Faculty:

BCS502 – COMPUTER NETWORKS Prof. Ashok Herur

[email protected]
Mobile: 91641 01399
 Module 3

Network Layer

Computer Networks - Prof Ashok Herur 2

 Main topics in Module 3 – Network layer

• Network Layer Services

• Packet Switching

• IPv4 addressing, IPv4 datagram, IPv6 datagram

• Introduction to Routing algorithms

• Unicast Routing algorithms

• Multicast Routing algorithms

Computer Networks - Prof Ashok Herur 3

Network layer services

Computer Networks - Prof Ashok Herur 4

Network layer services
• Functions of the Network Layer:
• The Network layer makes sure that the packets are sent all the way from the
source to the destination, in as efficient a manner as possible.
• Data link layer just moved the frames from one node to the adjacent one.

• To be able to do this, the Network layer must:

• Know the topology of the network (the set of all routers and links);
• Balance the load to avoid overloading some links while leaving others
under-utilised.

Computer Networks - Prof Ashok Herur 5

Review of Packet switching.
• Every message is segmented by the Transport layer to comply with the packet
size limit imposed by the networks.
• Each packet has information of the message to which it belongs, as well as the
destination address.

• There is no dedicated, end-to-end connection.

• The packets are then received, buffered, processed and forwarded by the
numerous routers along the way.
• Hence, there is a certain delay encountered at each router.

Computer Networks - Prof Ashok Herur 6

 Review of Packet switching

• Advantages:
• Since the bandwidth of a link is not dedicated (but shared with packets
from numerous hosts), the utilisation of the bandwidth is much higher.
• No set-up time.

• Two methods of Packet switching :

• Connectionless (Datagram Packet switching);
• Connection-oriented (Virtual Circuit Packet switching).

Computer Networks - Prof Ashok Herur 7

Packet switching – Hosts sharing the bandwidth

Computer Networks - Prof Ashok Herur 8

IPv4 and IPv6

Computer Networks - Prof Ashok Herur 9

Network layer is called INTERNET layer in TCP/IP model

Computer Networks - Prof Ashok Herur 10

 Internet Protocol (IP)
• IPv4 and IPv6 are the most popular protocols used on the Internet, with IPv6
succeeding IPv4.

• They define the logical addressing of the hosts as well as the datagram
(packet) format.

• The hosts as well as the routers understand this protocol.

• The logical address is used to route the datagram from router to router, till
they are delivered to the destination host.

Computer Networks - Prof Ashok Herur 11

IPv4 addresses
• Every device on the WAN is identified by an unique address.
• All IPv4 addresses are 32 bits long.
• They are represented as four octets in dotted decimal format.
• Example: 132.10.68.73
• The IP address has two components:
• The Network ID – Allotted by a global authority
• The Host ID – Allotted locally by the network administrator

Computer Networks - Prof Ashok Herur 12

IPv4 address classes
• IP addresses are divided into 5 classes: A,B, C, D, E.

• Class D addresses are used for multicasting while class E is not used

• The Network ID is 1, 2 and 3 bytes long in Class A, B and C respectively, with

the remaining part being the Host ID.

• Therefore Class A networks are very large and can accommodate a huge
number of hosts within them, while Class C networks are very small.

Computer Networks - Prof Ashok Herur 13

IPv4 address classes

Computer Networks - Prof Ashok Herur 14

IPv4 address classes

Computer Networks - Prof Ashok Herur 15

IPv4 address classes

Computer Networks - Prof Ashok Herur 16

Private IP addresses
• Private IP addresses are valid only within an organization
• The same addresses can be used in multiple organizations

• When leave the network of the organization, the private addresses are
converted to public addresses by the proxy server, and vice-versa.

• Private addresses:
– Class A: 10.0.0.0 to 10.255.255.255 --- 1 network
– Class B: 172.16.0.0 to 172.31.255.255 --- 16 networks
– Class C: 192.168.0.0 to 192.168.255.255 --- 256 networks

Computer Networks - Prof Ashok Herur 17

 Address Resolution Protocol (ARP)
• Routers use the network portion of an IP address to route the packet to the
network containing the destination host.
• When the packet reaches the ‘gate’ of the network (belonging to an
organization or an ISP), the Gateway device uses the host portion of the IP
address to find out where the particular host is, and then deliver the packet.
• Since hosts are uniquely and permanently identified by their physical (MAC)
addresses, the IP addresses should be mapped to the physical addresses for
eventual delivery of the packet.
• ARP helps in this mapping.

