BACKBONE NETWORKS

Report by Shriram Suryanarayanan
COEN 335 High Performance Networking 1 | P a g e
HIGH SPEED BACKBONE DESIGN AND
ROUTING
A Report by Shriram Suryanarayanan
Santa Clara University
Winter 2015
This report is submitted to Dr. Keyvan Moataghed of Santa Clara University for the fulfilment of requirements for
the High Performance Networking course (COEN 335)

AUDIENCE
This report on High Speed Backbone Design and Routing is for the Designers, Developers, and Product Engineers
of High Speed Networks technology. This report gives a brief overview on the frameworks and networks elements
involved in designing high speed backbone networks.
This report assumes that the reader has basic knowledge of computer networks and IP model.
This report gives a brief introduction on the elements of a backbone and covers a separate chapter on each of those.
It also talks about different type of router designs and need for higher bandwidth for time sensitive applications. This
report also talks about the quality of services and the design methodologies to achieve higher throughput alongside
the bandwidth.

INDEX
Contents
1. Introduction…………………………………………………………………………………………………………….5
2. Fiber Optics……………………………………………………………………………………………………………..7
3. Layer 2 and Layer 3 switches……………………………………………………………………………………9
4. Layer 3: Routers……………………………………………………………………………………………………..12
5. DWDM- Dense wavelength division multiplexing……………………………………………………15
6. Quality of service……………………………………………………………………………………………………19
7. Reselience………………………………………………………………………………………………………………23
8. Future aspects of backbone networks…………………………………………………………………….25
Abbreviations…………………………………………………………………………………………………………26
References……………………………………………………………………………………………………………..28

FIGURE LIST
Figure1: Multiport switches of passive optical networks
Figure 2: Layer 2 switches
Figure 3: Combined Layer 2 and 3 switches
Figure 4: Bus based architecture with multiple forward engines
Figure 5: Switch based architecture with multiple processors
Figure 6: DWDM diagram
Figure 7: Point to point DWDM networks
Figure 8: Wavelength routing with electronic TDM
Figure 9: TOS field
Figure 10: DS field

1. INTRODUCTION
In computer networking, a backbone is a central channel designed to transfer network data at high speeds. To
facilitate long distance reliable communication backbone networks have been designed. In earlier days in data
networking only star topology was used and it was quite simple to design and operate as the data traffic was very
less. With the core of the network moving to high speed client server model and peer to peer communication, the
network had to be extended to a wider spectrum which included providing connectivity to regional distribution
networks and wide area networks. The major two requirements while designing a backbone network are: reliability
and scalability. Any high speed backbone network involving wide area network would include great deal of
hardware equipment and environmental stability, so the design of the backbone networks should be carried out with
a great amount of detail. Since the networks involves large amount of real time traffic, the need for resilience and
stability is also always high. This document would discuss in detail the network elements involved in high speed
backbone networks.
Backbones networks consists of routers and switches connected by fiber optic or Ethernet cables. Computers in a
local network do not connect to a backbone directly; instead connect through internet provided by internet service
providers or through any large organization.
The basic framework behind designing high speed backbone networks are:
(1) Fibre optic cables or Ethernet cables connecting the computers to the network: These cables are designed
for long distance and high bandwidth communication networks. It is a network cable containing strands of
glass fibres inside an insulated casing.
(2) Layer 2 (switches): is a physical device that connects multiple computers in a single network. Ethernet
switches are the commonly switches typically in a home network before home routers became quite
popular. It operates at the data link layer of the OSI model.
(3) Routers: are physical devices that connect multiple networks together. It is typically a layer 3 gateway
device that operates at the network layer of the OSI model. By maintaining configuration information in a
piece of storage called routing table, routers can route incoming and outgoing packets to their correct
destinations.
(4) Layer 3(switches): Typically switches operate the layer 2 of the OSI model, but to improve the
performance of the routers inside the LAN, layer 3 switches were introduced. The hardware inside the layer
3 switches merges some of the functionalities of a layer 2 switch and a router. The software logic employed
behind a router is replaced with some hardware to increase the performance of the layer 3 switches. To cut
down the cost with the physical devices and increase the performance, layer 3 switches were introduced.

