or
PUT to a MIB variable
d Example: reboot defined as side-effect of PUT

SNMP Encoding
d SNMP uses a standard known as Abstract Syntax Notation.1
(ASN.1)
d Variable-length encoding
d Example: integer encoded as length and value
2
222222222222222222222222222222222222222222222222222222
Decimal Hexadecimal Length Bytes Of Value
Integer Equivalent Byte (in hex)
2
222222222222222222222222222222222222222222222222222222
27 1B 01 1B
2
222222222222222222222222222222222222222222222222222222
792 318 02 03 18
2
222222222222222222222222222222222222222222222222222222
24,567 5FF7 02 5F F7
2
222222222222222222222222222222222222222222222222222222
190,345 2E789 03 02 E7 89
2
222222222222222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

MIB Variable Names
d Are hierarchical
d Begin with standard prefix
d Identify a specific protocol and variable
d Example: counter for IP packets received has name
iso.org.dod.internet.mgmt.mib.ip.ipInReceives
d Name is encoded as integers:
1.3.6.1.2.1.4.3

Arrays In A MIB
d ASN.1 does not define an array type
d Many MIB variables correspond to conceptual array
– Routing table
– ARP cache
– Set of network interfaces
d Trick
– The “index” is appended onto variable name
– Manager software uses GET-NEXT operation to move
through array

Example Of Indexing
d IP routing table assigned variable name
standard-prefix.ip.ipRoutingTable
d Each field has a name
d Issuing GET_NEXT operation gets first routing table entry
d For example, name of destination address field variable is
standard-prefix.ip.ipRoutingTable.ipRouteEntry.field.IPdestaddr

A Plethora Of MIBs
d Initially
– One MIB
– Defined variables for IP, TCP, UDP, ICMP
d Now
– Many MIBs
– Variables for routers, switches, modems, printers, hosts,
and other network elements

Summary
d Streamed transfer of real-time data incompatible with Internet’s
best-effort delivery
d Two approaches
– Isochronous network
– Timestamps and jitter buffer
d Real-Time Transport Protocol (RTP) uses timestamps and
sequence numbers

Summary
(continued)
d Many IP telephony standards proposed
d Connection to PSTN causes debate
d H.323 and SIP standards are most widely used
d ENUM system uses DNS to convert phone number to IP
address

Summary
(continued)
d Quantitative measures of networks include delay, throughput,
goodput, and jitter
d Delay increases as utilization increases
d One can purchase more throughput, but not less delay
d Quality of Service (QoS) technologies provide guarantees on
performance
d The industry has moved away from fine-grain QoS (per-flow as
in ATM and IntServ) to coarse-grain QoS (DiffServ)
d Multi-Protocol Label Switching (MPLS) is used by tier-1 ISPs
to provide circuit-oriented networking

Summary
(continued)
d Network security is complex and difficult
d No network is completely secure
d Life goes on anyway
d Network management is complex and difficult
d Current tools are fairly primitive
d Life goes on anyway

MODULE VII
Emerging Technologies

Topics
d Software Defined Networking
d The Internet Of Things
d Other trends in networking

Software Defined Networking
(SDN)

What Is Software Defined Networking?
d One of the hottest topics in networking
d According to marketing SDN is
– A way to eliminate all human error
– A technology that improves overall routing
– An approach that eliminates 66% to 80% of operational
costs
d In reality SDN is
– A technology that gives programmers more control over
network equipment
– An approach with the potential to make some
improvements in network configuration and management

Motivation For SDN
d Switch from element management to network management
d Move from proprietary to open standards
d Automate and unify network-wide configuration
d Change from per-layer to cross-layer control
d Accommodate virtualization used in data centers

Background And Definitions
d Terminology adopted from network equipment engineers
d Data plane
– Refers to packet processing mechanisms
– Typical functions include packet classification and
packet forwarding
– Operates at wire speed
d Control plane
– Refers to management
– Typical functions include interacting with network
manager and modifying forwarding tables
– Operates slowly and only when changes are needed

Conceptual Organization Of Network Devices
control plane
(software)
data plane
(hardware)
control plane loads
new configuration
into the hardware
data plane passes
management packets
up to control plane
packets
arrive
packets
leave
d Data plane may use ASIC hardware for speed
d Control plane includes a TCP/IP stack

Control Plane Interface Modules
CLI WEB SNMP
data plane
(hardware)
common interface (internal)
...
packets
arrive
packets
leave
d Managers can choose among command line interface, web
interface, and SNMP

The SDN Approach: An External Controller
CLI WEB SNMP sdn
data plane
(hardware)
common interface (internal)
packets
arrive
packets
leave
External
Controller
PC running Linux

In Practice
Controller 1 Controller 2 Controller 3 Controller 4 Controller 5
Domain 1 Domain 2 Domain 3 Domain 4 Domain 5
controller
to controller
controller
to element
d Each controller can operate multiple devices
d Controllers coordinate to provide consistent configuration