Computer Networks - Prof Ashok Herur 18

 Address Resolution Protocol (ARP)

Computer Networks - Prof Ashok Herur 19

 Subnetting
• Network administrators divide large networks, into smaller ones called sub-
networks (or just subnets), to improve network performance.

• Another reason for having subnets is to reduce the size of a broadcast

domain, when broadcast traffic begins to consume too much of the
available bandwidth.
• Broadcasts are sent to all hosts on a network or a subnetwork.

• Having subnets also leads to efficient routing of packets within the network.
• The subnet ID is created by borrowing bits from the host portion of the IP
address

Computer Networks - Prof Ashok Herur 20

 Subnetting – Example 1
The network 208.154.29.0 needs to be divided into 2 subnets;
Subnet 1 has to support 100 hosts while Subnet 2 has to support 80 hosts.
Design and calculate the range of addresses to be used in both the subnets.

• The network 208.154.29.0 is a Class C network

• The host ID consists of the last 8 bits of the IP address
• It can support a total of 254 hosts (when not sub-netted)
• How many bits are required for the subnet ID? ---- Only 1 bit in this case
• How many bits remain for the host ID portion? ---- 7 bits
• How many hosts can each subnet support ? ---- 126 hosts each
• So, the stated requirement of 100 hosts and 80 hosts can be met.

Computer Networks - Prof Ashok Herur 21

 Subnetting – Example 1

Computer Networks - Prof Ashok Herur 22

 Subnetting – Example 2
The network 208.154.29.0 needs to be divided into 3 subnets;
The three Subnets have to support a maximum of 50 hosts each.
Design and calculate the range of addresses to be used in the 3 subnets.

• The network 208.154.29.0 is a Class C network

• The host ID consists of the last 8 bits of the IP address
• It can support a total of 254 hosts (when not sub-netted)
• How many bits are required for the subnet ID? ---- 2 bits
• How many bits remain for the host ID portion? ---- 6 bits
• How many hosts can each subnet support ? ---- 62 hosts each
• So, the stated requirement of 50 hosts each can be met.

Computer Networks - Prof Ashok Herur 23

 Subnetting – Example 2

Computer Networks - Prof Ashok Herur 24

Masking
• A ‘mask’ is used to mask out the host portion of an IP address and get only
the network (and sub-network) ID

• A mask is a string of 1’s followed by a string of 0’s

• The number of 1’s equal the number of bits allotted for the network and
sub-network ID

• The masking operation is performed by a logic AND operation of the given IP

address and the mask for that network

Computer Networks - Prof Ashok Herur 25

Masking
• The mask for a Class A network with no subnets is 1111111100…….0, which
is 255.0.0.0; This is called as Default mask for Class A networks.
• The Default mask for a Class B network is 255.255.0.0

• The mask for a Class A network with 4 subnets will have 10 ones followed by
22 zeros: 11111111 11000000 00000000 00000000, which is 255.192.0.0
• The mask for a Class A network with 256 subnets will 255.255.0.0 (same
as the default mask for Class B)

Computer Networks - Prof Ashok Herur 26

Masking
• Find out the Network ID and the Sub-network ID (if any) for the host with an
IP address 140.37.163.25 if the mask is 255.255.224.0

• 140.37.163.25 – 10001100. 00100101. 10100011. 00011001

255.255.224.0 – 11111111. 11111111. 11100000. 00000000

• The AND operation of these 2 would give us the Class B Network ID as

140.37.0.0 and the Sub-network ID bits as 101

Computer Networks - Prof Ashok Herur 27

Masking

Computer Networks - Prof Ashok Herur 28

Classful addressing
• Classful network design for IPv4 defined the network address as one or more 8-bit
groups, resulting in the blocks of Class A, B, or C addresses.

• Thus, the smallest allocation and routing block contained only 256 addresses - too
small for most enterprises, and the next larger block contained 65536 addresses -
too large to be used efficiently by even large organizations.

• This led to inefficiencies in

• Address use: Large number of unused addresses in Class B and A
• Routing: The large number of allocated small (class-C) networks created large routing tables
and heavy demand on routing equipment.

Computer Networks - Prof Ashok Herur 29

Classless addressing
• Classless Inter-Domain Routing (CIDR) allocates address space to Internet
Service Providers (ISPs) and end users on any address bit boundary, instead
of on 8-bit segments.