(5) DWDM fibers- dense wavelength division multiplexing
Dense wavelength division multiplexing combine’s different sources together in a fibre optic cable and
each set of sources have their own operating wavelength to be transmitted to the receiver across the fiber.
The protocols involved behind designing the network are:
(1) MPLS-multiprotocol label switching
To build next generation intelligent networks that provide a wide range of business as well as enterprise
services and can seamlessly integrate with any of the existing networks like ATM, frame relay, IP. It
combines the aspects of routing with the performance of switching and delivers high scalable end to end
services with good quality of service and paving way to meet the demands of an ever growing IP network.
The configuration of MPLS over the existing network is quite simple and helps in internet providers and
subscribers.
(2) RIP- routing information protocol
It is a distance vector algorithm typically used in small networks. It tells the switch to learn about other
routes in the network. RIP calculates the number of hops between source and destination. It calculates the
route with the lowest metric and assigns that path for the packet flow. RIP can be used in networks to
calculate routes that are 15 hops are less that. It is less efficient than the counterpart OSPF when
implementing in wide area networks.
(3) OSPF-open shortest path first
It is a routing protocol designed by the Internet Engineering Task Force (IETF). When the routing table has
been changed to some reasons, the part of the table that was changed is multicasted to the all other hosts in
the network so that they can update their routing table correspondingly. With RIP, the routing table is sent
every specific time interval irrespective of whether a change has occurred or not. OSPF also gives the user
the permission to assign cost metrics for routers so that certain paths have preference.

2. FIBER OPTICS
Layer 1: Fibre optics
With the advent of new technologies and high speed gigabit Ethernet, the networking requirements have reached a
level where energy efficient cables have to be designed. The speed has increased to almost 40 to 100 Gigabits per
second that new optical technologies need to be invented.
The basic principle behind fibre optic communication:
Travelling of information is different when it comes to cellphone and fibre optic communication. The cellphones
send and receive radio waves to and from the signalling stations to connect to the other party. Whereas in fibre optic
communication which is used in high speed networks deploy thick strands of glass embedded with in a fibre. The
technology behind the whole process of communication is optical technology. A fibre optic cable consists of
hundreds of thin strands of cables; each of thickness compared to a human hair and can carry about millions of
telephone calls.
The basic advantages behind deploying fibre optic cables are:
(1) Bandwidth is a limited which needs to be used quite efficiently. With the advent voice over IP and other
high end technologies the need of higher bandwidth became increasingly necessary. Fibre optic cables
provide high bandwidth for all types of media (voice, data and video).
(2) Fiber optic is light weight compared to copper cables.
(3) Fibre optic has a larger life span compared to copper cables.
(4) Low loss as compared to copper cables. At higher frequencies the signal loss is also higher for copper
cables.
(5) Fibre is more secured when compared to copper and can be safely transmitted over long distances
(6) When optic fibre cables are deployed in industry or an organization setting, the network performance
would increase significantly when compared to deploying it in a home environment.
(7) As optic fibre cables do not electrical energy, it is not affected by any kind of electromagnetic interference.
The main disadvantage of copper cables is that it cannot be carried over long distances.
A recent advancement in fibre optic communication has allowed installing fibre at residents which is called as fibre
to the home (FTTH). The following section discusses the operation of optical LAN’s (used in residents and large
buildings).
Optical LANs
Due to the advancement in broadband connections, the number of users is increasing tremendously and there the
bandwidth to accommodate more users is a constraint. In the past few years, there is a technology shift in using
passive optical networks in homes and small work stations. As optical fibers are accommodating more users, the

cost is also effectively balanced when the whole installation cost is taken into account. By large, this idea of using
passive optical networks is extended to geographical locations like a large organization involving hundreds of users.
The technology used behind deploying optical LAN’s to homes and large corporate buildings are called as
fibre to the home (FTTH) and fibre to the office (FTTO). The following diagram depicts the installation of
a multiport user outlet switch.
It shows typical multiport switches of passive optical networks which has the ability to carry voice, video and data
from the user end. The user terminals have power over Ethernet for the gigabit switch.
FTTH is completely future proof and the capacity of the network is virtually unlimited. The last mile of the network
that reaches all our households would be replaced by FTTH in the years to come enabling more high speed data.
Europe and the US have already moved to this technology which enables high gigabit speed of up to 100 Gbps.
Optical fiber principle:
The optical fiber consists of a light carrying core which is surrounded by an insulating layer called cladding which
concentrates the beam of the light inside the core layer. The core and the cladding are transparent mediums. The
light inside the core is kept by total internal refection which causes the optical fiber to act as a wave guide.
Depending on the number of transverse modes the optical fiber may be classified into:
(1) Single mode optical fiber
(2) Multimode optical fiber
Single mode fibers are used in long distance communication with a link capacity of up to 1000m. The
multimode fibers are used in short distance communication which requires high transmission power.

