Network Layer

UNIT 1 INTRODUCTION TO LAYER

FUNCTIONALITY AND DESIGN ISSUES

1.0 Introduction
1.1 Objectives
1.2 Services of the Network Layer
1.3 Packet Switching
1.3.1 Virtual Circuit Approach (Connection-oriented Service)
1.3.2 Datagram Approach (Connection-less Service)
1.3.3 Comparison of Virtual Circuit and Datagram Approach
1.3.4 A view of some Network Service models
1.4 Network Addressing
1.4.1 IP Address
1.4.2 Hierarchy in Addressing
1.4.3 Getting an IP Address
1.5 Congestion
1.6 Routing
1.6.1 Classification of Routing Algorithms
1.7 Delay in Packet Switched Networks
1.7.1Types of delay
1.7.2Computation of delay
1.7.3Numerical
1.8 Summary
1.9 Solutions to the problems
1.10 Further Readings

1.0 INTRODUCTION

This chapter discusses about the network layer, which is the third layer of the OSI
model. Job of this layer is to send the packets from a source to destination. This layer
responds to the service requests of the transport layer and takes the services from the
data link layer. This chapter starts with an overview of the services of the network
layer. Switching is the backbone of network architecture. This important concept of
packet switching is elaborated with diagrams in section 1.3. How the address will be
assigned to a host and the different concepts of addressing are discussed in further
section. This is followed by congestion and routing concepts. Delay is an important
concept in packet switched networks. The various types of delay have been discussed.
The delay computation in different scenarios is illustrated with various examples in
further section.

1.1 OBJECTIVES
After completing this unit, a student will be able to

 Explain the basic concepts and services of network layer.

 define the concepts of packet switching
 Differentiatebetween virtual circuit and datagram approach.

1
Introduction to Layer
Functionality and Design  Elaborate and utilize the concepts of addressing
Issues  define congestion and policies to overcome the congestion in the
network layer
 explain the concept of routing;
 calculate the delay in a given network scenario

1.2 SERVICES OF THE NETWORK LAYER

The third layer of the protocol stack is the network Layer. This layer is responsible for
delivering the data from source machine to the destination machine that is end to end
communication. At the source machine, it takes the services from the transport layer.
Whereas on the destination side, the network layer provides the services to the upper
layer that is transport layer. The important services provided by the network layer are

a) Routing – As discussed above, this layer is responsible for machine to

machine communication. Thus, this layer decides the route that a packet has
to follow from source to destination. There could be various possible paths
from a source to a destination. Based on the chosen metric like delay, number
of hops, a particular path would be selected as the best route.

b) Packetization–At the source machine, network layer receives the segment

from the transport layer and send further to the data link layer. The received
segment needs to be divided further into small packets or send as a whole
packet, this decision is to be taken by the network layer by visualizing the
maximum transmission unit of the data link layer. Control information i.e.
header is to be added at the sending side so at the receiving side, packet is to
be reassembled or decapsulated correctly.

c) Forwarding – when a packet is send from a source to a destination, this

packet pass through a number of routers along the path. A router has a
number of interfaces. Which interface is to be selected for the packet is
decided by the network layer.

Let us understand the clear distinction between routing and forwarding with an
example. We are planning a drive from JIIT, Noida to IGNOU. There are various
possible paths like one is via GT road, another is via Indirapuram and so- on.
Which path is the best one as per the time taken or road conditions? This decision
process is routing and here our metric to decide the best route could be any one
like traffic conditions on the road, infrastructure of the road, etc. Suppose the
selected route is via Indirapuram, and the person started the journey. At one of the
intermediate junctions, there are various directions. Which direction to be chosen
at the junction is the forwarding decision taken by the router.

1.3 PACKET SWITCHING

There are two switching mechanisms that work in the backbone of the network, circuit
switching and packet switching. In today’s Internet, packet switching is utilized where
2
 Network Layer
telephone networks is an example which best describes the concept of circuit
switching. In circuit switching, the resources are reserved for a user where as in
packet switching, the resources are shared among different users on demand basis.
Network layer utilizes the packet switching as packet is the basic data unit used at this
layer.

Let us understand the concept of packet switching more clearly with the following
scenario. Consider two banks where bank 1 requirement is book an appointment
before coming to the bank. If you reach directly, you would not be entertained.If
already booked an appointment, your waiting time is negligible.There is no such
requirement for 2ndbank. As soon as you reach to the bank, you will be entertained
based on the number of people already waiting. The services will be provided to you
without any hassle, if no one is there. If already a large number of people are waiting,
then your waiting time would be large or in some situations, bank will say it’s already
full, kindly come on next day. But on the other hand, there is no hassle of calling
before leaving from home. The scenario of 2ndbank describes how packets will be
handled during packet switching.

Network layer receives the data from the transport layer and divide into manageable
units known as packets. Based on different forwarding mechanisms used by connected
devices to forward the packets from a given source to a particular destination, packet
switched networksare further divided into two categories: virtual circuit approach and
datagram approach.

1.3.1 Virtual Circuit Approach (Connection-oriented Service)

Before going into the detail of virtual circuit approach, first let us understand the
meaning of connection oriented service. Connection oriented service in which an
end to end logical connection would be established between the source machine and
destination machine. All the data between a source destination pair would be sent
through the same connection. After sending the data, connection would be terminated.
A connection oriented service has the following properties

a) All data would be sent in order and without any error to the destination
machine.
b) All the received data would be acknowledged by the destination machine.
c) The underlying service guarantees the in-order delivery of packets without
any loss or duplication of packets.
d) There is a retransmission policy which will handle the lost packets.

Due to all these properties, connection oriented service is also known as reliable
service. A connection oriented service is a three step process which are described as
follows

a) Connection establishment: This is a handshaking process which needs to be

executed before any data exchange among two entities. Suppose, person A
wants to talk to other unknown person B. Before any informal or formal talk,
they will exchange formal hello messages. Similarly, here in connection
oriented service, source machine will send the connection request message
(control message) to the intended destination machine. In receipt of this,
3
Introduction to Layer
Functionality and Design
destination machine will send an acknowledgement message to the source
Issues machine. Source machine will send a confirmation of the received
acknowledgment message. Purpose of this 3 step control messages exchange
is to prepare both the entities for handling the data transfer further.

b) Data transfer:Once the connection gets establishment, data could be

transferred among the two entities.

c) Connection termination: Once the data transfer is over, a connection

termination request will be send by the source machine to the destination
machine. Connection termination is also a three step process similar to the
connection establishment phase. Source machine will send the connection
termination request to the destination machine. Destination machine will send
the ACK of termination request and its own termination request. In the last
step, sender sends the confirmation of received acknowledgement and
termination packet.