• CIDR is based on variable-length subnet masking (VLSM), which allows a

network to be divided into subnets of various sizes.

• In the CIDR notation for an IP address, the network address (or routing prefix)
is written with a suffix that indicates the number of bits of the prefix, such as
192.168.2.0/24.

Computer Networks - Prof Ashok Herur 30

Classless addressing
• The /24 indicates that the network and subnetwork address is 24 bits long
• Or that the host address is (32-24) = 8 bits in length

• 192.168.2.0/24 is the same as the older notation which mentioned the

network ID and the Mask: 192.168.2.0/255.255.255.0

• The IPv4 block 182.168.100.0/22 represents 210 = 1024 addresses from

192.168.100.0 to 192.168.103.255

Computer Networks - Prof Ashok Herur 31

IPv4 datagram header

Computer Networks - Prof Ashok Herur 32

 IPv4 Datagram Header

Computer Networks - Prof Ashok Herur 33

IPv4 Header
• The IP header is built up in blocks of 32 bits, and is always an integral number
of 32-bit words.
• The IP header is divided up into various fields, each having a defined function.
• The 4-bit version field indicates the release version of the IP that is used in this
datagram; IPv4 would have it as ‘0100’
• The 4-bit header length field identifies the length of the IP header in
multiples of 32 bits (or 4 bytes)
• The minimum value for a valid header is 5 (means 5*32 bits = 20 Bytes),
Maximum is 15 (means 15*32 bits = 60 Bytes).

Computer Networks - Prof Ashok Herur 34

IPv4 Header
• The Service type field specifies the parameters for the type of service
requested by the packet from the network.
• This field specifies things like:
• The priority to be given to the datagram;
• Delay to be minimised;
• Cost to be minimised;
• Throughput to be maximised;
• Reliability to be maximised.
• The networks use these parameters to define the handling of the datagram
during its transport.

Computer Networks - Prof Ashok Herur 35

MTU, and Fragmentation
• Every network specifies the maximum size of the data field in a frame. This
limit is called the maximum transmission unit (MTU) of that network.

• Any datagram encapsulated in a frame must therefore be smaller than the

MTU for that network.

• What happens in the scenario depicted below?

Computer Networks - Prof Ashok Herur 36

IPv4 Header
• The Total length field indicates the length of the entire packet, in bytes.
• With a 16-bit field, it allows packets as long as 65,535 bytes.

• When a packet is divided into smaller fragments to comply with the MTU
restriction, each fragment of the packet will have the same Identification field
as the parent packet.

• Of the three bits in the Flag field, the first one is not currently defined/ used.

• The second flag indicates whether fragmentation is allowed. It is called the

Don't fragment (DF) flag. If the flag is set to 1, fragmentation is not allowed,
and if it is set to 0 fragmentation is allowed.

Computer Networks - Prof Ashok Herur 37

IPv4 Header
• If the DF flag is set to 1, a datagram will be lost if it has to cross a network that
cannot handle its size.
• The third flag is called the More fragments (MF) bit. It is used to indicate that
there are more fragments to follow, so if a datagram is fragmented this is set
on all except on the last fragment.
• The Fragment offset field is 13 bits long, and is used to indicate the position of
the data in a fragment relative to the beginning of data in the original
datagram.

Computer Networks - Prof Ashok Herur 38

IPv4 Header
• The TTL (Time-To-Live) field contains a initial integer when a packet is sent out
by the source.
• This number is decremented by 1 by every router which forwards it.
• When the TTL value becomes zero, routers discard them.
• The TTL field thereby puts a limit to the number of hops (life) a packet can go
through and deals with stray packets which keep looping in the network.

Computer Networks - Prof Ashok Herur 39

IPv4 Header
• The Header Checksum field contains the checksum computed on all the bits of
the IP header (not the Data)
• This checksum is verified by every router before forwarding the packet.
• If the checksum doesn’t match, it indicates an error(s) in the header portion,
and the router therefore discards the packet
• It does not attempt to inform the source that the packet is being dropped.