3. Layer 2 and Layer 3 switches
Layer 2: Switches
The role of the switches in layer 2 is to switch the frames to the correct output port based on the destination MAC
address. Devices within the same subnet do not need any routing to reach their local peers. The address resolution
protocol takes care of finding the corresponding MAC address within the local cluster for which a particular device
is trying to reach. It then sends the packet to correct destination output port based on the resolved MAC address.
The layer 2 switches take care of segmentation of local area networks (LAN) based on the MAC address. The major
difference between a multiport and a layer 2 switch is the involvement of a hardware that enables multiple switching
in the network at the same time.
The same can be depicted from the below diagram,
Consider a layer 2 switch that has 4 ports and has 4 computers connected to it. Computer A wants to establish a
connection with computer B and computer wants with computer D. Since the layer 2 switches are hardware enabled,
multiple frames can be switched in parallel. A and B as well as C and D can communicate with each other. This is
highly effective when computers A and B use TCP (Transmission control protocol) as their transport protocol and C
and D use the internet protocol (IP). The layer 2 switches are designed to be protocol independent.
Virtual LANs
Forwarding of frames is efficient when the computers are within the same sub domain. What if communication has
to happen between A and D? Layer 2 switches can achieve it? Yes by configuring layer 2 switches are virtual
LAN’s. The configuration can be achieved by ports or through MAC addresses. The port based VLAN’s assume that
the frames coming in are from the same VLAN and the MAC based VLAN uses MAC addresses to determine the
VLAN membership. In the above diagram, to eliminate the need of a router while forwarding frames to other
domains extra piece of software is embedded in the Layer 2 hardware.
Characteristic of layer 2 switches
Apart from storing and forwarding the frames onto the output ports, the Layer 2 switches also act as an IP end node
for SNMP (simple network management protocol), web based management etc. The switches have a MAC address

that can be considered as an IP end node and routing decisions can be taken. To avoid looping of several switching
paths, the Layer 2 switches also support spanning tree protocol and eliminate the loop by closing down on the
switching interfaces.
There are 3 types of layer 2 switching methods:
1. Store and forward switching
The frames are copied onto the buffers and a cyclic redundant check is performed to determine the errors in
the frames. If there is any deviation found, that particular frame is discarded.
Note
If the length of the Ethernet frame is less than 64 bytes or greater than 1518 bytes, that frame is discarded.
2. Cut through switching
The frame’s destination MAC address is copied onto the buffer and the output port is determined as soon as
the switch starts reading the frame. Due to this the delay in forwarding the frame is eliminated. Unlike the
store and forward switching concept that discards bad frames through CRC, in this method the frames are
forwarded though they are corrupt and CRC is done at the destination end which eliminates latency inside
the switching mechanism. The quality of service is better across the network in case of store and forward
method as it eliminates bad frames by dropping them then and there in the network.
3. Fragment free switching
It is a compromise between the store and forward switching and the cut through switching. This method
was developed to solve the late collision problem. The switch stores the first 64 bytes of the segmented
frame as most of the late collision problem occur in this region.
Layer 3 switches
The layer 2 switches and the bridges operate at the data link layer of the OSI model whereas the layer 3 switches,
like the routers, operate at the network layer. The difference between a Layer 3 switch and a router is that the switch
makes decision using the port level IP addresses whereas the router using the routing table to route packets.
Multilayer switching is a technique used at both the datalink and the network layer. The store and forward technique
is used to forward frames for multilayer switching. The entire frame sequence is received, stored in the buffer and
the CRC redundant check is done to see if the frames are corrupted and then switches to the output port to the
destination.
Layer 3 switching operation
A Layer 3 switch switches packets that are encapsulated inside an Ethernet frame and makes decisions depending
upon the network traffic. The switching device has a built in hardware that makes it faster in switching the packets
compared to a router. When the frame reaches the switch, the frame is forwarded to the exact output port by looking
at the MAC address. The time to live(TTL) and error check sum are done at the hardware level and TTL logic is also
taken care by the switching device and the count is decremented before it is passed on to the next hop. The
calculation of the checksum and the switching are done at very high speeds. The routers communicate with the