Transmission Control Protocol (TCP) is an example of connection oriented

protocol which works at the transport layer. Connection oriented service is provided at
the transport layer as well as the network layer. However, there are some subtle
differences. At the transport layer, only two end systems are involved in connection
establishment and setting of parameters, where as in the network layer, along with end
systems, connecting devices i.e. routers along the path are also involved in setup
process. Connection oriented service at the transport layer is implemented in the two
end systems where as in the network layer it is implemented in all the routers in the
chosen path along with the end systems.

In virtual circuit approach, a virtual connection would be established from source to

destination on which all packets among this source destination pair would be sent.
Therefore, it is also known as connection oriented service.

A network layer packetcontains the source and destination address as a part of header
information because it provides logical communication among the machines.In virtual
circuit approach, along with the source and destination addresses, packet contains a
VC-ID. It is necessary to mention VC-ID in the packet as all packets has to follow the
same virtual connection. When packet reaches to a router it will consult the
forwarding table on the basis of VC-ID mentioned in the packet and decide the output
port.

A virtual circuit approach involves three phases. The phases are a) setup phase b) data
transfer and then connection termination. All are explained as follows:

a) Setup phase: Establishing a virtual circuit implies the following needs to be

done.
i. Deciding the path between a source and destination
ii. Assigning a virtual circuit identifier (VC-ID) to each link along the
path
iii. Change in the forwarding table of all intermediate routers along the
path with respect to virtual circuit

Let us understand this process with the figure 1.

4
 Network Layer

4 1
R1 R3
A
1 2
3
2 3

3 2
1 2
4 1
R2 3 R4 B

Figure 1: Sending of Request packet in virtual connection establishment process

As shown in figure 1, machine A wants to send the data to machine B. Let us chosen
path between machine A and B is A-R1-R2-R4-B. Thus, a virtual connection needs to
be established among A and B by involving all the intermediate routers. The process
is as follows

1) Machine A will chose a VC-ID from its available list of VC-ID’s and send the
request packet to R1. As shown in figure 1, chosen VC-ID by A is 5.
2) As soon as Router R1 will receive this request packet, it will create an entry
for this virtual circuit in its forwarding table as shown in figure 1. In this
entry, Router R1 notes that the packet is coming from incoming port 1 and
incoming VC-ID 5. Outgoing port is 2 and leave blank in place of outgoing
VC-ID.
3) Now, R1 will forward this request packet to R2. In the similar manner, R2
will create an entry of this virtual circuit request in its forwarding table.
Suppose, the chosen VC-ID by R1 is 15, thus the values of incoming port,
incoming VC-ID, outgoing port and outgoing VC-ID are 1, 15, 3 and blank
respectively.
4) R2 will forward the packet to R4. R4 complete the three entries of its
forwarding table in the similar manner as shown in figure 1.
5) R4 sends the packet further to machine B. Machine B will chose a VC-ID and
let this value is 60. In future communications, the VC-ID 60 is an indication
for B that this packet comes from machine A.

All these five steps show the forwarding of request packet for setting the
virtual connection from source machine A to destination machine B. But this
forwarding completes the only three entries in the forwarding table. To
complete the 4th entry of forwarding table, B will send an acknowledgment
packet back to A via same path that is B-R4-R2-R1-A. The process can be
visualized in figure 2 and explained as follows.
5
Introduction to Layer
Functionality and Design
Issues 1) Destination machine B sends an acknowledgement packet carrying VC-
ID 60 to R4. By knowing this value, Router R4 will complete the 4th
column i.e. outgoing VC-ID of its forwarding table as shown in figure 2.
2) Router R4 will forwardthis acknowledgement packet to router R2. This
packet contains the incoming VC-ID 35 which will be copied at the place
of outgoing VC-ID in the table of R2.
3) Similar process will happen at R1. Router R1 receives the incoming VC-
ID 15 from the R2 table. It will be copied at the place of outgoing VC-ID
in the table of R1.
4) Finally, R1 forwards the acknowledgement packet to machine A which
carries incoming VC-ID as 5. This VC-ID is chosen by A only in the
initial process. Machine A knows that this VC-ID is to be used for
communication to B.

4 1
5 R1 R3
A
1 2
3
2 3

15

3 2
1 2
4 1 60
3 35 R4
R2 B

Figure 2: Sending of Acknowledgment packet in virtual connection

establishment process

As discussed initially in the setup phase the virtual circuit establishment implies three
works (deciding the path, assignment of VC-ID to each link, change in forwarding
table) to be done. So, as explained in figure 1 and 2, all three mentioned works has
been completed.

b) Data transfer:All the packets between A and B will be sent through the same
established virtual circuit between them. As a result, thus all reach to the
destination in order. Each intermediate router changes the value of VC-ID
by seeing the forwarding table as shown in figure 3. As soon as the packet is
reached to a router, it will see the VC_ID of this packet. In this example it is
5. Thus, R1 will see its forwarding table for the VC_ID 5 and incoming port

6
 Network Layer
1. It can be visualized from figure 3, for these values as an index; the
outgoing port is 2 and VC_ID is 15. R1 will change the VC_ID value in the
packet and forward it further. Similar process will be followed at the other
routers as well and finally the packet will be delivered to the destination
machine B via established virtual connection. The figure3 shows the process
for one packet. The same process would be followed by all the packets.

5, PKT 4 1
R1 R3
A
1 2
3
2 3

15, PKT

3 2
1 2
4 1
3 35,PKT R4
R2 B
60,PKT

Figure 3: Packet transfer over an established virtual connection

c) Connection termination: Once A has sent all the packets to B, machine A will
send a termination request packet to B and in return, B will send an
acknowledgment of the same. As a result, all the routers delete the entry from the
routing table.

1.3.2 Datagram Approach (Connection-less Service)

Datagram approach is used in today’s Internet scenario. This approach follows the
concepts of connection-less service. So, let us first understand the basic concepts of
connection less service.