SDN Communication
d Two conceptually separate types
– Controller to network element
– Controller to controller
d Protocols used can differ

OpenFlow
d Specification for controller-to-element communication
d Devised at Stanford
d Now a de facto industry standard for SDN
d Defines
– Secure communication (over SSL)
– Message format
– Items to be managed
d Completely unlike SNMP

OpenFlow Model
d Uses flow table abstraction
– Data plane is assumed to have a sequence of flow tables
– Each flow table specifies how to parse packets and
handle them
d OpenFlow allows manager to set values in each flow table
d Important note: flow table model closely matches
classification hardware found in Ethernet switches

Classification
d Alternative to packet demultiplexing
d Examines headers from multiple layers at the same time
d Uses an array of pairs
(pattern, action)
d Where
– Pattern is a pattern that is matched against packets
– Action specifies steps to be taken if the match succeeds

Classification Hardware
classification engine
packet to be matched
pattern 1
pattern 2
pattern 3
pattern N (default that matches any packet)
action 1
action 2
action 3
action N
.
.
.
.
.
.
d Hardware checks all patterns in parallel
d Result is extremely high speed classification

TCAM
d Acronym for Ternary Content Addressable Memory
d Hardware technology used for high-speed classification
d Pattern is ternary because value for each bit can be 0, 1, or
“don’t care”
d TCAM matches all patterns at once, and performs the action
on the first matching table entry

Example Of IPv4 Classification
d The challenge
– A frame arrives
– What is the minimum number of steps needed to
determine whether the frame carries an IPv4 datagram
destined for a web server?
d The answer
– Check whether the frame type field specifies IPv4
(0x0800)
– Check whether the IP protocol field specifies TCP (6)
– Check whether the TCP destination port specifies a web
server (80)

IPv6 Classification
d Simplest case (only a base header)
– Frame type field specifies IPv6 (0x86DD)
– Next Header field specifies TCP (6)
– TCP destination port specifies a web server (80)
d Additional patterns needed for extension headers
d Example: base header plus a route header
– Frame type field specifies IPv6 (0x86DD)
– Next Header field specifies Route Header (43)
– Next Header field specifies TCP (6)
– TCP destination port specifies a web server (80)

Example Items In An OpenFlow Pattern
Field Meaning
2
22222222222222222222222222222222222222222222222222222222222222222
Layer 2 fields
Ingress Port Switch port over which the packet arrived
Metadata 64-bit field of metadata used in the pipeline
Ether src 48-bit Ethernet source address
Ether dst 48-bit Ethernet destination address
Ether Type 16-bit Ethernet type field
VLAN id 12-bit VLAN tag in the packet
VLAN priority 3-bit VLAN priority number
ARP opcode 8-bit ARP opcode
Layer 3 fields
MPLS label 20-bit MPLS label
MPLS class 3-bit MPLS traffic class
IPv4 src 32-bit IPv4 source address
IPv4 dst 32-bit IPv4 destination address
IPv6 src 128-bit IPv6 source address
IPv6 dst 128-bit IPv6 destination address
IPv4 Proto 8-bit IPv4 protocol field
IPv6 Next Header 8-bit IPv6 next header field
TOS 8-bit IPv4 or IPv6 Type of Service bits

Example Items In An OpenFlow Pattern
(continued)
Field Meaning
2
22222222222222222222222222222222222222222222222222222222222222
Layer 4 fields
TCP/UDP/SCTP src 16-bit TCP/UDP/SCTP source port
TCP/UDP/SCTP dst 16-bit TCP/UDP/SCTP destination port
ICMP type 8-bit ICMP type field
ICMP code 8-bit ICMP code field

Examples Of SDN Functionality
d End-to-end layer 2 paths
d Forwarding based on source as well as destination
d All traffic from a specific MAC address sent along a specific
path
d Segregation of traffic based on application type
d Multipath forwarding based on hash of 4-tuple
d Transport of nonstandard layer 3 protocols

Internet Of Things
d Awkward term used for embedded systems on the Internet
– Generally not operated by humans
– Can access one another or cloud services
d Examples
– Scientific sensor systems
– Home automation systems
– Smart grid
– Retail systems

Technology Characteristics
d Low power
– Energy harvesting (e.g., door latch)
– Multi-year battery life
d Wireless communication
– Necessary in many situations
– Enables mobility

Wireless Mesh Network
d Useful when individual nodes have very low power (limited
range)
d Allows a set of nodes to communicate even if some nodes
cannot communicate directly
d Each node agrees to forward packets on behalf of neighbors

Example Wireless Mesh Technology
d ZigBee IP
– Created by ZigBee Alliance
– Uses IEEE 802.15.4 wireless radios
– Intended for smart grid
d ZigBee protocol stack
– Goal is to run IPv6, TCP, and HTTP
– Includes many other protocols