Layer 3 switches using routing protocols such as OSPF (open shortest path first) and RIP (routing information
protocol)
Combined Layer 2 and layer 3 switches
The traditional way of a Layer 2 switch is to forward frames to the correct output port without modifying the
destination MAC address. In case of Virtual LAN’s, the router is overloaded and might hang off at times leading to
network congestion. To ease it off, the functionalities of a Layer 2 switch and a Layer 3 switch are combined
together. The below diagram depicts the switching of frames between two different subnets.
The stations A and B are connected to one subnet and stations C and D are connected to different subnet. When
station B and D are to communicate, the frame is set to the incoming port 2 of the combined switch of
Layer2/Layer3, the switch looks on to the routing table and changes the Header of the IP packet to the destination
MAC address of the station D. In this way the routers have been replaced by Layer 3 switches in some of the
scenarios like virtual VLAN’s.
Characteristics
As the Layer 3 switches perform the functionalities of a Layer 2 switch the VLAN configuration has to be
done so that they can know about the MAC addresses on the ports. In Layer 3 switching, the ports of the
switch can be configured to each of the subnets that the stations belong to and this can be done by looking
into the source IP address of the incoming frame.
Are the Layer 3 switches so fast that it can replace a router? The answer is NO. The Layer 3 switches are
more efficient only for switching frame when inside a VLAN or an enterprise building. They get the
routing information with the help of the nearby routers. To connect to the wide area network and perform
congestion control mechanisms when the network is congested, routers are a must. Apart from this, the
routers perform some specific functions like inter autonomous routing using protocols such as border
gateway protocol

4. Layer 3: Routers
Network Layer: Routers
Communication between different computers in the wide area network and interconnection between LAN’s is not
possible with the help of Layer 2 switches. The most established device that takes care of inter-networking between
LAN’s and that connect to the internet is the network layer device routers.
The datagram that needs to be sent over the interconnection of several networks and we do not know the
destination’s physical location. The role of the routers is to forward the datagrams to the next router by looking onto
the routing table. Hand off takes place where the datagram is forwarded to the next router. This process is called as
next hop routing.
High Speed routing
In the era of high speed networks, the internet service providers, enterprise networks are prone to speed of up to
several gigabits and the routers that connect these to the backbone should be capable enough to handle such high
data traffic. The routers in the network need to provide more differential services to support multimedia services.
With the use of data, voice and video in the same network, there is an urge for more bandwidth. There is a paradigm
shift to design routers that to support high data traffic.
Router designs
The need to design fast routers is in the high as the processing of routing table lookup are getting complex. The
need to design hardware based route look up table and forwarding is needed for high speed data transfer and reduce
the overhead processing on the routers. In this section we will see some of the design architectures for next
generation routers.
Bus based router architectures and multiple processors
With the help of multiple processors, there is an improvement in the shared-bus router architecture that speeds up
the packet forwarding when the packet arrives at the input terminal. The load on the system bus is reduced to a great
extent and packet forwarding is speeded up by using a route cache of frequently visited addresses.
Bus based router architectures with multiple forward engines
To achieve multiple high packet processing rates, multiple forward engines are connected in parallel. The following
diagram shows the same.

Forwarding engine row bus
Control bus
Forward engine column bus
Data bus
NI: network interface
FW: Forwarding engine
The forwarding engines work in parallel and it performs the routing function and decides through which output
interface the packet should be forwarded. The packet is then moved from the source interface module to the
destination interface module. Forwarding only the IP header eliminates any unnecessary overhead on the forwarding
engine. The choice of this architecture removes the bottleneck on the interfaces of the router and can increase the
port density of the router.
-Switch based router architectures with multiple processors
To even further reduce the overhead on the interface of the routers, the IP routers were designed with switch fabric.
This method provides sufficient bandwidth for packets between interface cards and improves the transmission
bandwidth to several orders of magnitude. Multi gigabit router is an example for this type of configuration. This
design has dedicated forward engines with route caches in them. The multi gigabit router consists of multiple
interface cards and forwarding engines connected to a high speed switch.
FG
FG
FG
Resource control
NI NI NI