Connection less service, as its name implies no virtual connection would be made
between source and destination. Both the entities does not do any handshaking. When
a machine wants to send the packet to another machine, it simply starts sending. The
message is divided into manageable units called packets where each packet is treated
individually. Each packet can follow same or different paths, thus they may reach out
of order or can be lost in between. Sender machine does not have any clue regarding
the loss of packets as there is no provision of acknowledgement of packets. Due to all
these properties, connection less service is an unreliable service.

7
Introduction to Layer
Functionality and Design
Although this service is unreliable but it is required in some situations like where we
Issues want immediate transfer of data or where loss of some packets does not affect the
overall quality of message or the situation where less overhead is required. The
overhead of handshaking or sending acknowledgements is not present in connection
less service.User Datagram protocol (UDP) and Internet protocol (IP) are the
examples of connection less protocols which works at the transport layer and network
layer respectively.

As datagram approach is a connection less service, thus all packets either belonging
to same source destination pair or different, are treated individually. Here, packet is
called as a datagram. Datagram contains the source and destination address.
Forwarding decision is taken individually for each packet on the basis of destination
address. Each router looks into the forwarding table for the mentioned destination
address in the datagram. It returns the output interface based on matching on which
the datagram will be forwarded further. If more than one entry are matched then based
on the principle of longest prefix matching, the output interface would be selected.

In datagram approach, routing tables were modified by the routing algorithms.

Routing algorithms is an important aspect of network layer which will be discussed
later.

Destination address follows hierarchical addressing structure. How the destination

address is actually extracted and processed to decide the path, let us understand it
more clearly with an analogy. Suppose we want to go to a particular location 116-A,
H Block, Vikaspuri, New Delhi. We started our journey from Noida and at the first
junction if we ask for 116-A, H Block, Vikaspuri, New Delhi. Nobody will be able to
tell the exact location or if somebody tried to do so, he or she will tell you only the
road going towards Delhi. After reaching to Delhi, if u ask now for Vikaspuri,
someone can instruct you by taking this path, you can reach to the desired location.
Again the same situation, when you enter into Vikaspuri and you ask for 116-A, H
Block , the person can tell you only about the path directions to reach H-block, not
exactly 116-A. In the similar manner, part of destination address will be extracted and
used to decide the output interface.

1.3.3 Comparison of Virtual Circuit and Datagram Approach

Both the approaches havetheir own advantages and disadvantages. Both the
approaches can be compared on the basis of following points.

a) Setup time: Connection is to be setup in case of virtual circuit approach

where as in datagram approach, no setup phase is required. Due to setup phase
in virtual circuit approach, the sequencing of packets can be easily
maintained. On the other hand, in datagram approach there is no setup
overhead so sending of packets can be started immediately.

b) Routing decision:As the virtual circuit has been established between a

specified source and destination, so no routing decision has to be taken for
individual packets. Thus, packets can be forwarded more quickly in virtual
circuit approach whereas in datagram approach, for every packet, the routing
8
 Network Layer
decision is to be made. In virtual circuit approach, output port is decided by
looking into the VC_ID of a packet where as in datagram approach, output
port is decided by looking into the destination address.

c) Reliability:In virtual circuit approach, if a router gets failed, all the

connection passing through that router or whose state information is
maintained in this routergets lost. However, in datagram approach only the
packets waiting in the queue of that router gets lost.

d) Routing tables: While establishing a virtual circuit, state of a connection

needs to be updated in all the intermediate routers. Routing table is indexed
by VC_ID where as in datagram approach, routing table is indexed by
destination address and routing algorithms update the routing tables.

e) Load balancing: In datagram approach, packets may travel different paths.

Thus traffic can be balanced over multiple routes.

f) Reservation of resources: In virtual circuit approach, resources are reserved

so delivery can be guaranteed. If there are packets then allocated resources
(like buffers, bandwidth, etc.) would be directly used. Whereas in datagram
approach, the resources are shared on the demand basis. If everyone is trying
to use the resources and resources are limited, then it may lead to congestion
or packet loss. Congestion can be easily avoided in virtual circuit approach.

Table 1 provides a brief overview about the difference in virtual circuit and
datagram approach.

Table 1: Comparison of virtual circuit and datagram approach

Virtual Circuit Approach Datagram approach

Route is decided for all packets of a Route is decided for each packet
conversation between S and D
Overload may block connection setup Overload increase packet delay
and increase packet delay
Connection set up delay along with Only packet transmission delay
packet transmission delay
Forwarding decision based on VC_ID Forwarding decision based on
destination address
Congestion avoidance is easy Congestion avoidance is difficult

1.3.4 A view of some Network Service models

Till now, we had discussed the overview of network layer services. This section
discusses some of the network architectures to get an idea about their services.

Internet is most widely used network architecture. Internet’s network layer provides
best effort service. Best effort service implies it will try but does not guarantee
anything. Therefore, it can be visualized from table 2, Internet network service model
does not guarantee on any issue like ordering of packets, packet loss, bandwidth etc. It
does not even preserve the timings difference among packets when the packets reach
9
Introduction to Layer
Functionality and Design
at the receiver side. But, there are other network architectures which provide more
Issues than best effort service. Table 2 compares three network service architectures on the
basis of their services. For more details please refer [1]

Table 2: Comparison of Network Service models

Internet ATM ATM

Service model Best effort CBR ABR
Guaranteed No Constant rate Minimum
Bandwidth Guarantee Guarantee
Delay Guarantee No Yes No
Sequencing of Any order In order In order
packets
Packet Loss No Yes No
Guarantee
Congestion No No chances of Provides
indication congestion congestion
indication

ATM CBR (constant bit rate) works on the principle of a virtual pipe between
source and destination. Thus, it is able to provide some of the services like
ordering of packets, guaranteed bandwidth to each user, etc. No packet would
be lost and there are no chances of congestion as the resources are reserved
while establishing the connection. ATM ABR (available bit rate) provides a
minimum amount of bandwidth guarantee and delivers packets in order. But it
does not provide any guarantee about the loss of packets and jitter among
packets. Thus, ATM ABR is a little bit better than best effort service model of
Internet.

 Check Your Progress 1

Choose the correct option.

Q1. ______________ Approach takes the forwarding decision based on destination
address.
a) Virtual circuit
b) Datagram

Q2. ________ is a connection less protocol used at the transport layer.

a) IP
b) TCP
c) UDP

Q3. ________ is an example of packet switched datagram networks.

a) Internet
b) Telephone networks

Q4. Compare Virtual circuit approach with the datagram approach. Provide at least
two differences.

...................................................................................................

