Chapter 3 v6.0

Transport Layer 3-1
Chapter 3
Transport Layer
Computer
Networking: A Top
Down Approach
6th
edition
Jim Kurose, Keith Ross
Addison-Wesley
March 2012
A note on the use of these ppt slides:
We’re making these slides freely available to all (faculty, students, readers).
They’re in PowerPoint form so you see the animations; and can add, modify,
and delete slides (including this one) and slide content to suit your needs.
They obviously represent a lot of work on our part. In return for use, we only
ask the following:
 If you use these slides (e.g., in a class) that you mention their source
(after all, we’d like people to use our book!)
 If you post any slides on a www site, that you note that they are adapted
from (or perhaps identical to) our slides, and note our copyright of this
material.
Thanks and enjoy! JFK/KWR
All material copyright 1996-2012
J.F Kurose and K.W. Ross, All Rights Reserved

Transport Layer 3-2
Chapter 3: Transport Layer
our goals:
 understand
principles behind
transport layer
services:
 multiplexing,
demultiplexing
 reliable data transfer
 flow control
 congestion control
 learn about Internet
transport layer protocols:
 UDP: connectionless
transport
 TCP: connection-oriented
reliable transport
 TCP congestion control

Transport Layer 3-3
Chapter 3 outline
3.1 transport-layer
services
3.2 multiplexing and
demultiplexing
3.3 connectionless
transport: UDP
3.4 principles of reliable
data transfer
3.5 connection-oriented
transport: TCP
 segment structure
 flow control
 connection management
3.6 principles of congestion
control
3.7 TCP congestion control

Transport Layer 3-4
Transport services and protocols
 provide logical communication
between app processes
running on different hosts
 transport protocols run in
end systems
 send side: breaks app
messages into segments,
passes to network layer
 rcv side: reassembles
segments into messages,
passes to app layer
 more than one transport
protocol available to apps
 Internet: TCP and UDP
application
transport
network
data link
physical
logicalend-end
transport
application
transport
network
data link
physical

Transport Layer 3-5
Transport vs. network layer
 network layer: logical
communication
between hosts
 transport layer:
logical
communication
between processes
 relies on, enhances,
network layer
services
12 kids in Ann’s house sending
letters to 12 kids in Bill’s
house:
 hosts = houses
 processes = kids
 app messages = letters in
envelopes
 transport protocol = Ann
and Bill who demux to in-
house siblings
 network-layer protocol =
postal service
household analogy:

Transport Layer 3-6
Internet transport-layer protocols
 reliable, in-order
delivery (TCP)
 flow control
 connection setup
 unreliable, unordered
delivery: UDP
 no-frills extension of
“best-effort” IP
 services not available:
 delay guarantees
 bandwidth guarantees
 Not possible at the
transport layer alone
application
transport
network
data link
physical
application
transport
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
network
data link
physical
logicalend-end
transport

Transport Layer 3-7
Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

Transport Layer 3-8
Multiplexing/demultiplexing
process
socket
use header info to deliver
received segments to correct
socket
demultiplexing at receiver:handle data from multiple
sockets, add transport header
(later used for demultiplexing)
multiplexing at sender:
transport
application
physical
link
network
P2P1
transport
application
physical
link
network
P4
transport
application
physical
link
network
P3

Transport Layer 3-9
How demultiplexing works
 host receives IP datagrams
 each datagram has source IP
address, destination IP
address
 each datagram carries one
transport-layer segment
 each segment has source,
destination port number
 host uses IP addresses &
port numbers to direct
segment to appropriate
socket
source port # dest port #
32 bits
application
data
(payload)
other header fields
TCP/UDP segment format

Transport Layer 3-10
Connectionless demultiplexing
 recall: created socket has
host-local port #:
DatagramSocket mySocket1
= new
DatagramSocket(12534);
 when host receives UDP
segment:
 checks destination port #
in segment
 directs UDP segment to
socket with that port #
 recall: when creating
datagram to send into
UDP socket, must specify
 destination IP address
 destination port #
IP datagrams with same
dest. port #, but different
source IP addresses
and/or source port
numbers will be directed
to same socket at dest

Connectionless demux: example
DatagramSocket
serverSocket = new
DatagramSocket
(6428);
transport
application
physical
link
network
P3
transport
application
physical
link
network
P1
transport
application
physical
link
network
P4
DatagramSocket
mySocket1 = new
DatagramSocket
(5775);
DatagramSocket
mySocket2 = new
DatagramSocket
(9157);
source port: 9157
dest port: 6428
source port: 6428
dest port: 9157
source port: ?
dest port: ?
source port: ?
dest port: ?

Connection-oriented demux
 TCP socket identified
by 4-tuple:
 source IP address
 source port number
 dest IP address
 dest port number
 demux: receiver uses
all four values to direct
segment to appropriate
socket
 server host may support
many simultaneous TCP
sockets:
 each socket identified by
its own 4-tuple
 web servers have
different sockets for
each connecting client
 non-persistent HTTP will
have different socket for
each request

Connection-oriented demux: example
transport
application
physical
link
network
P3
transport
application
physical
link
P4
transport
application
physical
link
network
P2
source IP,port: A,9157
dest IP, port: B,80
source IP,port: B,80
dest IP,port: A,9157
host: IP
address A
host: IP
address C
network
P6P5
P3
source IP,port: C,5775
dest IP,port: B,80
dest IP,port: B,80
three segments, all destined to IP address: B,
dest port: 80 are demultiplexed to different sockets
server: IP
address B

Connection-oriented demux: example
transport
application
physical
link
network
P3
transport
application
physical
link
transport
application
physical
link
network
P2
source IP,port: A,9157
dest IP, port: B,80
source IP,port: B,80
dest IP,port: A,9157
host: IP
address A
host: IP
address C
server: IP
address B
network
P3
dest IP,port: B,80
dest IP,port: B,80
P4
threaded server

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

UDP: User Datagram Protocol [RFC 768]
 “no frills,” “bare bones”
Internet transport
protocol
 “best effort” service, UDP
segments may be:
 lost
 delivered out-of-order
to app
 connectionless:
 no handshaking
between UDP sender,
receiver
 each UDP segment
handled independently
of others
 UDP use:
 streaming multimedia
apps (loss tolerant, rate
sensitive)
 DNS
 SNMP
 reliable transfer over
UDP:
 add reliability at
application layer
 application-specific error
recovery!