Switch fabric
In the forwarding engine mechanism if the packet’s outbound interface is not found in the route cache, then the
CPU’s routing table is used to determine the same. If all the incoming packets are of the same type, then the
overhead on the main CPU increases drastically leading to a slow path.
Therefore the performance of a router using the route cache technique depends on
(1) How big the cache is
(2) What the performance of the slow path is (as some of the traffic might need this route).
As we can see from the design of the high speed routers, the architecture decides how fast a router would function.
The major deciding factors:
(1) The router needs enough processing power to process several million packets per second.
(2) It has to be keenly noted that the certain functions that re very critical can be processed in the hardware
and remaining functionalities can be pushed to the software.
Line card
Line card
Line card
Route Processor
Forwarding engine
Forwarding engine
Forwarding engine

5. Dense wavelength division multiplexing
It is a technology that puts different sources of data on the same optical fiber. The different channels of data can
have its own light wavelength. Using DWDM, up to 80 channels of data can be multiplexed into a single light
stream can be carried over the optical fiber. -A multiplexed time division signal is carried by each of the channels.
We are multiplexing different channels supporting different data rates; it is possible to have IP data, ATM traffic
within the same optical fiber. The corresponding channel can be demultiplexed at the receiving end. Therefore
DWDM is protocol and bit rate independent. As the demand for transmission bandwidth increases, upgrades can be
done either by simple equipment upgrades or by increasing the wavelength of the fiber.
DWDM has the following advantages
(1) DWDM being physical layer architecture, it can support multiple data formats such as IP data, ATM traffic
etc.
(2) It can meet the demands of point to point links in metropolitan and enterprise networks.
(3) It can provide high bandwidth services for most of the intensive multimedia applications.
DWDM system functions
As we said earlier DWDM can carry signals of different wavelength around its core and the following diagram
depicts the same.
The DWDM has the following components:
(1) Signal generator: It contains a solid state laser that generates the narrow bandwidth signal that carries the
digital data.
(2) Combiners: The combiner multiplexes are the incoming signals and transmitted over the optical fiber.

(3) Transmitting signals: The signals are multiplexing are transmitted over the optical fiber cable. If more
signals are multiplexed, loss will be more. Control variables such as channel variables, laser power levels
can minimize these effects.
(4) Optical amplifiers: If the signal is to be carried over the optical fiber it needs to be amplified to certain level
for its propagation over the fiber.
(5) Separators: At the receiving end, a separator is placed to demultiplex the signal. The demultiplexed signals
are received by a photodetector.
We can divide the DWDM backbone networks into two groups
(1) Point to point DWDM
(2) DWDM wavelength routing with electronic time division multiplexing
Point to point DWDM networks
The most common type of DWDM is the point to point network. It consists of electronic nodes coupled with
DWDM nodes. The electronic node can be any IP router or an ATM switch. The signals are multiplexed and
demultiplexed at the receiving end by the DWDM nodes located in between the electronic nodes are shown in
the figure. The DWDM node consists of multiplexers/demultiplexers, electrical to optical convertors at the
transmission side and optical to electrical convertors at the receiving side. Number of channels can be
embedded onto a DWDM point to point network each having a different wavelength. One wavelength channel
can be considered as a single light beam carrying data back and forth.
One disadvantage of point to point is that the bandwidth of the channel is not fully utilized.

Wavelength routing with electronic TDM
To fully utilize the bandwidth of the used channels, this type of DWDM architecture was used. The wavelength
routing consists of wavelength routers to configure the network topology and the electronic network nodes are used
for multiplexing and demultiplexing. This type of combined optical and electrical network can be used in IP
networks, ATM networks and SONET networks. The network topology can be changed according to the
requirement of the current traffic and can be static for a longer duration when required. The advantage in this
architecture is to utilize the maximum bandwidth of the available channels. The traffic calculation among the
DWDM channels should be taken care into account while designing the wavelength routers.
DWDM active access optical networks
In this type of DWDM networks, each channel has more bandwidth that it can carry several types of services. This
can be thought of applying to TDM to all the individual channels and can support integrated services. The
asynchronous transfer mode can be used as the TDM protocol in the channels which can tremendously increase the
channel capacity of the channels. Though this type of ATM introduced enhances the bandwidth capacity, the
installation of ATM switches at the access points would increase the operational cost to an extent though the fiber
cost are kept on the lower side.
DGDGOooooooOFNODJFNJ
OPTICAL DOMAIN
ETDM
node
ETDM
node
ETDM
node
ETDM
node
WR
W
R
T
D
M
n
o
d
e
W
R
WR