10
 .................................................................................................... Network Layer

....................................................................................................

1.4 NETWORK ADDRESSING

Network layer provides end to end communication i.e. it delivers the packets from
source machine to the destination machine. This communication could be at a global
level, thus, a unique identifier for every machine is required. This identifier is the
logical address of machine, also known as Internet address or IP address. In actual
terms, this address is not associated to the machine; it is associated to the interface.
The portion between machine and the link is called as an interface. Generally, a host is
connected to a single network through a link so it has one interface thus one IP
address. On the other hand, a router is connected to many networks or hosts so it has
multiple interfaces and each interface will have an IP address.

1.4.1 IP address

IP address is a 32 bit address. Along with the property of uniqueness, IP address

should be universally acceptable also, that is who so ever wants to communicate,
follow a common format. As the IP address is of 32 bits, thus 232 unique addresses are
possible. This much amount of address space impliesat an instant of time,
approximately 4 billion machines with unique addresses could be connected.

IP address is generally written in dotted decimal notation (base 256). The other two
notations are binary (base 2) and hexadecimal (base 16). As shown in figure 4, binary
notation is just writing of all 32 bits in binary form. However, to increase its
readability, the bits are written in a group of 8 bits that is a byte and some space will
be provided between each byte. If we write decimal value of each byte and put a dot
to separate the group is referred as dotted decimal notation. This is most commonly
used notation. If we write hexadecimal value with respect to a group of 4 bits, then
that notation is called as hexadecimal notation.

11000000 00100000 00000011 00001101 Binary

192.31.3.13 Dotted decimal

C020030D Hexadecimal

11
Introduction to Layer
Figure 4: IP Address notations
Functionality and Design
Issues
1.4.2 Hierarchy in addressing

Network addressing follows hierarchical structure. The hierarchical structure can be

easily visualized in our daily life examples like postal service, telephone networks. In
postal service, posts have been distributed on the basis of written postal address.
Postal address is extracted on the basis of country, state, district, city, street, building
and house number. The address is always extracted in the reverse order. First, all the
posts have been separated on the basis of country then state will be looked upon and
so on. Similarly, the telephone networks also follow the hierarchy in telephone
number. First few digits signify the country code which is followed by the area code
and then the connection number itself.

In the similar manner, IP address is divided into two parts where first part signifies the
network portion and second part is the host address. Network portion can be fixed or
variable. If the network portion is fixed, it refers to classful addressing which is
widely used in earlier days. But nowadays people switch onto the concepts of variable
network portion which refers to classless addressing. The next chapter discusses the
concept of classful and classless addressing in complete detail. Suppose 𝑏 bits are
used to denote network address, then the remaining (32 − 𝑏) bits would be used to
denote host address.

32 bits

IP Address

Network portion Host portion

b bits 32-b bits

Figure 5: Parts of IP address

1.4.3 Getting an IP address

An IP address can be assigned to a host in two ways either manually or dynamically.

a) Manual Assignment: Network administrator allocates an IP address to

the host from the available block of addresses. It would not be changed
until the administrator himself changes it. This is also called static
assignment.
b) Dynamic Assignment: When a host machine joins the network, an IP
address would be automatically assigned by some protocol like Dynamic
Host Configuration Protocol (DHCP). DHCP works like a plug and play
protocol in the sense that as soon as someone joins the network, an
address would be allocated and free the address when the host machine
leaves the network. This is a dynamic assignment in the sense that every
time a host joins a network, it will get a new address.

12
 Network Layer

 Check Your Progress 2

Choose the correct option.

Q1. The ________ protocol is used to assign dynamic IP address.

a) Internet protocol
b) Transmission Control Protocol
c) Dynamic Host Configuration Protocol

Q2. If a host portion is of 8 bits then how many bits denote the network portion?

a) 32 bits
b) 24 bits
c) 16 bits

Q3. IP address is generally written in __________ notation.

a) Dotted decimal
b) Binary
c) Hexadecimal

Q4. How a host gets an IP address?

...................................................................................................

....................................................................................................

....................................................................................................

1.5 CONGESTION

If the packets are coming at a faster rate than the handling capacity of the network,
this leads to a situation known as congestion. Initially, when the packet arrival rate
starts getting higher than the packet processing rate, queue starts filling up. As a
result, packet delivery time gets increased. If the same situation continues, queue
becomes full and packet drop starts. In this situation, source does not receive any
acknowledgement and for a large number of the packets, the timer is up; which leads
to unnecessary retransmissions. Sometimes the situation becomes worse and reaches
to a deadlock point and whole system gets collapsed.

To understand the situation of congestion, let us see the behaviour of two important
performance metrics i.e. delay and throughput with respect to the capacity of the
network. Figure 6 shows the delay and throughput as a function of load.

13
Introduction to Layer
Functionality and Design
Issues

No
Congest
congestion
Congested Area

Throughput
Area
No

Delay
congestion

System Load System Load

Figure 6: Delay and Throughput vs Load

Initially, when the less number of packets, they would be delivered without any delay
and throughput is good. As the system load increases, packets experience queuing
delay which in turn affects throughput as well. When the system load reaches the
capacity, queue becomes full and some packets get discarded. As a result, delay
approaches towards infinite value and throughput starts decreasing.

The issue of congestion is not only handled at the network layer; it is also handledat
the transport layer. The main idea behind congestion is try to avoid the situation i.e.
take preventive measures before reaching to a threshold and if the situation happens,
try to come out of that situation. First one is known as congestion avoidance phase
and second one is congestion removal. The policies used for congestion avoidance
phase is called as open loop congestion control policies. Closed loop congestion
control policies are for congestion removal phase.

Some of the open loop congestion control policies are as follows

a) Admission policy: A router can visualize the possibility of congestion and if

there are chances of congestion then the new virtual connection request can be
rejected.

b) Retransmission policy: When the sender does not receive any

acknowledgement of the sent packet, it does the retransmission. For how long
the sender has to wait or after how many lost packets, the packet needs to be
retransmitted, all these kinds of retransmission policies should be designed in
such a way that it will not add more congestion in the network.

c) Acknowledgement policy: Receiver’s acknowledgement policy can also

control the congestion at some level. For example, if receiver sends the
acknowledgement packet after receiving some packets, it will slow the sender
as well as not add a burden of sending acknowledgement packets.

d) Discard policy: If a router implements the good discarding policy then it can
also prevent congestion at some level. The good discarding policy which does

14
 Network Layer
not impact the overall quality of transmission. For example, in a multimedia
transmission, some less priority packets gets discarded at the time of chances
of congestion then it will not impact the overall quality of transmission.