UDP: segment header
32 bits
application
data
(payload)
UDP segment format
length checksum
length, in bytes of
UDP segment,
including header
 no connection
establishment (which can
add delay)
 simple: no connection
state at sender, receiver
 small header size
 no congestion control:
UDP can blast away as
fast as desired
why is there a UDP?

UDP checksum
sender:
 treat segment contents,
including header fields,
as sequence of 16-bit
integers
 checksum: addition
(one’s complement sum)
of segment contents
 sender puts checksum
value into UDP
checksum field
receiver:
 compute checksum of
received segment
 check if computed
checksum equals checksum
field value:
 NO - error detected
 YES - no error detected.
But maybe errors
nonetheless? More later
….
Goal: detect “errors” (e.g., flipped bits) in transmitted
segment

Internet checksum: example
example: add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0
1 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 0
1 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sum
checksum
Note: when adding numbers, a carryout from the most
significant bit needs to be added to the result

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

Principles of reliable data
transfer important in application, transport, link layers
 top-10 list of important networking topics!
 characteristics of unreliable channel will determine
complexity of reliable data transfer protocol (rdt)

transfer important in application, transport, link layers

 important in application, transport, link layers
transfer

Reliable data transfer: getting started
send
side
receive
side
rdt_send(): called from above,
(e.g., by app.). Passed data to
deliver to receiver upper layer
udt_send(): called by rdt,
to transfer packet over
unreliable channel to receiver
rdt_rcv(): called when packet
arrives on rcv-side of channel
deliver_data(): called by
rdt to deliver data to upper

we’ll:
 incrementally develop sender, receiver sides of
reliable data transfer protocol (rdt)
 consider only unidirectional data transfer
 but control info will flow on both directions!
 use finite state machines (FSM) to specify sender,
receiver
state
1
state
2
event causing state transition
actions taken on state transition
state: when in this “state”
next state uniquely
determined by next
event
event
actions
Reliable data transfer: getting started

rdt1.0: reliable transfer over a reliable channel
 underlying channel perfectly reliable
 no bit errors
 no loss of packets
 separate FSMs for sender, receiver:
 sender sends data into underlying channel
 receiver reads data from underlying channel
Wait for
call from
above packet = make_pkt(data)
udt_send(packet)
rdt_send(data)
extract (packet,data)
deliver_data(data)
Wait for
call from
below
rdt_rcv(packet)
sender receiver

 underlying channel may flip bits in packet
 checksum to detect bit errors
 the question: how to recover from errors:
 acknowledgements (ACKs): receiver explicitly tells sender
that pkt received OK
 negative acknowledgements (NAKs): receiver explicitly tells
sender that pkt had errors
 sender retransmits pkt on receipt of NAK
 new mechanisms in rdt2.0 (beyond rdt1.0):
 error detection
 receiver feedback: control msgs (ACK,NAK) rcvr-
>sender
rdt2.0: channel with bit errors
How do humans recover from “errors”
during conversation?

 underlying channel may flip bits in packet
 checksum to detect bit errors
 the question: how to recover from errors:
 acknowledgements (ACKs): receiver explicitly tells sender
that pkt received OK
 negative acknowledgements (NAKs): receiver explicitly tells
sender that pkt had errors
 sender retransmits pkt on receipt of NAK
 new mechanisms in rdt2.0 (beyond rdt1.0):
 error detection
 feedback: control msgs (ACK,NAK) from receiver to
sender
rdt2.0: channel with bit errors

rdt2.0: FSM specification
Wait for
call from
above
sndpkt = make_pkt(data, checksum)
udt_send(sndpkt)
extract(rcvpkt,data)
deliver_data(data)
udt_send(ACK)
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) && isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) &&
corrupt(rcvpkt)
Wait for
ACK or
NAK
Wait for
call from
below
sender
receiver
rdt_send(data)
Λ

rdt2.0: operation with no errors
Wait for
call from
above
snkpkt = make_pkt(data, checksum)
udt_send(sndpkt)
deliver_data(data)
udt_send(ACK)
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) &&
corrupt(rcvpkt)
Wait for
ACK or
NAK
Wait for
call from
below
rdt_send(data)
Λ

rdt2.0: error scenario
Wait for
call from
above
snkpkt = make_pkt(data, checksum)
udt_send(sndpkt)
deliver_data(data)
udt_send(ACK)
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) &&
corrupt(rcvpkt)
Wait for
ACK or
NAK
Wait for
call from
below
rdt_send(data)
Λ

rdt2.0 has a fatal flaw!
what happens if
ACK/NAK corrupted?
 sender doesn’t know
what happened at
receiver!
 can’t just retransmit:
possible duplicate
handling duplicates:
 sender retransmits
current pkt if ACK/NAK
corrupted
 sender adds sequence
number to each pkt
 receiver discards (doesn’t
deliver up) duplicate pkt
stop and wait
sender sends one packet,
then waits for receiver
response

rdt2.1: sender, handles garbled ACK/NAKs
Wait for
call 0 from
above
sndpkt = make_pkt(0, data, checksum)
udt_send(sndpkt)
rdt_send(data)
Wait for
ACK or
NAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
( corrupt(rcvpkt) ||
isNAK(rcvpkt) )
udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)
&& notcorrupt(rcvpkt)
&& isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isNAK(rcvpkt) )
rdt_rcv(rcvpkt)
&& isACK(rcvpkt)
Wait for
call 1 from
above
Wait for
ACK or
NAK 1
Λ
Λ