802.15.4 Wireless Characteristics
d Goal is low power, and result is
– Extremely low data rate
– Extremely small MTU
– Limited distance
222222222222222222222222222222222222222222
Property Value
222222222222222222222222222222222222222222
Networking paradigm Packet switching
222222222222222222222222222222222222222222
Maximum data rate 250 Kbps
222222222222222222222222222222222222222222
Payload size (MTU) 102 octets
222222222222222222222222222222222222222222
Maximum distance 10 meters
222222222222222222222222222222222222222222
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

ZigBee IP Mesh Routing
d One or more border routers
– Connect to global Internet
– Are more powerful than other nodes
d Set of ZigBee IP routers (ZIP routers)
– Attach to appliances
– Form a mesh
– Forward traffic to border router

Choosing A Path
d ZIP router must choose a path to a border router
d Cannot simply choose node with the strongest transmission
signal
B1 B2
N
Internet
d Additional protocol used to find which node receives
strongest signal (MLE)

Running IPv6 Over A ZigBee Network
d IPv6 can only run over networks that offer an MTU of 1280
or higher, but 802.15.4 has an MTU of 102
d Solution
– Additional protocol named 6LoWPAN
– Shim layer between IP and device driver

6LoWPAN Operation
d Sending side
– Divides datagram into series of blocks
– Transmits each block in a packet
d Receiving side
– Joins blocks into a datagram
– Delivers entire datagram to IPv6
d Notes
– Division into block does not use IP fragmentation
– Unlike fragmentation, division and regrouping is
performed at each hop

ZigBee IP Mesh Routing
d ZIP nodes forward packets toward the border router
d Border router
– Can send outgoing packets to the Internet
– Forwards other packets across the mesh
d If two ZIP nodes communicate
– Packet goes to border router first
– Border router forwards to destination

Border Router Operation
d To forward across the mesh the border router
– Learns the topology of the mesh
– Computes a path through the mesh to each ZIP node
– Uses IPv6 source routing
d IPv6 source routing
– Requires IP-in-IP tunneling (header modification
prohibited)
– Places an extension header on outer datagram with series
of hops
– Each ZIP node only needs to know its neighbors

Computing Source Routes
d All nodes run Routing Protocol for lossy and Low power
networks (RPL)
d Each node reports its parent to the border router
d RPL code on border router creates a Destination Oriented
Directed Acyclic Graph (DODAG)
d DODAG is used to compute source routes

Example DODAG
1
2
3 4
5
6
7
8
9
Border router
d Arcs in DODAG point to parent (path toward border router)
d Source route to node X is reverse of the path from X to
border router

Does ZigBee IP Make Sense?
d Choosing IPv6 instead of IPv4 means
– Much larger datagram headers
– The use of 6LoWPAN to divide a datagram into MTU-
size pieces
– Sending more data over a slow network
– The need for RPL routing protocols
– Larger memories (and lower battery life)
d Using TCP and HTTP over IPv6 means
– Using DNS to resolve names
– Unnecessary overhead
– Unnecessary memory footprint

But Wait, There’s More!
d Smart grid applications must be secure, so ZigBee IP
includes security protocols, including TLS
d IPv6 Neighbor Discovery doesn’t work in a mesh network,
so ZigBee IP includes a modification known as 6LoWPAN-
ND
d IEEE 802.15.4 allows short (16-bit) MAC addresses, so
ZigBee IP includes a mechanism that allows a border router
to prevent address collisions

Major Items In The ZigBee Protocol Stack
IEEE 802.15.4
6LoWPAN adaptation
IPv6, ICMPv6, and 6LoWPAN-ND RPL
TCP and UDP
TLS PANA mDNS and DNS-SD MLE
Application Protocols
d Resulting stack is large
d Design is more general-purpose than necessary
d Technology may be a triumph of politics and economics

A Few Key Technologies
d Content Caching
d Peer-To-Peer Communication
d Universal Representation (XML)
d Wireless networks that support mobility
d Higher-speed access technologies (1 Gbps)
d Cloud computing and cloud data centers

Web Load Balancers
.
.
.
Internet
connection
load balancer
physical severs
site running a web server
shared
database
d Load balancer distributes HTTP requests across servers
d Path from servers back to client may be higher speed

Overlay Networking
A
B
C
D
E
F
G
H
I
A
B
C
D
E
F
G
H
I
Internet
(a) (b)
(a) Physical connection of computers to the Internet
(b) Logical network imposed by overlay routing

Other Trends
d Switch to digital telephony and digital video
d Increased use of social networking and social media
d Distributed data centers and migration

Lecture_Notes.pdf

More Related Content

Similar to Lecture_Notes.pdf (20)

Recently uploaded (20)

Lecture_Notes.pdf