Computer Networks - Prof Ashok Herur 40

IPv6 datagram header

Computer Networks - Prof Ashok Herur 41

 IPv6 Datagram Header

Computer Networks - Prof Ashok Herur 42

IPv6 Header
• Version field is the same as in IPv4 (of course, the value is now 6 (0110).
• Traffic class is same as “Type of Service” in IPv4.
• Payload length is same as “Total length” in IPv4.
• Hop Limits is same as TTL in IPv4.
• IP address length increased from 32 bits to 128 bits.
• Flow Label field is used by a source to label the packets belonging to the
source and destination, in order to request special handling by intermediate
IPv6 routers.
• Between a source and destination, multiple flows may exist because many
processes might be running at the same time.
Computer Networks - Prof Ashok Herur 43
IPv6 address classes
• IPv6 does not have classes, but its address space is divided into different types
based on how they are used. Some types of IPv6 addresses include:
• IPv6 multicast address: A major address class in IPv6, with the prefix FF00:
allocated to all IPv6 multicast addresses.
• IPv6 unicast address: Used to identify a single network interface. Each
node on an IPv6 network has at least one unique unicast address.
• IPv6 anycast address: Used to address multiple interfaces on a single
multicast address.

Computer Networks - Prof Ashok Herur 44

Routing algorithms

Computer Networks - Prof Ashok Herur 45

 Routing algorithms
• A routing algorithm helps decide which output line, an incoming packet
should be transmitted on.

• In Datagram Packet switching, this decision must be made anew for every
arriving packet since the ‘best route’ may have changed since the last packet.

• In VC Packet switching, these decisions (at every router along the way) are
made when the VC is being set up. Thereafter, the packets are forwarded on
the pre-decided route.

• The output of the routing algorithm is stored as a Routing (look-up) Table, one
for each router.
Computer Networks - Prof Ashok Herur 46
 Issues in Routing
• Finding the ‘best’ route is not always easy.
• The topology keeps changing: Lines may be added, may go down and would
be brought back later.
• A packet may have to travel through heterogeneous networks.
• Delay variability is high.
• Have to deal with Congestion in the network.
• Multicasting and broadcasting is difficult.

Computer Networks - Prof Ashok Herur 47

 Routing algorithms
• Desirable features of a routing algorithm:
• Correctness
• Simplicity
• Robustness (be able to cope with changing topology and traffic)
• Stability (quickly reach an equilibrium {route} and stay there)
• Fairness (to all users)
• Efficiency (at a global level)

• Often, Fairness and Efficiency are contradictory goals.

Computer Networks - Prof Ashok Herur 48

Fairness versus Efficiency
• Let’s say that there is
enough traffic between A
and A’, B and B’, C and C’ to
saturate the horizontal links.

• To maximise the total flow,

X to X’ traffic should be cut
off since it would ‘eat up’
the bandwidth on 3 links.

• Globally efficient but not

fair to X and X’

Computer Networks - Prof Ashok Herur 49

 Routing algorithms
• Routing algorithms try to:
• Minimise the packet delay (mainly depends on the number of hops).
• Reduce the physical transmission distance, as far as possible.
• Reduce the bandwidth consumed per packet.
• Improve the overall network throughput.

• Routing algorithms can be classified as:

• Non-adaptive (Static) algorithms
• Adaptive (Dynamic) algorithms

Computer Networks - Prof Ashok Herur 50

 Non-adaptive routing algorithms
• Non-adaptive routing algorithms compute the route from node (router) I to
node J, in advance and download that to all the routers along the way.

• The chosen routes remain static for a long period of time (could be days or
weeks too).

• They do not base their routing decisions on estimates of the current traffic or
short-term changes in topology.

• Since it does not respond to changes or link failures, it is only used in

situations where the routing choice is clear (no choice) or has very few
options.
Computer Networks - Prof Ashok Herur 51
 Adaptive routing algorithms
• Adaptive routing algorithms base their routing decisions on estimates of the
current traffic and changes in the topology.

• These algorithms differ from each other in:

• Where they get their information (eg., from adjacent routers or from all
routers in their sub-net)
• When they change the routes (eg., when the topology changes or every
ΔT seconds as the load changes)
• What metric is used for optimization (eg., distance, number of hops, or
estimated transit time).

Computer Networks - Prof Ashok Herur 52

The Optimality principle
• It states that if router J is on
the optimal path from router I
to router K, then the optimal
path from J to K also falls along
the same route.

• If there was a better route

from J to K (say, r1) then the
optimal path from I to K would
have been (I - J - r1).