Wait for
0 from
below
sndpkt = make_pkt(NAK, chksum)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
not corrupt(rcvpkt) &&
has_seq0(rcvpkt)
rdt_rcv(rcvpkt) && notcorrupt(rcvpkt)
&& has_seq1(rcvpkt)
deliver_data(data)
sndpkt = make_pkt(ACK, chksum)
udt_send(sndpkt)
Wait for
1 from
below
&& has_seq0(rcvpkt)
deliver_data(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) && (corrupt(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
not corrupt(rcvpkt) &&
has_seq1(rcvpkt)
rdt_rcv(rcvpkt) && (corrupt(rcvpkt)
udt_send(sndpkt)
sndpkt = make_pkt(NAK, chksum)
udt_send(sndpkt)
rdt2.1: receiver, handles garbled ACK/NAKs
ACKing pkt 0
ACKing pkt 1

rdt2.1: discussion
sender:
 seq # added to pkt
 two seq. #’s (0,1) will
suffice. Why?
 must check if received
ACK/NAK corrupted
 twice as many states
 state must “remember”
whether “expected”
pkt should have seq #
of 0 or 1
receiver:
 must check if received
packet is duplicate
 state indicates whether
0 or 1 is expected pkt
seq #
 note: receiver can not
know if its last
ACK/NAK received
OK at sender

rdt2.2: a NAK-free protocol
 same functionality as rdt2.1, using ACKs only
 instead of NAK, receiver sends ACK for last pkt
received OK
 receiver must explicitly include seq # of pkt being ACKed
 duplicate ACK at sender results in same action as
NAK: retransmit current pkt

rdt2.2: sender, receiver fragments
Wait for
call 0 from
above
udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) &&
isACK(rcvpkt,1) )
rdt_rcv(rcvpkt)
&& isACK(rcvpkt,0)
Wait for
ACK
0
sender FSM
fragment
&& has_seq1(rcvpkt)
deliver_data(data)
sndpkt = make_pkt(ACK1, chksum)
udt_send(sndpkt)
Wait for
0 from
below
rdt_rcv(rcvpkt) &&
(corrupt(rcvpkt) ||
has_seq1(rcvpkt))
udt_send(sndpkt)
receiver FSM
fragment
Λ
Acking pkt 1

rdt3.0: channels with errors and loss
new assumption:
underlying channel can
also lose packets
(data, ACKs)
 checksum, seq. #,
ACKs, retransmissions
will be of help … but
not enough
approach: sender waits
“reasonable” amount of
time for ACK
 retransmits if no ACK
received in this time
 if pkt (or ACK) just delayed
(not lost):
 retransmission will be
duplicate, but seq. #’s
already handles this
 receiver must specify seq
# of pkt being ACKed
 requires countdown timer

rdt3.0 sender
udt_send(sndpkt)
start_timer
rdt_send(data)
Wait
for
ACK0
rdt_rcv(rcvpkt) &&
isACK(rcvpkt,1) )
Wait for
call 1 from
above
udt_send(sndpkt)
start_timer
rdt_send(data)
rdt_rcv(rcvpkt)
&& isACK(rcvpkt,0)
rdt_rcv(rcvpkt) &&
isACK(rcvpkt,0) )
rdt_rcv(rcvpkt)
&& isACK(rcvpkt,1)
stop_timer
stop_timer
udt_send(sndpkt)
start_timer
timeout
udt_send(sndpkt)
start_timer
timeout
rdt_rcv(rcvpkt)
Wait for
call 0from
above
Wait
for
ACK1
Λ
rdt_rcv(rcvpkt)
Λ
Λ
Λ

sender receiver
rcv pkt1
rcv pkt0
send ack0
send ack1
send ack0
rcv ack0
send pkt0
send pkt1
rcv ack1
send pkt0
rcv pkt0
pkt0
pkt0
pkt1
ack1
ack0
ack0
(a) no loss
sender receiver
rcv pkt1
rcv pkt0
send ack0
send ack1
send ack0
rcv ack0
send pkt0
send pkt1
rcv ack1
send pkt0
rcv pkt0
pkt0
pkt0
ack1
ack0
ack0
(b) packet loss
pkt1
X
loss
pkt1
timeout
resend pkt1
rdt3.0 in action

rdt3.0 in action
rcv pkt1
send ack1
(detect duplicate)
pkt1
sender receiver
rcv pkt1
rcv pkt0
send ack0
send ack1
send ack0
rcv ack0
send pkt0
send pkt1
rcv ack1
send pkt0
rcv pkt0
pkt0
pkt0
ack1
ack0
ack0
(c) ACK loss
ack1
X
loss
pkt1
timeout
resend pkt1
rcv pkt1
send ack1
(detect duplicate)
pkt1
sender receiver
rcv pkt1
send ack0
rcv ack0
send pkt1
send pkt0
rcv pkt0
pkt0
ack0
(d) premature timeout/ delayed ACK
pkt1
timeout
resend pkt1
ack1
send ack1
send pkt0
rcv ack1
pkt0
ack1
ack0
send pkt0
rcv ack1 pkt0
rcv pkt0
send ack0ack0
rcv pkt0
send ack0
(detect duplicate)

Performance of rdt3.0
 rdt3.0 is correct, but performance stinks
 e.g.: 1 Gbps link, 15 ms prop. delay, 8000 bit packet:
 U sender: utilization – fraction of time sender busy sending
U
sender =
.008
30.008
= 0.00027
L / R
RTT + L / R
=
 if RTT=30 msec, 1KB pkt every 30 msec: 33kB/sec thruput
over 1 Gbps link
 network protocol limits use of physical resources!
Dtrans =
L
R
8000 bits
109
bits/sec
= = 8 microsecs

rdt3.0: stop-and-wait operation
first packet bit transmitted, t = 0
sender receiver
RTT
last packet bit transmitted, t = L / R
first packet bit arrives
last packet bit arrives, send
ACK
ACK arrives, send next
packet, t = RTT + L / R
U
sender =
.008
30.008
= 0.00027
L / R
RTT + L / R
=