Sometimes, even after taking all the preventive measures, congestion

occurred. In this situation, closed loop congestion control policies to be used
to avoid stucking into deadlock. These policies are

a) Sending of Choke packet: A choke packet is a control packet sent by the

router to the source node. This packet informs the sender about
congestion occurrence. This method is implemented by protocols like
ICMP, etc.

b) Signaling: In this method a signal would be sent by the congested node to

inform the sender about congestion. Rather than sending an explicit
packet like choke packet, here, a signal will be sent in the existing packets
carrying data.

c) Implicit signaling: In this method, no specific information is sent to the

source rather sender itself guesses about congestion. For example, if the
sender does not receive any acknowledgements of several packets with in
timeout period is a signal for the sender that there is congestion in the
network.

1.6 ROUTING

Job of the network layer is to send the datagram from a source end system to
destination end system. Data may travel through different paths or through multiple
hops to reach to the destination. The process of deciding about the path to reach to the
destination is known as routing. There are routing protocols or which helps in
constructing the forwarding table. The forwarding table or routing table is stored at
every end system and router. Whenever a router receives a packet, router consults its
routing table to decide the output interface. Looking into the routing table and
choosing the output interface, this process is known as forwarding. Filling up of
routing tables, their maintenance and regular updation is done at continuous intervals
by routing protocols or routing algorithms.Routing algorithms is a part of network
layer software.

There are various desirable properties ofa routing algorithm. These are as follows

1) Routing protocols decide the best route from a source S to destination D.

This best route can be best in terms of any of the metrics like delay,
throughput, packet loss, etc. This is similar to the analogy when we decide
our travelling route from one city to another. There are various paths as well
as various modes of transportation. We chose the one on the basis of cost,
15
Introduction to Layer
Functionality and Design
comfort, time etc. Similarly, here in communication networks, the path is
Issues decided by looking into various metrics.
2) Other than the metrics, paths should be decided or updated in between by
looking into the conditions of congestion into the network.
3) Routing is successful only when all the nodes are cooperative with each
other. So, cooperation among the nodes is must.
4) Suppose, a link gets down or a router become fail, then all the routing tables
to be quickly updated. This knowledge should be reflected in all the routing
tables, so that packet loss will be minimal. Quick convergence and stability
is very important in a routing algorithm.

To solve the routing problems, network isrepresented in terms of graphs. While

formulations as a graph, routers are represented as nodes and the links connecting the
routers are represented as edges. A graph G is represented as (V, E). Where, V is a set
of nodes and E is a set of edges connecting those nodes. Each edge is labeled with a
value representing its cost. This cost could be directly or indirectly related to the link
type or metric value. For example, cost could be directly proportional to the
congestion or inversely proportional to the bandwidth. Higher bandwidth link or less
congested link gives a low cost value of that link. If the nodes are connected by an
edge then the cost is associated with that edge else it is infinity. Undirected graphs are
considered for formulation. Thus, the cost associated with edge (a, b) is same as edge
(b, a). A path is a sequence of edges traversed from chosen starting node to chosen
destination node. The cost of a path is sum of all the edges cost of that path. Let us
understand this formation more clearly with the figure 7.

A
2

3 C
7
2
B 5
E
4
4
D

Figure 7: Graph representation of a network

It can be visualized from the figure 7, set V consist of routers {A, B, C, D,

E} and set E consist of edges such as (A,B), (A,C), (D, E), etc. Each edge is
labeled with a value known as cost. Let us calculate the path from node A to
node E. It can be visualized from the figure there are multiple paths from A
to E. Some of the paths and associated costs are as follows

Path Cost
A-C-E 2+7=9
A-B-C-E 3+2+7=12
16
 Network Layer
A-B-D-E 3+4+4 = 11

The cost of a path is sum of the costs of all traversed edges from A to E. On
the basis of low cost, routing algorithm chose the path A-C-E. Suppose, all
the edges have same unit cost, then the path with less number of hops is
chosen. For this scenario, again path A-C-E would be selected because it
has less number of hops in comparison to other paths.

Now, revise all this scenario of finding the paths in your mind and tell me,
how you have chosen the path. Have you tried all the combinations? I think
no, you have tried just 2-3 paths and convinced yourself that this is the least
cost path. All this work of your mind is done actually by a routing algorithm
but to do this job, routing algorithm shouldknow the complete knowledge
about the network.

1.6.1 Classification of Routing Algorithms

Routing algorithms can be classified into different types like centralized or

decentralized based on how the routes would be computed. Another categorization is
adaptive or non-adaptive routing algorithms which are based on strategy of updation
of routes. All these different types are discussed as follows

a) Global Routing Algorithm: These routing algorithms compute the best path
(least cost one or shortest one) by gathering the complete knowledge about
the network. It is also known as centralized routing algorithm. It can be run at
one central location or a replica to be run at multiple locations. All the
information about the nodes and links to be collected at the central algorithm
and then this algorithm computes the routes. The computed routing tables
would be distributed to all nodes. In this way, global optimal routes will be
computed and distributed to all. It will reduce the burden on each node. Link
state algorithms are a kind of global routing algorithms.

b) Decentralized Routing Algorithm: These routingalgorithms works in an

iterative, asynchronous and distributive manner. Complete network
knowledge is not available at a single location rather each node interacts with
its neighbors about the cost and their knowledge. Now, these neighbors
interact further with their neighbors and this process goes on. In this way, the
path to one or more destinations would be computed. Distance vector
routing algorithm is an example of decentralized routing algorithm.

c) Adaptive Routing Algorithm:These routing algorithms reflect the change in

the routing tables whenever there is a change in topology. That’s why these
algorithms are also known as Dynamic routing algorithms. These
algorithms also update the paths with respect to network traffic conditions.

d) Non-adaptive Routing Algorithm: These routing algorithms do not update

the routing table periodically or at the instance of change in topology.
Manually, the routes would be computed and stored. These routing algorithms

17
Introduction to Layer
Functionality and Design
do not respond to failures automatically, that’s why these algorithms are also
Issues known as Static algorithms.

 Check Your Progress 3

Choose the correct option.

Q1.________ process creates and maintains the routing table.