Computer Networks - Prof Ashok Herur 53

 Routing algorithms
• Shortest Path algorithm (also called Dijkstra’s algorithm)

• Flooding

• Distance Vector routing algorithm

• Link State routing algorithm

• Hierarchical routing algorithm

• Broadcast routing algorithm

• Multicast routing algorithm

Computer Networks - Prof Ashok Herur 54

 Shortest Path (Dijstra’s) routing algorithm
• It is a non-adaptive (static) routing algorithm.
• It is also called as SPF (Shortest Path First) algorithm.
• It helps identify the “shortest” path between a given Source Node and a
Destination Node.
• Procedure:
• Start with the Source Node.
• Mark it as permanent (‘seen’ node), and designate it as the working node.
• Set the tentative distance from the Source node to all other nodes to
“Infinity” (since, currently, we do not know how ‘far’ they are from the
Source node).

Computer Networks - Prof Ashok Herur 55

Shortest Path (Dijstra’s) routing algorithm
• While some nodes are still not marked permanent, do the following:
• Compute the tentative distance from the source to all nodes adjacent
to the working node.
• If this is shorter than the current tentative distance, then replace the
tentative distance and record the label of the working node there.
• Select the node with the smallest value and make it the new working
node. Designate the node permanent.

Computer Networks - Prof Ashok Herur 56

Shortest Path (Dijstra’s) routing algorithm - Example
• Find the shortest path
between node A and H.

• Using the labels,

backtrack from H to A,
and note down the
chosen path.

• Path: A – B – E – F – H
• Distance metric = 8 units

Computer Networks - Prof Ashok Herur 57

 Shortest Path (Dijstra’s) routing algorithm - Example

C • Example: Build the Routing

5 Table for Node A.
B
5
3
4
• Find the shortest paths from
4 D A to all the other Nodes.
E 4
6
A • Note the node to which A
6 should forward to reach the
2 1
3 0 3 respective nodes.

H • Note the distance metric to

F 1 G 2
0 those nodes.

Computer Networks - Prof Ashok Herur 58

Shortest Path (Dijstra’s) routing algorithm - Example
Destination Outgoing “Distance”
line
A - 0

B B 4

C F 8

D F 9

E F 5

F F 3

G F 12

H F 14

Computer Networks - Prof Ashok Herur 59

 Flooding algorithm
• It is a static flooding algorithm for distributing packets to every part of a
connected network in the shortest time.

• Flooding algorithms are used as a part of some routing protocols, generally

those used in ad-hoc wireless networks.

• Each node tries to forward every message to every one of its neighbours.
• This results in every message eventually being delivered to all reachable
parts of the network.

Computer Networks - Prof Ashok Herur 60

 Flooding algorithm
• Real-world flooding algorithms have to be more complex than this, since
precautions have to be taken to
• avoid wasted duplicate deliveries
• avoid infinite loops
• allow messages to eventually expire from the system.

• Selective Flooding Flood only in the direction of the destination.

• Practically useful in a few settings like Military Applications

Computer Networks - Prof Ashok Herur 61

 Distance Vector routing algorithm
• Distance Vector routing algorithm (also known as Bellman-Ford algorithm) is
an adaptive algorithm.
• These routing algorithms operate by having each router maintain a Table
(Vector) giving the best known “distance” to each destination and which line to
use to get there.
• The “distance” metric is normally the number of hops or the propagation time.
• The table at each router has one entry for each router in the subnet. For each
entry, it has two parts:
• The preferred outgoing line to use for that destination
• An estimate of the time or ‘distance’ to that destination

Computer Networks - Prof Ashok Herur 62

 Distance Vector routing algorithm
• Periodically (every 5-10 minutes), each router does the following:
• Exchange their Tables with all the neighbouring routers every 5-10 minutes.
• Estimate the “distance” to each neighbour.
• If the distance metric is hops, the distance to each neighbour is just one hop.
• If the distance metric is propagation delay, the router can measure it directly by
sending ECHO packets to each neighbour (the neighbour just Time-stamps and
sends it back)
• The Routing table is then updated using the Tables received from the neighbours
and the distance estimate to the neighbours.