Pipelined protocols
pipelining: sender allows multiple, “in-flight”, yet-
to-be-acknowledged pkts
 range of sequence numbers must be increased
 buffering at sender and/or receiver
 two generic forms of pipelined protocols: go-Back-N,
selective repeat

Pipelining: increased utilization
first packet bit transmitted, t = 0
sender receiver
RTT
last bit transmitted, t = L / R
first packet bit arrives
last packet bit arrives, send ACK
ACK arrives, send next
packet, t = RTT + L / R
last bit of 2nd
packet arrives, send ACK
last bit of 3rd
packet arrives, send ACK
3-packet pipelining increases
utilization by a factor of 3!
U
sender =
.0024
30.008
= 0.00081
3L / R
RTT + L / R
=

Pipelined protocols: overview
Go-back-N:
 sender can have up to
N unacked packets in
pipeline
 receiver only sends
cumulative ack
 doesn’t ack packet if
there’s a gap
 sender has timer for
oldest unacked packet
 when timer expires,
retransmit all unacked
packets
Selective Repeat:
 sender can have up to N
unack’ed packets in
pipeline
 rcvr sends individual ack
for each packet
 sender maintains timer
for each unacked packet
 when timer expires,
retransmit only that
unacked packet

Go-Back-N: sender
 k-bit seq # in pkt header
 “window” of up to N, consecutive unack’ed pkts allowed
 ACK(n): ACKs all pkts up to, including seq # n - “cumulative
ACK”
 may receive duplicate ACKs (see receiver)
 timer for oldest in-flight pkt
 timeout(n): retransmit packet n and all higher seq # pkts in
window

GBN: sender extended FSM
Wait
start_timer
udt_send(sndpkt[base])
udt_send(sndpkt[base+1])
…
udt_send(sndpkt[nextseqnum-
1])
timeout
rdt_send(data)
if (nextseqnum < base+N) {
sndpkt[nextseqnum] = make_pkt(nextseqnum,data,chksum)
udt_send(sndpkt[nextseqnum])
if (base == nextseqnum)
start_timer
nextseqnum++
}
else
refuse_data(data)
base = getacknum(rcvpkt)+1
If (base == nextseqnum)
stop_timer
else
start_timer
rdt_rcv(rcvpkt) &&
notcorrupt(rcvpkt)
base=1
nextseqnum=1
rdt_rcv(rcvpkt)
&& corrupt(rcvpkt)
Λ

ACK-only: always send ACK for correctly-received
pkt with highest in-order seq #
 may generate duplicate ACKs
 need only remember expectedseqnum
 out-of-order pkt:
 discard (don’t buffer): no receiver buffering!
 re-ACK pkt with highest in-order seq #
Wait
udt_send(sndpkt)
default
rdt_rcv(rcvpkt)
&& notcurrupt(rcvpkt)
&& hasseqnum(rcvpkt,expectedseqnum)
deliver_data(data)
sndpkt = make_pkt(expectedseqnum,ACK,chksum)
udt_send(sndpkt)
expectedseqnum++
expectedseqnum=1
sndpkt =
make_pkt(expectedseqnum,ACK,chksum)
Λ
GBN: receiver extended FSM

GBN in action
send pkt0
send pkt1
send pkt2
send pkt3
(wait)
sender receiver
receive pkt0, send ack0
receive pkt3, discard,
(re)send ack1rcv ack0, send pkt4
rcv ack1, send pkt5
pkt 2 timeout
send pkt2
send pkt3
send pkt4
send pkt5
Xloss
(re)send ack1
(re)send ack1
rcv pkt2, deliver, send ack2
ignore duplicate ACK
0 1 2 3 4 5 6 7 8
sender window (N=4)
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8

Selective repeat
 receiver individually acknowledges all correctly
received pkts
 buffers pkts, as needed, for eventual in-order delivery
to upper layer
 sender only resends pkts for which ACK not
received
 sender timer for each unACKed pkt
 sender window
 N consecutive seq #’s
 limits seq #s of sent, unACKed pkts

Selective repeat: sender, receiver windows

Selective repeat
data from above:
 if next available seq # in
window, send pkt and
start timer
timeout(n):
 resend pkt n, restart
timer
ACK(n) in [sendbase,sendbase+N-
1]:
 mark pkt n as received
 if n smallest unACKed
pkt, advance window base
to next unACKed seq #
sender
pkt n in [rcvbase, rcvbase+N-1]
 send ACK(n)
 out-of-order: buffer
 in-order: deliver (also
deliver buffered, in-order
pkts), advance window to
next not-yet-received pkt
pkt n in [rcvbase-N,rcvbase-1]
 ACK(n)
otherwise:
 ignore
receiver

Selective repeat in action
send pkt0
send pkt1
send pkt2
send pkt3
(wait)
sender receiver
receive pkt3, buffer,
send ack3rcv ack0, send pkt4
rcv ack1, send pkt5
pkt 2 timeout
send pkt2
Xloss
send ack4
send ack5
rcv pkt2; deliver pkt2,
pkt3, pkt4, pkt5; send ack2
record ack3 arrived
0 1 2 3 4 5 6 7 8
sender window (N=4)
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
0 1 2 3 4 5 6 7 8
record ack4 arrived
record ack4 arrived
Q: what happens when ack2 arrives?