Q2. Which routing algorithm works in an iterative, asynchronous and distributive

manner?
a) Link state routing algorithm
b) Distance Vector routing algorithm
c) Static algorithm

Q3. Define congestion.

...................................................................................................

....................................................................................................

Q4. What are the various policies that can be used to avoid congestion?

...................................................................................................

....................................................................................................

....................................................................................................

1.7 Delay in Packet Switched Networks

The packet transmission process starts from a source and ends at a desired destination.
In this transmission process, packet travels through a number of intermediate routers
and paths. Thus, a packet will not reach immediately to the destination. Rather it
experiences a number of delays.

1.7.1 Types of delay

A packet experience four types of delays which are explained as follows:

a) Transmissiondelay: A source machine transmits a packet means that source

machine put one by one bit of that packet on the link. A packet has certain
length thus, it can’t be put on the link in one go. Total time experienced by the
source machine in this process is known as transmission delay. If a packet
length is denoted as L and transmission rate is denoted by R, then
transmission delay is calculated as𝐿⁄𝑅 .

b) Propagation delay:As soon asthe bit is put on the link, this bit has to travel
through a number of intermediate links. For a single intermediate link,
propagation delay is calculated as distance of this link divided by the speed of

18
 Network Layer
the link. Speed of the link depends on the physical type of link. Generally,
speed is considered as 3*10 8 m/s, which is propagation speed of the vacuum.

c) Processing delay: This is the amount of time taken by a router or destination

machine to process a packet. Packet content would be checked for error
detection. Packet would be processed to the upper layer protocol if it is a
destination machine. If it is a router, it would be processed to the selected
outgoing port. Generally, the value of processing delay depends on the speed
of the router.

d) Queuing delay:As its name suggests, this is the amount of time a packet
waits for its turn to get to be transmitted. Each router has an input queue for
incoming port and an output queue for outgoing port. Summation of both the
waiting times is known as queuing delay. Queuing delay mainly depends on
the packets already waiting for their turn. If there is no packet in the queue,
queuing delay is zero.

1.7.2 Computation of delay

Total delay is the summation of all the above types of delays defined in above
subsection. The following notations represent four delays.

 𝑑𝑡 − 𝑇𝑟𝑎𝑛𝑠𝑚𝑖𝑠𝑠𝑖𝑜𝑛 𝑑𝑒𝑙𝑎𝑦
 𝑑𝑝 − 𝑃𝑟𝑜𝑝𝑎𝑔𝑎𝑡𝑖𝑜𝑛 𝑑𝑒𝑙𝑎𝑦
 𝑑𝑝𝑟𝑜𝑐 − 𝑃𝑟𝑜𝑐𝑒𝑠𝑠𝑖𝑛𝑔 𝑑𝑒𝑙𝑎𝑦
 𝑑𝑞 − 𝑄𝑢𝑒𝑢𝑖𝑛𝑔 𝑑𝑒𝑙𝑎𝑦

To compute the total delay experienced by a packet from source to

destination, it has to been how many links and routers in between. If there are
𝑘 links in the path from source to destination, it implies there are 𝑘 − 1
routers in between. For all the 𝑘 links, transmission, propagation and
processing delays to be computed but queuing delay is to be computed at
router only. Total delay is computed as follows

𝑇𝑜𝑡𝑎𝑙 𝑑𝑒𝑙𝑎𝑦 = 𝑘(𝑑 𝑇 + 𝑑𝑃 + 𝑑𝑝𝑟𝑜𝑐 ) + (𝑘 − 1)𝑑𝑞

1.7.3 Numerical

Q1. Suppose two hosts Y and Z are directly connected by a link. Length of this link is
10,000 Km and this link transmission rate is 1Mbps. The propagation speed of the
link is 2.5*108 m/s. Based on this information answer the following parts

a) Y sends a file of 400K bits to Z. How long does it take to send the file assuming
it is sent continuously?
b) Suppose now the file is broken up into 10 packets with each packet containing
40K bits. Z sends an ACK for each packet and Y cannot send a packet until the
preceding one is acknowledged. Transmission time of an ACK packet is
negligible. How long does it take to send the file?

19
Introduction to Layer
Functionality and Design
Answer:
Issues
a) File size = 400,000 bits , Transmission rate = 1 Mbps

Thus, transmission delay is

𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒 400000
= 𝑡𝑟𝑎𝑛𝑠 𝑟𝑎𝑡𝑒 = 106
= 400 𝑚𝑠𝑒𝑐

Distance = 10,000 Km, Speed = 2.5 *108m/s

Thus, propagation delay is

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 10 ∗ 106
= = = 40 𝑚𝑠𝑒𝑐
𝑠𝑝𝑒𝑒𝑑 2.5∗ 108

In this scenario, there is no processing delay or queuing delay. Therefore, total delay
is

Total delay = Trans delay + Propagation delay = 400 + 40 = 440 msec.

b) Packet size = 40,000 bits , Transmission rate = 1 Mbps

Thus, transmission delay for one packet is

𝑓𝑖𝑙𝑒 𝑠𝑖𝑧𝑒 40000
= 𝑡𝑟𝑎𝑛𝑠 𝑟𝑎𝑡𝑒 = 106
= 40 𝑚𝑠𝑒𝑐

Distance = 10,000 Km, Speed = 2.5 *108m/s

Thus, propagation delay is

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 10 ∗ 106
= 𝑠𝑝𝑒𝑒𝑑
= 2.5∗ 108 = 40 𝑚𝑠𝑒𝑐

There are two important points to be note down

i. Second packet is sent only when Y receives the acknowledgement of first packet,
thus twice of propagation delay is used.
ii. Acknowledgement can be sent only when the first packet is received completely
at the receiver end. Thus, twice of transmission delay is used.

Therefore, for one packet total delay is = 2*Trans delay + 2*Propagation delay

And the total delay for all 10 packets is = 10*(2*Trans. delay + 2*Propagation delay)

= 10*(2*40 + 2*40) =1600 msec.

Q2. Compute the end to end delay for circuit switching and packet switching for a
network. This network is having 5 hops to switch a message of 1200 bits where all the
links have a data rate of 4800bps. Size of the packet is 1024 bits along with a header
of 32 bits. In case of circuit switching, consider 0.5sec as a call setup time. Hop to hop
delay is .02 sec. Assume zero processing delay.