Computer Networks - Prof Ashok Herur 63

Example 1 - DVR algorithm – Router J updating it’s Table

Computer Networks - Prof Ashok Herur 64

Example 2 - DVR algorithm – Router E updating it’s Table

C
5
B
5
3 4
4 D
6 E 4
A
6
2 1
3 0 3

H
F 1 G 2
0

Computer Networks - Prof Ashok Herur 65

Existing Distance Vector of Router E
Destination Outgoing Delay (ms)
line
A F 5

B C 8

C C 3

D D 4

E - 0

F F 2

G D 7

H D 9

Computer Networks - Prof Ashok Herur 66

Distance Vectors received by router E & Measurements
Destination Delay Delay Delay Delay
from A from C from D from F
A 0 9 9 3
B 4 5 14 7
C 8 0 5 5
D 9 5 0 6
E 5 3 4 2
F 3 5 6 0
G 12 4 3 9
H 14 6 5 11

5 3 5 3 Delay from E
Computer Networks - Prof Ashok Herur 67
 Updated Distance Vector of router E
Destination Outgoing Delay (ms)
line
• The entry for Self (E A A 5
to E) will not be B C 8
determined by the
C C 3
updating method
shown in the D D 5

previous slide. E - 0

F F 3
• It will always be “0”.
G C 7

H C 9
Computer Networks - Prof Ashok Herur 68
 Count-to-infinity problem
• One of the main drawbacks of the Distance Vector Routing algorithm is that
while “good news” travels fast, “bad news” take a long time to reach all the
routers in the subnet.
• The latter part is called “Count-to-infinity” problem.

• “Good news” is, for example, a link that was down earlier, coming back into
service.

• “Bad news” is, for example, a link going down.

Computer Networks - Prof Ashok Herur 69

Count-to-infinity problem – “Good news travels fast”
• “A” was initially down and all the other
nodes recorded it as “∞” (a “.” in the
figure).

• During the first exchange of Tables,

after A comes back, B will know the
fact and mention that A is just one hop
away.

• On subsequent exchanges, the other

too learn about it.

• If there are N nodes in a linear stretch,

they will all know about it after (N-1)
exchanges.
Computer Networks - Prof Ashok Herur 70
Count-to-infinity problem – “Bad news takes a long time”
• “A” was initially working and all the other nodes
recorded the respective number of hops to A.

• “A” suddenly goes down.

• During the first exchange of Tables, B realises

that A is not directly reachable.

• However, C tells B that it can reach A in 2 hops;

So, B concludes that it can reach A in 3 hops
(one hop to C and 2 hops from C to A).

• On the next exchange, C will update its Table to

say that it can reach A in 4 hops (one hop to B
and another 3 hops from B to A).
Computer Networks - Prof Ashok Herur 71
 Count-to-infinity problem – “Bad news takes a long time”
• This progresses very slowly as shown in the figure.

• It takes a long time to reach Infinity (generally defined as 16 hops).

• A few approaches have been proposed to overcome this, but in practice,

none of them are effective.
• One such approach was to prevent routers from advertising their best
paths back to their neighbours from which they heard about them.

Computer Networks - Prof Ashok Herur 72

 Link State routing algorithm
• Distance vector routing had a major drawback in that it took too long to
converge especially to ‘bad’ news (like a link going down).

• Link State routing has the following 5 steps:

1. Discover the neighbours and learn their addresses
2. Measure the delay or cost metric to each of the neighbours
3. Construct a packet holding all that it has just learned
4. Send this packet to, and receive packets from, all other routers
5. Compute the shortest path to every other router

Computer Networks - Prof Ashok Herur 73

 Link State routing algorithm
• Building the Link State packets can either be done periodically or when some
significant event occurs (such as a line or neighbour going down or coming
back again).

• In effect, the complete topology is known to every router.

• Then, Dijkstra’s algorithm can be run at each router to find the shortest path
to every other router.

• Steps 1 and 2 can be completed by sending an ECHO packet to the neigbours.

Computer Networks - Prof Ashok Herur 74

Link State routing algorithm
Step 3. Construct a packet holding all that it has just learned

Computer Networks - Prof Ashok Herur 75

Link State routing algorithm
Step 4. Distribution of Link State Packets.
• Flooding is used to distribute the packets.
• To keep the flood in check, Sequence Number and Age fields are used.
• For every new Link State Packet sent by a source router the Sequence
number is incremented by one.
• If a packet with a sequence number lower than the highest one
seen so far arrives, it is discarded as obsolete, as the router has
more recent data.