Thus the optical point to point DWDM networks have been used widespread and have proved itself dominant in
most of the multimedia applications that require high bandwidth. Whereas the all optical network to become more
predominant the processing optical power should be more. Implementing TDM in active DWDM might increase the
installation cost. Going further there is a possibility to implement active DWDM across the backbone network to
increase the channel capacity.

6. Quality of service
The next generation multimedia applications require bandwidth in terms of gigabit and the corresponding quality of
service is more intensive to support these applications. The network routes are selected in such a way to ensure that
the QOS to the existing connection are not hampered as well as provide better QOS to the incoming high priority
multimedia applications. The aim of the quality of service is to ensure the requested quality of service reaches the
end user application in the best possible route.
The quality of service of any network decides how well the network resources respond to the inbound traffic and
how efficiently data are prioritized as per the need of the service. If the delay in the network, jitter can be minimized
the required quality of service can be achieved. There are basically two qualities of service architectures that are
predominantly used in the backbone networks.
(1) Differentiated services
(2) Integrated services
Differentiated services
In case of differentiated services, priority for the packets is directly specified in the TOS (type of service) of the IP
header. In differential services, the packets are divided into classes and the quality of service is provided depending
on the priority number marked on the packet header. This type of service is per hop based. A six bit-bit pattern in the
IPv4 TOS field defines the class of service based on bit patterns.
The signaling of resource allocation is eliminated which results in a scalable coarse grained QOS. Based on the
priority listed on the packets, the latency, jitter and network resources are decided.

Differentiated services architecture
It employs a small well defined set of building blocks from which a variety of behaviors is built. When the packets
are transmitted over the network, each packet is marked by a small bit pattern determining the class of service to be
received. The architecture to deliver an end to end QOS is based on the following two aspects.
(1) Packet marking
In this, the packet are marked with a 6 bit pattern in the TOS header and can offer variety of class of
services up to (2^6). The field is known as the differential services field (DS) and the precedence of the bits
is known as differential services code point (DSCP). The last two fields are unused and only first 6 bit
patterns are used to determine the class of service. The following diagram explains the same.
Per hop behaviors
A meaningful QOS can be provided to the packets that have the same DSCP value and they follow a
specific direction and are called as behavior aggregate. Packets from multiple applications could have the
same DSCP number and belong to the same BA. Per hop behavior refers to the packet scheduling, shaping
and policing behavior of a node.
Integrated services
To support and provide quality of service to real time media applications such as voice over IP, multimedia
processing integrated type of model is required. The integrated services use the resource reservation and the
admission control to provide end to end QOS across the network. The resource reservation protocol ensures
the bandwidth requested by the media application is available throughout the end to end path. If it is
available, then the application starts to send its packets to the network. Based on the end to end signaling,
we have the following functions provided by the integrated services
(1) Admission control
Based on the existing traffic, determine if a new flow can be granted the without affecting the existing
flow and give the best possible QOS for the new flow.