Answer: This answer is divided into two parts a) Computation of delay in circuit
switching scenario b) Computation of delay in packet switching scenario

a) Computation of delay in circuit switching scenario

20
 Network Layer

Call set up time is = 0.5 sec

Propagation delay is = .02 sec
1200
Transmission delay is = = 0.25 sec
4800
Total delay = call set up time + message delivery time
= 0.5 + 5*(propagation delay) + transmission delay
=0.5+ 5*(0.02) + 0.25 = 0.85 sec

b) Computation of delay in packet switching scenario

The given packet size is 1024 bits. It implies 32 bits of header and the leftover 992
bits are data bits. Thus, to send the total message of 1200 bits, two packets are
required.

 First packet is of 1024 bits (992 bits of data and 32 bits of header).
 Second packet is of 240 bits (1200 – 992= 208 bits of data and 32 bits of
header).

Propagation delay is = 5*0.02 = 0.1 sec

1024 240
Transmission delay at first hop = 4800 + 4800 = 0.213 + 0.05 = 0.263 sec
1024
Transmission delay at rest of the hops is = 4 ∗ 4800 = 4 ∗ 0.213 = 0.852 sec. because
at the rest of the hops there is no transmission delay for 2 nd packet or any other
number of packets.

The total delay is 0.1 + 0.263 + 0.852 = 𝟏. 𝟐𝟏𝟓 sec.

 Check Your Progress 4

Choose the correct option.

Q1. If there is no buffer at the router, each incoming packet directly forwarded further
onto the outgoing port. In this situation which kind of delay is negligible?
a) Processing delay
b) Queuing delay
c) Transmission delay
d) Propagation delay

Q2. Host X is connected to Y via switch S. The link bandwidth is 10Mbps and
propagation delay on each link is 20μs. S is a store and forward switch, it begins
retransmitting a received packet 35μs after it has finished receiving it. Calculate the
total time required to transmit 10,000 bits from X to Y.

a) As a single packet
b) As two 5,000 bit packets sent on right after the other

...................................................................................................

....................................................................................................

....................................................................................................
21
Introduction to Layer
Functionality and Design
....................................................................................................
Issues

1.8 SUMMARY

In this unit, we understood the concepts of packet switching. Network layer follows
the concept of packet switching as packet is the basic data unit used at this layer.
There are two types of packet switching techniques, virtual circuit and datagram
approach. In virtual circuit approach, before sending any data between a source
destination pair, end to end logical connection needs to be established between them.
Datagram approach is a connection less service. There is no handshaking between
source and destination and each packet follows its own route.

Each machine is identified by a logical address, known as IP address. It is a 32 bit

address which is unique for an individual. Due to overload of the network, network
layer faces a problem which is known as congestion. Open loop congestion control
policies for avoiding the congestion and Closed loop congestion control policies for
congestion removal phase has been studied. A very important job of network layer is
route the packets, thus concepts of routing has been discussed. It is followed by the
computation of delay metric which is a very important in network layer as packet
travels through a number of paths and routers.

1.9 SOLUTIONS/ANSWERS

Check your progress 1

1) b

2) c

3) a

4) Virtual circuit approach decides the output port on the basis of VC_ID of a
packet whereas datagram approach decides the output port on the basis of
destination address mentioned in the packet. If a router gets failed, only the
packets waiting in the queue of that router gets lost in datagram approach.
However, in virtual circuit approach, all the connection passing through that
router or whose state information is maintained in this router gets lost.

Check your progress 2

1) c

2) b

22
 Network Layer
3) a

4) A host gets an IP address either manually which is assigned by a network

administrator or dynamically assigned by the DHCP protocol. Manual
assignment is also known as static assignment as the allocated address
can’t be changed until the administrator himself wants to change the
same. In manual assignment, network administrator allocates an IP
address to the host from the available block of addresses.
During dynamic assignment, Dynamic Host Configuration Protocol
assigns the IP address to the machine as soon as the host joins the
network and frees the address when the host machine leaves the network.
Therefore, whenever host joins the network, it will get a new IP address.

Check your progress 3

1) Routing

2) b

3) Network Layer faces this problem of congestion when the number of

packets sent to the network is greater than the capacity of the
network. Network is not able to handle the packets as packets are
coming at a faster rate. Then, packet loss starts and sometimes it leads
to a deadlock situation and the whole system gets collapsed.

4) There are policies used to avoid the congestion known as open loop
congestion control policies. But if congestion occurred, then some of
the policies are used to remove the congestion.

The following open loop congestion control policies can be used to avoid
congestion. Receiver’s acknowledgement policy can control the congestion at some
level. For example, if receiver sends the acknowledgement packet after receiving
some packets, it will slow the sender as well as not add a burden of sending
acknowledgement packets. Another policy implemented by router that the router can
visualize the possibility of congestion and if there are chances of congestion then the
new virtual connection request can be rejected. Sender can implement the
retransmission policy to avoid the problem of congestion. For how long the sender has
to wait or after how many lost packets, the packet needs to be retransmitted, all these
kinds of retransmission policies should be designed in such a way that it will not add
more congestion in the network. Sometimes, even after taking all the preventive
measures, congestion occurred. In this situation, closed loop congestion control
policies to be used to avoid stucking into deadlock.

These policies are mainly about informing all about the situation of congestion. One
of the policy is sending a choke packet. A choke packet is a control packet sent by the
router to the source node. This packet informs the sender about congestion
occurrence. Another is signaling in which a signal would be sent by the congested
node to inform the sender about congestion. Rather than sending an explicit packet
like choke packet, here, a signal will be sent in the existing packets carrying data.

23
Introduction to Layer
Functionality and Design Check your progress 4
Issues
1) b

2) Propagation delay =20μs

𝐿 10,000
Transmission delay = = = 1000μs
𝑅 10∗106

a) Total time (as a single packet) =20+1000+35+20+1000=2075μs

𝐿 5,000
b) Transmission delay =𝑅 = 10∗106 = 500μs
Total time for first packet = 20+500+35+20+500=1075μs
Total time for second packet = 500μs

Thus, total delay = 1075+500 = 1575μs

1.10 FURTHER READINGS

[1] Kurose, J. F., & Ross, K. W. (2012). “Computer networking: A top-down

approach featuring the Internet”, Boston: Addison-Wesley.

[2] Forouzan, B.A., & Mosharraf, F. (2012), “Computer Networks: A top-down

approach”, McGraw Hill.