Computer Networks - Prof Ashok Herur 76

Link State routing algorithm
• The Age field is a preset number when the packet is formed.
• Each router decrements it by one before flooding it further.
• When the Age becomes zero, it is discarded.
• The idea is to make sure that a packet doesn’t live on for a very long
time.
Step 5. Computation of the new shortest paths
• This is done by using Dijkstra’s algorithm.
• OSPF (Open Shortest Path First) is a very popular Link State Protocol.

Computer Networks - Prof Ashok Herur 77

 Hierarchical routing
• As networks grow, the router routing table grows proportionally.
• This would consume more memory and more search and updation time.
• In a hierarchical network, routers are organised into regions, with each router
only knowing all the details of how to route packets to destinations within its
own region.
• To route packets to destinations in other regions, they send the packets to a
designated router within their own region.
• Regions may be further combined into clusters, and so on.

Computer Networks - Prof Ashok Herur 78

Hierarchical routing

Region 1 Region 2
1B

1A 1C

Region 3 Region 4 Region 5

Computer Networks - Prof Ashok Herur 79

Hierarchical Routing Table for 1A

Destination Outgoing Hops

line
1A - -

1B 1B 1

1C 1C 1

2 1B 2

3 1C 2

4 1C 3

5 1C 4

Computer Networks - Prof Ashok Herur 80

 Hierarchical routing - Examples
• With an eye on minimizing the length of the Routing Table, how do you
group 120 routers into:
1. Regions (2-level hierarchy)?
2. Regions and Clusters (3-level hierarchy)?

• What will be the length of the Routing Table in both cases?

• Answers:
• 12 regions of 10 routers each; Table of each router will have 10 entries for each router
in its own region, and 11 entries for each of the other 11 regions. A total of 21 entries.
• 4 Clusters each having 5 Regions, with each region having 6 routers each; Table of each
router will have 6 entries for each router in its own region, 4 entries for each of the
other 4 regions in its own cluster, and 3 entries for each of the other 3 clusters. A
total of 13 entries.
Computer Networks - Prof Ashok Herur 81
 Routing Information Protocol (RIP)
• RIP is a variation of Distance Vector Routing (DVR) algorithm, where in the
“distance” metric is always the number of hops.
• The maximum hop count allowed in the RIP is 15.
• It is mostly used in small to medium-sized networks.
• Routing Tables are broadcast to all the neighbouring routers every 30
seconds.
• That information is used by every router to update it’s own routing table.

Computer Networks - Prof Ashok Herur 82

 Border Gateway Protocol (BGP)
• BGP (Border Gateway Protocol) is the protocol that enables the global routing
system of the internet.
• It manages how packets get routed from network to network by exchanging
routing and reachability information among “edge routers”.
• An edge router is a specialized router located at a network boundary that
enables an internal network to connect with another network.
• BGP enables peering to send packets between autonomous systems (ASes),
which are networks managed by a single enterprise or service provider.
Together, these ASes make up the public internet.

Computer Networks - Prof Ashok Herur 83

 Border Gateway Protocol (BGP)
• BGP is an exterior gateway protocol, which means it is designed to share
routing information between different ASes.
• Alternatively, an interior gateway protocol sends information within a
single AS.
• BGP helps provide redundancy by enabling routers to quickly adapt and send
packets through another connection if one internet path goes down.
• It is often used in large networks, such as internet service provider networks,
wide area networks and infrastructure-as-a-service (IaaS) environments.

Computer Networks - Prof Ashok Herur 84

 Multicast routing
• Multicast is a method of group communication where the sender sends data
to multiple receivers or nodes present in the network simultaneously.
• Multicasting uses spanning trees to deliver the packets to members of a
particular multicast group.
• Link State routing helps construct these spanning trees since each router
knows the entire topology, and which router belongs to which multicast
group.

Computer Networks - Prof Ashok Herur 85

Multicast routing - Example

Computer Networks - Prof Ashok Herur 86

 Multicast OSPF (MOSPF) routing
• It is an Interior Gateway Protocol (IGP) specifically designed to distribute
unicast topology information among routers belonging to a single
Autonomous System.
• OSPF is based on link-state routing algorithm.
• MOSPF routers are required to implement a "local group database" which
maintains a list of directly attached group members.
• One of the routers is designated as a Designated Router (DR) and another
one as a Backup Designated Router (BDR).
• The DR is responsible for communicating group membership information to
all other routers in the OSPF area by flooding Group-Membership
information.
Computer Networks - Prof Ashok Herur 87
Computer Networks - Prof Ashok Herur 88