Selective repeat:
dilemma
example:
 seq #’s: 0, 1, 2, 3
 window size=3
receiver window
(after receipt)
sender window
(after receipt)
0 1 2 3 0 1 2
0 1 2 3 0 1 2
0 1 2 3 0 1 2
pkt0
pkt1
pkt2
0 1 2 3 0 1 2 pkt0
timeout
retransmit pkt0
0 1 2 3 0 1 2
0 1 2 3 0 1 2
0 1 2 3 0 1 2X
X
X
will accept packet
with seq number 0
(b) oops!
0 1 2 3 0 1 2
0 1 2 3 0 1 2
0 1 2 3 0 1 2
pkt0
pkt1
pkt2
0 1 2 3 0 1 2
pkt0
0 1 2 3 0 1 2
0 1 2 3 0 1 2
0 1 2 3 0 1 2
X
will accept packet
with seq number 0
0 1 2 3 0 1 2 pkt3
(a) no problem
receiver can’t see sender side.
receiver behavior identical in both cases!
something’s (very) wrong!
 receiver sees no
difference in two
scenarios!
 duplicate data
accepted as new in
(b)
Q: what relationship
between seq # size
and window size to
avoid problem in (b)?

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

TCP: Overview RFCs: 793,1122,1323, 2018, 2581
 full duplex data:
 bi-directional data flow
in same connection
 MSS: maximum segment
size
 connection-oriented:
 handshaking (exchange
of control msgs) inits
sender, receiver state
before data exchange
 flow controlled:
 sender will not
overwhelm receiver
 point-to-point:
 one sender, one receiver
 reliable, in-order byte
steam:
 no “message
boundaries”
 pipelined:
 TCP congestion and
flow control set window
size

TCP segment structure
32 bits
application
data
(variable length)
sequence number
acknowledgement number
receive window
Urg data pointerchecksum
FSRPAU
head
len
not
used
options (variable length)
URG: urgent data
(generally not used)
ACK: ACK #
valid
PSH: push data now
(generally not used)
RST, SYN, FIN:
connection estab
(setup, teardown
commands)
# bytes
rcvr willing
to accept
counting
by bytes
of data
(not segments!)
Internet
checksum
(as in UDP)

TCP seq. numbers, ACKs
sequence numbers:
byte stream “number” of
first byte in segment’s
data
acknowledgements:
seq # of next byte
expected from other side
cumulative ACK
Q: how receiver handles
out-of-order segments
A: TCP spec doesn’t say,
- up to implementor source port # dest port #
sequence number
checksum
rwnd
urg pointer
incoming segment to sender
A
sent
ACKed
sent, not-
yet ACKed
(“in-flight”)
usable
but not
yet sent
not
usable
window size
N
sender sequence number space
sequence number
checksum
rwnd
urg pointer
outgoing segment from sender

TCP seq. numbers, ACKs
User
types
‘C’
host ACKs
receipt
of echoed
‘C’
host ACKs
receipt of
‘C’, echoes
back ‘C’
simple telnet scenario
Host BHost A
Seq=42, ACK=79, data = ‘C’
Seq=79, ACK=43, data = ‘C’
Seq=43, ACK=80

TCP round trip time, timeout
Q: how to set TCP
timeout value?
 longer than RTT
 but RTT varies
 too short: premature
timeout, unnecessary
retransmissions
 too long: slow reaction
to segment loss
Q: how to estimate RTT?
 SampleRTT: measured
time from segment
transmission until ACK
receipt
 ignore retransmissions
 SampleRTT will vary, want
estimated RTT “smoother”
 average several recent
measurements, not just
current SampleRTT

RTT: gaia.cs.umass.edu to fantasia.eurecom.fr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RTT(milliseconds)
SampleRTT Estimated RTT
EstimatedRTT = (1- α)*EstimatedRTT + α*SampleRTT
 exponential weighted moving average
 influence of past sample decreases exponentially fast
 typical value: α = 0.125
RTT(milliseconds
RTT: gaia.cs.umass.edu to fantasia.eurecom.fr
sampleRTT
EstimatedRTT
time (seconds)

More on the Averaging
 Input: x[k], k = 1, 2, …
 Output: y[k], y[0]=0

 timeout interval: EstimatedRTT plus “safety margin”
 large variation in EstimatedRTT -> larger safety margin
 estimate SampleRTT deviation from EstimatedRTT:
DevRTT = (1-β)*DevRTT +
β*|SampleRTT-EstimatedRTT|
(typically, β =
0.25)
TimeoutInterval = EstimatedRTT + 4*DevRTT
estimated RTT “safety margin”

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

TCP reliable data transfer
 TCP creates rdt service
on top of IP’s unreliable
service
 pipelined segments
 cumulative acks
 single retransmission
timer
 retransmissions
triggered by:
 timeout events
 duplicate acks
let’s initially consider
simplified TCP sender:
 ignore duplicate acks
 ignore flow control,
congestion control

TCP sender events:
data rcvd from app:
 create segment with
seq #
 seq # is byte-stream
number of first data
byte in segment
 start timer if not
already running
 think of timer as for
oldest unacked
segment
 expiration interval:
TimeOutInterval
timeout:
 retransmit segment
that caused timeout
 restart timer
ack rcvd:
 if ack acknowledges
previously unacked
segments
 update what is known
to be ACKed
 start timer if there are
still unacked segments

TCP sender (simplified)
wait
for
event
NextSeqNum = InitialSeqNum
SendBase = InitialSeqNum
Λ
create segment, seq. #: NextSeqNum
pass segment to IP (i.e., “send”)
NextSeqNum = NextSeqNum + length(data)
if (timer currently not running)
start timer
data received from application above
retransmit not-yet-acked segment
with smallest seq. #
start timer
timeout
if (y > SendBase) {
SendBase = y
/* SendBase–1: last cumulatively ACKed byte */
if (there are currently not-yet-acked segments)
start timer
else stop timer
}
ACK received, with ACK field value y