(2) Classification
Provide packets that need a particular level of QOS.
(3) Policing
If the network is congested with heavy traffic packets can be dropped. It also provides verification for
user permission to request RSVP services.
(4) Packet scheduler
It will order the packets based on the required quality of service.
QOS routing and different network components
The QOS routing is a resource reservation to provide the guaranteed service. To meet the individual
connection and reduce the call blocking rate are very parameters for QOS routing.
QOS and resource reservation
The QOS and the resource reservation go hand in hand. The required QOS can be provided if and only if
the necessary resources like the bandwidth, router processing power is reserved once the connection gets
started. The job of routing is to ensure that the bandwidth requested by the end node is available across the
path of the routing and make sure that data transmission of the new connection is not affected by other
existing connections. Before even assuring the end nodes of bandwidth, the complete path must be selected.
QOS routing and QOS negotiation
Whenever the routing algorithm fails to find a feasible path for the reservation requested by the end
application due to a feasible path that is already been given to an existing connection, the QOS can give an
alternate feasible path to the end application with lower QOS. The QOS routing would calculate the best
available path by negotiating with the application. If the negotiation with the application is successful, the
QOS routing would go further to allocate the best available path.
QOS routing and the best effort traffic
The primary of the QOS routing is to maximize the available total number of feasible paths and minimize
the call blocking rate. Other aspect is to providing best effort traffic when the network is congested and
increase the responsiveness and the throughput. The throughput of the best effort traffic will suffer if the
overall traffic is not estimated properly. If a particular link has a mild QOS traffic and heavy best effort
traffic, imposing more paths in this link would have a bearing on the best effort traffic and the link would
start to drop packets irrespective of the priority assigned to the packets (in case of best effort traffic).

Integration with network components:
Guaranteed services can be provided to any application with the help of other network components in the
network. The routing algorithms should be designed in a way how the resources are allocated and
scheduled. The following properties are necessary for a better QOS traffic.
(1) Generality
Different multimedia applications require different kind of data rates. Designing a general routing
algorithm captures the common messaging and routing structure.
(2) Extensibility
As the applications are growing in size, there is new service types needed to be supported by the
routing algorithms. It is very important to design algorithms to support such extensibility of
applications.
(3) Simplicity
During the time of congestion, the logic behind debugging the network should be kept simple as the
algorithm should be made understandable.
On the whole the QOS routing is an important network function for high speed multimedia
applications. Routing algorithms should be designed in such a way that
(1) Find routes that satisfy the QOS constraints
(2) Make efficient use of the network resources to offer high QOS to end applications

7. Resilience
Unlike the traditional PSTN telephony, the data carried by the backbone networks are either faulty due to the
malfunctioning of the intermediate routers or have security breaches. Due to the development of predominant
internet applications like voice over IP, web browsing, e commerce, the lack of an effective backbone reliability
pose a significant challenge in the market of interactive applications. The internet backbone has got the following
characteristics:
(1) The mean time to failure is quite low as compared to public switched telephone network.
(2) The internet traffic is rerouted every 3 days and the time taken to repair once a fault is detected is less than
two minutes.
(3) The total number of network paths that contribute the long term unavailability of network bandwidth is
very less.
In a cluster of IP routers that handle multiple paths between different nodes including DWDM elements, there is a
possibility to embed resilient functionality in one or more layer of the IP stack to ensure fault protection. Consider
there are two routers R1 and R2 and they are connected by a transmission circuit. In the event of failure of the
transmission connection, the IP layer resilience mechanism should divert the route by routing it via some other
router. The way in which the IP routing resilience mechanism depends on a number of issues discussed in this
section. The resilience of IP-DWDM backbone architecture can be achieved in the following three ways.
(1) Unprotected architecture
(2) Protected architecture
Unprotected architecture
The routing protocols such as ISIS and SPF (shortest path first) maintain and work together to find an alternative
route. The fault recovery has basically three steps
(1) Fault detection
(2) Fault propagation
(3) Shortest path first calculation
Fault detection is quite fast as it detects the link failure and fault propagation informs all other routers in the
network regarding the particular link failure. The role of the shortest path first calculation is to find an
alternate path for the packets to travel from one router to another. The SPF calculation is done along the
route path where all the routers are involved. The unprotected architecture depends on reconvergence of the
dynamic protocols for the node failures.

Protected architecture
The most common type of protection in an IP-DWDM network is 1+1 protection. It provides a dedicated alternate
link between 2 nodes if 1 link fails. This is quite simple to implement but the network resources would add up to the
installation cost. In case of long distances, the regeneration and additional wavelength translators would even more
add up to the total cost in setting it up.
To ensure that the cost is reduced, optical layer protection can be used which reduces the cause of link failures and
minimize transmission loss. But with the technology this would be a costly option and switching from 1+1
protection to optical layer protection is not quite feasible.