TCP: retransmission scenarios
lost ACK scenario
Host BHost A
Seq=92, 8 bytes of data
ACK=100
X
timeout
ACK=100
premature timeout
Host BHost A
ACK=100
Seq=92, 8
bytes of data
timeout
ACK=120
ACK=120
SendBase=100
SendBase=120
SendBase=120
SendBase=92

TCP: retransmission scenarios
X
cumulative ACK
Host BHost A
ACK=100
timeout
ACK=120

TCP ACK generation [RFC 1122, RFC 2581]
event at receiver
arrival of in-order segment with
expected seq #. All data up to
expected seq # already ACKed
arrival of in-order segment with
expected seq #. One other
segment has ACK pending
arrival of out-of-order segment
higher-than-expect seq. # .
Gap detected
arrival of segment that
partially or completely fills gap
TCP receiver action
delayed ACK. Wait up to 500ms
for next segment. If no next segment,
send ACK
immediately send single cumulative
ACK, ACKing both in-order segments
immediately send duplicate ACK,
indicating seq. # of next expected byte
immediate send ACK, provided that
segment starts at lower end of gap

TCP fast retransmit
 time-out period often
relatively long:
 long delay before
resending lost packet
 detect lost segments
via duplicate ACKs.
 sender often sends
many segments back-
to-back
 if segment is lost, there
will likely be many
duplicate ACKs.
if sender receives 3
additional ACKs for
same data
(“triple duplicate ACKs”),
resend unacked
segment with smallest
seq #
 likely that unacked
segment lost, so don’t
wait for timeout
TCP fast retransmit
(“triple duplicate ACKs”),

X
fast retransmit after sender
receipt of triple duplicate ACK
Host BHost A
ACK=100
timeout
ACK=100
ACK=100
ACK=100
TCP fast retransmit

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

TCP flow control
application
process
TCP socket
receiver buffers
TCP
code
IP
code
application
OS
receiver protocol stack
application may
remove data from
TCP socket buffers ….
… slower than TCP
receiver is delivering
(sender is sending)
from sender
receiver controls sender, so
sender won’t overflow
receiver’s buffer by transmitting
too much, too fast
flow control

TCP flow control
buffered data
free buffer spacerwnd
RcvBuffer
TCP segment payloads
to application process
 receiver “advertises” free
buffer space by including
rwnd value in TCP header
of receiver-to-sender
segments
 RcvBuffer size set via
socket options (typical default
is 4096 bytes)
 many operating systems
autoadjust RcvBuffer
 sender limits amount of
unacked (“in-flight”) data to
receiver’s rwnd value
 guarantees receive buffer
will not overflow
receiver-side buffering

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

Connection Management
before exchanging data, sender/receiver “handshake”:
 agree to establish connection (each knowing the other willing
to establish connection)
 agree on connection parameters
connection state: ESTAB
connection variables:
seq # client-to-server
server-to-client
rcvBuffer size
at server,client
application
network
connection state: ESTAB
connection Variables:
seq # client-to-server
server-to-client
rcvBuffer size
at server,client
application
network
Socket clientSocket =
newSocket("hostname","port
number");
Socket connectionSocket =
welcomeSocket.accept();

Q: will 2-way handshake
always work in
network?
 variable delays
 retransmitted messages
(e.g. req_conn(x)) due to
message loss
 message reordering
 can’t “see” other side
2-way handshake:
Let’s talk
OK
ESTAB
ESTAB
choose x
req_conn(x)
ESTAB
ESTAB
acc_conn(x)
Agreeing to establish a connection

Agreeing to establish a connection
2-way handshake failure scenarios:
retransmit
req_conn(x)
ESTAB
req_conn(x)
half open connection!
(no client!)
client
terminates
server
forgets x
connection
x completes
retransmit
req_conn(x)
ESTAB
req_conn(x)
data(x+1)
retransmit
data(x+1)
accept
data(x+1)
choose x
req_conn(x)
ESTAB
ESTAB
acc_conn(x)
client
terminates
ESTAB
choose x
req_conn(x)
ESTAB
acc_conn(x)
data(x+1) accept
data(x+1)
connection
x completes server
forgets x

TCP 3-way handshake
SYNbit=1, Seq=x
choose init seq num, x
send TCP SYN msg
ESTAB
SYNbit=1, Seq=y
ACKbit=1; ACKnum=x+1
choose init seq num, y
send TCP SYNACK
msg, acking SYN
ACKbit=1, ACKnum=y+1
received SYNACK(x)
indicates server is live;
send ACK for SYNACK;
this segment may contain
client-to-server data
received ACK(y)
indicates client is live
SYNSENT
ESTAB
SYN RCVD
client state
CLOSED
server state
LISTEN

TCP 3-way handshake: FSM
closed
Λ
listen
SYN
rcvd
SYN
sent
ESTAB
Socket clientSocket =
newSocket("hostname","port
number");
SYN(seq=x)
Socket connectionSocket =
welcomeSocket.accept();
SYN(x)
SYNACK(seq=y,ACKnum=x+1)
create new socket for
communication back to client
SYNACK(seq=y,ACKnum=x+1)
ACK(ACKnum=y+1)
ACK(ACKnum=y+1)
Λ

TCP: closing a connection
 client, server each close their side of connection
 send TCP segment with FIN bit = 1
 respond to received FIN with ACK
 on receiving FIN, ACK can be combined with own FIN
 simultaneous FIN exchanges can be handled

FIN_WAIT_2
CLOSE_WAIT
FINbit=1, seq=y
ACKbit=1; ACKnum=y+1
ACKbit=1; ACKnum=x+1
wait for server
close
can still
send data
can no longer
send data
LAST_ACK
CLOSED
TIME_WAIT
timed wait
for 2*max
segment lifetime
CLOSED
TCP: closing a connection
FIN_WAIT_1 FINbit=1, seq=xcan no longer
send but can
receive data
clientSocket.close()
client state server state
ESTABESTAB

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

congestion:
 informally: “too many sources sending too much
data too fast for network to handle”
 different from flow control!
 manifestations:
 lost packets (buffer overflow at routers)
 long delays (queueing in router buffers)
 a top-10 problem!
Principles of congestion control

Causes/costs of congestion: scenario 1
 two senders, two
receivers
 one router, infinite
buffers
 output link capacity: R
 no retransmission
 maximum per-connection
throughput: R/2
unlimited shared
output link buffers
Host A
original data: λin
Host B
throughput: λout
R/2
R/2
λout
λin R/2delay λin
 large delays as arrival rate,
λin, approaches capacity

 one router, finite buffers
 sender retransmission of timed-out packet
 application-layer input = application-layer output: λin = λout
 transport-layer input includes retransmissions : λin λin
finite shared output
link buffers
Host A
λin : original data
Host B
λoutλ'in: original data, plus
retransmitted data
‘

idealization: perfect
knowledge
 sender sends only when
router buffers available
link buffers
retransmitted data
copy
free buffer space!
R/2
R/2
λout
λin
Host B
A

retransmitted data
copy
no buffer space!
Idealization: known loss
packets can be lost,
dropped at router due
to full buffers
 sender only resends if
packet known to be lost
A
Host B

retransmitted data
free buffer space!
Idealization: known loss
packets can be lost,
dropped at router due
to full buffers
 sender only resends if
packet known to be lost
R/2
R/2λin
λout
when sending at R/2,
some packets are
retransmissions but
asymptotic goodput
is still R/2 (why?)
A
Host B

A
λin
λoutλ'in
copy
free buffer space!
timeout
R/2
R/2λin
λout
some packets are
retransmissions
including duplicated
that are delivered!
Host B
Realistic: duplicates
 packets can be lost, dropped
at router due to full buffers
 sender times out prematurely,
sending two copies, both of
which are delivered

R/2
λout
some packets are
retransmissions
including duplicated
that are delivered!
“costs” of congestion:
 more work (retrans) for given “goodput”
 unneeded retransmissions: link carries multiple copies of pkt
 decreasing goodput
R/2λin
Realistic: duplicates
 packets can be lost, dropped
at router due to full buffers
 sender times out prematurely,
sending two copies, both of
which are delivered

 four senders
 multihop paths
 timeout/retransmit
Q: what happens as λin and λin
’
increase ?
link buffers
Host A λout
Host B
Host C
Host D
λ'in: original data, plus
retransmitted data
A: as red λin
’
increases, all arriving
blue pkts at upper queue are
dropped, blue throughput  0

another “cost” of congestion:
 when packet dropped, any “upstream
transmission capacity used for that packet was
wasted!
C/2
C/2
λout
λin
’

Approaches towards congestion control
two broad approaches towards congestion control:
end-end congestion
control:
 no explicit feedback
from network
 congestion inferred
from end-system
observed loss, delay
 approach taken by
TCP
network-assisted
congestion control:
 routers provide
feedback to end systems
single bit indicating
congestion (SNA,
DECbit, TCP/IP ECN,
ATM)
explicit rate for
sender to send at

Case study: ATM ABR congestion control
ABR: available bit rate:
 “elastic service”
 if sender’s path
“underloaded”:
 sender should use
available bandwidth
 if sender’s path
congested:
 sender throttled to
minimum guaranteed
rate
RM (resource management)
cells:
 sent by sender, interspersed
with data cells
 bits in RM cell set by
switches (“network-assisted”)
 NI bit: no increase in rate
(mild congestion)
 CI bit: congestion
indication
 RM cells returned to sender
by receiver, with bits intact

Case study: ATM ABR congestion control
 two-byte ER (explicit rate) field in RM cell
 congested switch may lower ER value in cell
 sender transmits at max supportable rate on path
 EFCI bit in data cells: set to 1 in congested switch
 if data cell preceding RM cell has EFCI set, receiver sets
CI bit in returned RM cell
RM cell data cell

Chapter 3 outline
3.1 transport-layer
services
demultiplexing
3.3 connectionless
transport: UDP
data transfer
transport: TCP
 flow control
control

TCP congestion control: additive increase
multiplicative decrease
 approach: sender increases transmission rate (window
size), probing for usable bandwidth, until loss occurs
 additive increase: increase cwnd by 1 MSS every
RTT until loss detected
 multiplicative decrease: cut cwnd in half after loss
cwnd:TCPsender
congestionwindowsize
AIMD saw tooth
behavior: probing
for bandwidth
additively increase window size …
…. until loss occurs (then cut window in half)
time

TCP Congestion Control: details
 sender limits transmission:
 cwnd is dynamic, function
of perceived network
congestion
TCP sending rate:
 roughly: send cwnd
bytes, wait RTT for
ACKS, then send
more bytes
last byte
ACKed sent, not-
yet ACKed
(“in-flight”)
last byte
sent
cwnd
LastByteSent-
LastByteAcked
< cwnd
sender sequence number space
rate ~~
cwnd
RTT
bytes/sec
Actually, max. in-flight data
is limited by min(cwnd,
rcv_wnd)

TCP Slow Start
 when connection begins,
increase rate
exponentially until first
loss event:
 initially cwnd = 1 MSS
 double cwnd every RTT
 done by incrementing
cwnd for every ACK
received
 summary: initial rate is
slow but ramps up
exponentially fast
Host A
one segment
RTT
Host B
time
two segments
four segments

TCP Slow Start
 When connection begins,
CongWin = 1 MSS
 Example: MSS = 500 bytes &
RTT = 200 msec
 initial rate = 20 kbps
 available bandwidth may be
>> MSS/RTT
 desirable to quickly ramp up
to respectable rate
When connection begins,
increase rate
exponentially fast until
first loss event