8. Future aspects of a good backbone network
In the last decade there has been a tremendous increase in the usage of bandwidth due to IP acting as a gateway for
interactive applications. The future traffic load might vary as services like IP television, video on demand would be
dominate the IP backbone traffic. The telecommunication industry is growing at a rapid pace and the need for more
bandwidth is on the higher side. To meet the high demands, we should design the traffic models keeping these kinds
of services in mind.
(1) Internet protocol- television
(2) Video on demand
(3) Peer to peer
(4) Content delivery networks
(5) Virtual private networks
A service aspect of a network model has to be designed which can accommodate the type of services mentioned
above. As these types of services include generating content from a central site, predominantly a star oriented model
is to be used. (For services like P2P, user generated content and CDN).
Let us see some of the factors that determine in designing a good backbone network for high end services.
Reliability
The growth of internet has been tremendous over the last decade. The speed of the routers has increased from
2.5Gbps till 100Gbps. As the speed increases, the complexity in building more reliable routers is on the high for
supporting Video on demand, IPTV etc. The internet is made up of countless number of nodes and a failure of a
single node would lead loss to several million dollars. So it must be reliable and should be able to trouble shoot in
the event of any failure.
Scalability
The internet traffic is increasing at a rapid rate every 12 months and the need to address real time data traffic is high.
So the issue of scalability should be addressed or less the service providers would find it really difficult to expand to
address the high demands of their customers. The design of the network is very important to suit the business growth
as well as provide reliable communication without multiplying the number of routers in the network making it even
more complex. Rather than increasing the number of routers in the clusters, it would be feasible if the capacity of a
single device can be increased. The network performance, service growth can be met through this.
Energy Conservation
Energy saving techniques like smart power supplies, smart line card set up, employing smart service and dormancy
process; we can save up to 23% less power in turn would stabilize the energy requirements as well.

ABBREVIATIONS
OSI- Open Systems Interconnection
LAN- Local Area Network
DWDM- Density Wavelength Division Multiplexing
MPLS- Multi Protocol Label switching
RIP- Routing Information Protocol
ATM- Asynchronous transfer mode
IP- Internet Protocol
OSPF- Open Shortest Path First
IETF- Internet Engineering Task Force
FTTH- Fiber to the Home
FTTO- Fiber to the Office
MAC- Media Access Control
VLAN- Virtual Local Area Network
SNMP- Simple Network Management Protocol
CRC- Cyclic Redundancy Check
TCP- Transmission Control Protocol
TTL- Time to Live
SONET- Synchronous Optical Networking
WR- Wavelength Router
TDM- Time Division Multiplexing
EDTM- Electronic Time Division Multiplexing
QOS- Quality of Service

TOS- Type of Service
DS- Differential Services
DSCP- Differential Services Code Point
RSVP- Reservation Protocol
PSTN- Public Switched Telephone Network
ISIS- Intermediate System to Intermediate System
SPF- Shortest Path First
CDN- Content Delivery Networks
IPTV- Internet Protocol Television

REFERENCES
(1) Cisco’s white paper on differential services
http://www.cisco.com/en/US/technologies/tk543/tk766/technologies_white_paper09186a00800a3e2f_p
s6610_Products_White_Paper.html
(2) Cisco’s introduction to integrated services
http://www.cisco.com/c/en/us/products/ios-nx-os-software/integrated-services/index.html
(3) Service oriented models for future backbone networks
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5755639
(4) Cisco’s introduction on evolution of switches
http://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_1-2/switch_evolution.html
(5) Cisco’s white paper on QOS routing
http://www.cise.ufl.edu/~sgchen/papers/QoSRoutingSurvey99.pdf
(6) IP routing concepts
http://www.tcpipguide.com/free/t_IPRoutingConceptsandtheProcessofNextHopRouting.htm
(7) Resilient characteristics of internet backbone
http://www.tik.ee.ethz.ch/file/1371851fc3fca76756817a5ad3e6aac6/ISW00-1.pdf
(8) An overview of DWDM networks
http://canrev.ieee.ca/canrev37/song_eng.pdf -
(9) Cluster Routers- the best choice of future backbone networks
http://wwwen.zte.com.cn/endata/magazine/ztetechnologies/2010/no10/articles/201010/t20101021_1931
99.html
(10) IP router architecture- An overview
http://www.cs.virginia.edu/~cs757/papers/awey99.pdf

BACKBONE NETWORKS

More Related Content

What's hot (20)

Similar to BACKBONE NETWORKS (20)

BACKBONE NETWORKS