TCP: detecting, reacting to loss
 loss indicated by timeout:
 cwnd set to 1 MSS;
 window then grows exponentially (as in slow start)
to threshold, then grows linearly
 loss indicated by 3 duplicate ACKs: TCP RENO
 dup ACKs indicate network capable of delivering
some segments
 cwnd is cut in half window then grows linearly
 TCP Tahoe always sets cwnd to 1 (timeout or 3
duplicate acks)

Q: when should the
exponential
increase switch to
linear?
A: when cwnd gets
to 1/2 of its value
before timeout or
3 duplicated
ACKs.
Implementation:
 variable ssthresh
 on loss event, ssthresh
is set to 1/2 of cwnd just
before loss event
TCP: switching from slow start to CA

Summary: TCP Congestion Control
timeout
ssthresh = cwnd/2
cwnd = 1 MSS
dupACKcount = 0
retransmit missing segment
Λ
cwnd > ssthresh
congestion
avoidance
cwnd = cwnd + MSS (MSS/cwnd)
dupACKcount = 0
transmit new segment(s), as allowed
new ACK
.
dupACKcount++
duplicate ACK
fast
recovery
cwnd = cwnd + MSS
duplicate ACK
ssthresh= cwnd/2
cwnd = ssthresh + 3
dupACKcount == 3
timeout
ssthresh = cwnd/2
cwnd = 1
dupACKcount = 0
ssthresh= cwnd/2
cwnd = ssthresh + 3
dupACKcount == 3cwnd = ssthresh
dupACKcount = 0
New ACK
slow
start
timeout
ssthresh = cwnd/2
cwnd = 1 MSS
dupACKcount = 0
cwnd = cwnd+MSS
dupACKcount = 0
new ACKdupACKcount++
duplicate ACK
Λ
cwnd = 1 MSS
ssthresh = 64 KB
dupACKcount = 0
New
ACK!
New
ACK!
New
ACK!

 When CongWin is below Threshold, sender in slow-
start phase, window grows exponentially.
 When CongWin is above Threshold, sender is in
congestion-avoidance phase, window grows linearly.
 When a triple duplicate ACK occurs, Threshold set to
CongWin/2 and CongWin set to Threshold.
 When timeout occurs, Threshold set to CongWin/2
and CongWin is set to 1 MSS.

TCP sender congestion control
Event State TCP Sender Action Commentary
ACK receipt
for previously
unacked
data
Slow Start
(SS)
CongWin = CongWin + MSS,
If (CongWin > Threshold)
set state to “Congestion
Avoidance”
Resulting in a doubling of
CongWin every RTT
ACK receipt
for previously
unacked
data
Congestion
Avoidance
(CA)
CongWin = CongWin+MSS *
(MSS/CongWin)
Additive increase, resulting
in increase of CongWin by
1 MSS every RTT
Loss event
detected by
triple
duplicate
ACK
SS or CA Threshold = CongWin/2,
CongWin = Threshold,
Set state to “Congestion
Avoidance”
Fast recovery,
implementing multiplicative
decrease. CongWin will not
drop below 1 MSS.
Timeout SS or CA Threshold = CongWin/2,
CongWin = 1 MSS,
Set state to “Slow Start”
Enter slow start
Duplicate
ACK
SS or CA Increment duplicate ACK count
for segment being acked
CongWin and Threshold
not changed

 end-end control (no network
assistance)
 sender limits transmission:
LastByteSent-LastByteAcked
≤ CongWin
 Roughly,
 CongWin is dynamic, function of
perceived network congestion
How does sender perceive
congestion?
 loss event = timeout or 3
duplicate acks
 TCP sender reduces rate
(CongWin) after loss
event
three mechanisms:
 AIMD
 slow start
 conservative after timeout
events
rate =
CongWin
RTT
Bytes/sec

TCP throughput: One Estimate
 avg. TCP thruput as function of window size, RTT?
 ignore slow start, assume always data to send
 W: window size (measured in bytes) where loss occurs
 avg. window size (# in-flight bytes) is ¾ W
 avg. thruput is 3/4W per RTT
W
W/2
avg TCP thruput =
3
4
W
RTT
bytes/sec

TCP Futures: TCP over “long, fat pipes”
 example: 1500 byte segments, 100ms RTT, want
10 Gbps throughput
 requires W = 83,333 in-flight segments
 If a TCP connection gets hit by segment losses,
throughput in terms of segment loss probability, L
[Mathis 1997]:
➜ to achieve 10 Gbps throughput, need a loss rate of L
= 2·10-10
– a very small loss rate!
 new versions of TCP for high-speed
TCP throughput = 1.22 . MSS
RTT L

fairness goal: if K TCP sessions share same
bottleneck link of bandwidth R, each should have
average rate of R/K
TCP connection 1
bottleneck
router
capacity R
TCP Fairness
TCP connection 2

Why is TCP fair?
two competing sessions:
 additive increase gives slope of 1, as throughout increases
 multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Connection2throughput
congestion avoidance: additive increase
loss: decrease window by factor of 2
congestion avoidance: additive increase
loss: decrease window by factor of 2

Fairness (more)
Fairness and UDP
 multimedia apps often
do not use TCP
 do not want rate
throttled by congestion
control
 instead use UDP:
 send audio/video at
constant rate, tolerate
packet loss
Fairness, parallel TCP
connections
 application can open
multiple parallel
connections between two
hosts
 web browsers do this
 e.g., link of rate R with 9
existing connections:
 new app asks for 1 TCP, gets rate
R/10
 new app asks for 9 TCPs, gets R/2

Chapter 3: summary
 principles behind
transport layer services:
 multiplexing,
demultiplexing
 flow control
 instantiation,
implementation in the
Internet
 UDP
 TCP
next:
 leaving the
network “edge”
(application,
transport layers)
 into the network
“core”

Chapter 3 v6.0

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to Chapter 3 v6.0 (20)

Recently uploaded (20)

Chapter 3 v6.0

Editor's Notes