24
 Network Layer
UNIT 2 ROUTING ALGORITHMS

2.0 Introduction
2.1. Objectives
2.2. Flooding
2.3. Shortest Path Routing Algorithm
2.4. Distance Vector Routing
2.4.1. Comparison
2.4.2. The Count-to-Infinity Problem
2.5. Link State Routing
2.6. Hierarchical Routing
2.7. The Internet Protocol (IP)
2.7.1. IPV4 addressing
2.7.2. Datagram Format
2.7.3. IP Datagram Fragmentation
2.7.4. IP V6
2.7.5. Internet control message protocol
2.7.6. Dynamic host configuration protocol
2.7.7. IP Security
2.8. Routing with Internet
2.8.1. Inter Autonomous System Routing in the Internet: RIP & OSPF
2.8.2. Inter Autonomous System Routing BGP
2.9. Multicast Routing
2.10. Mobile IP
2.11. Summary
2.12. Solution/Answers
2.13. Further Readings

2.0 INTRODUCTION

Network layer is responsible for finding the optimal route from source to a
destination. Multiple paths may exist between a pair of source and destination.
A path with minimum cost is considered to be the optimal route. Routing
algorithms construct and maintain a table called Routing Table which is
referred while looking for a route. A routing algorithm is responsible for
selecting the most appropriate route in the network between source and
destination. Router is the network layer device which is responsible for
performing routing for the network. A cost is associated with each path in the
form of bandwidth, delay, congestion, security etc. Router performs routing
and selects the minimum cost path for all the remote networks. A router is
implemented with a number of algorithms to find the optimal routes.Based on
the criteria of optimal path according to the requirement of the network traffic,
appropriate routing algorithm can be chosen.

1
Routing Algorithms

In this unit section 2.3 is about flooding which uses broadcasting. In section
2.4 shortest path routing algorithmi.e. Dijkstra’s algorithm is discussed. In
section 2.5 Distance vector routing algorithm: Bellman-Ford Algorithm is
discussed. In this section comparison between Dijkstra’s algorithm and
Bellman-Ford Algorithm and count-to-infinity problemis also discussed in this
section. Section 2.6 covers a link state routing protocol and its working. In the
section 2.7 hierarchical routing is discussed. Section 2.8 deals with the Internet
Protocol (IP). In this section Ipv4 and IPv6 along with ICMP, DHCP and IP
security are covered. Section 2.9 discusses routing in the Internet and protocols
RIP, OSPF and BGP. Section 2.10 is about multicast routing. In section 2.11
Mobile IP is introduced. Section 2.12 summarizes the chapter. In Section 2.13
review questions and their solutions are covered. Section 2.14 lists further
readings.

2.1 OBJECTIVES

After completing this section, one should be able to:

 understand how the working of shortest path routing algorithm;
 construct a spanning tree;
 understand the working of distance vector routing and link state routing;
 understand hierarchical routing;
 understand Internet Protocols IPv4 and IPv6, and
 understand and implement multicast routing.

2.2 FLOODING

Whenever any link’s or router’s state is changed either up or down, leads to

change in the topology. And whenever a topology change happens the same is
required to be communicated to all the nodes of the topology. This is done by
sending a topology change message to all the nodes (very large in numbers) in
the network using broadcasting. In networking such type of broadcasting is
known as flooding. Flooding is of two types: uncontrolled flooding and
controlled flooding. Uncontrolled flooding does not restrict sender to send the
packet to the node from which it received the same packet. In controlled
2
 Network Layer
flooding a node does not forward the packet back to the node from which it has
received the packet. Another technique to reduce the duplicate packets in the
network is to make the provisions so that a node relays a packet only once. To
implement this the sender adds it ID and a unique auto increment sequence
number to each packet. Whenever a packet is received by a node, it stores the
sequence number and the ID of origin node. Before relaying the packet, node
checks the sequence number of the newly received packet and the sequence
number stored for the origin sender. A packet will only be relayed if the
sequence number of the packet is greater than the sequence number stored for
the origin sender. By doing so the duplicate packets in the network are reduced
to a great extent. Now days uncontrolled flooding is not used in general. In
flooding the source node sends a packet (of information to be shared) to all its
neighbours: the nodes connected with a direct link. These nodes further send
the packet totheirneighbours, and this process continues till each node in the
network receives the packet.

t=1 R2 t=1 R5
R2 R5
t=0 t=2
t=0 t=1
t=1 t=2 t=2 R1 R4
R1 R4 t=2
R7
R7
t=1 t=2
t=2 t=0 t=1
t=2
t=0
t=1 R3 R6
R3 R6

R1 to R7: Routers, Pkt: Packet, R1: Origin node

(a) (b)

Figure 1: Packet Flooding (a) without spanning tree (b) With spanning tree
Construction
Considering the figure 1(a).Suppose a change in topology is observed by node
R1. R1 will send the notification packets toR2 and R3. R2 will send the packet
to R4 and R5. R2 will send the packet to R1 (as it has received the same from
R1). In similar way node R3will send the packet to R4 and R6. Node R4 has
received the same notification packet originated by same origin with same
sequence number. So R4 will further send the packet arrives first and will
discard the later one. Similarly R6 will discard the later one and forwards the
first received packet to R7. Node R5 receives packets both from R2 and R4, so
in same way the packet arrives first will be forwarded and later one will be
discarded. R7 will receive the packet from both R5 and R6.
Another way to reduce the redundancy of packets in the network and to avoid
the cyclic forwarding of the packets is to construct a logical spanning tree of
the topology.
Considering L as the number of bi-directional links of the network, for a
packet to be broadcasted the total number ofpacket transmission lies between L
and 2L. Arrows on the links show packet transmissions with the time of
3
Routing Algorithms
transmission (assumed to be 01 unit for each packet) shown. In figure 1, the
flooding is shown with both the methods without and with spanning tree
construction. In flooding without spanning tree the number of packets
transmitted is in generally many more as in the case of with spanning tree.
Also, in both the cases, the broadcast packet reaches to all nodes within same
time. For a graph many spanning trees are possible, hence the flooding time
depends on the spanning tree constructed.

2.3 SHORTEST PATH ROUTING ALGORITHM

• In Shortest path routing method router builds a graph of the network,

where node of the graph represents the router and connecting links are the
lines joining the nodes. Shortest path between two nodes is calculated by
considering parameters as the number of hops, or the geographic distance
between them etc. Shortest path routing technique is based on the greedy
approach by considering the next best possible option without considering
the overall best option. Hence, sometimes it may be the case that overall
there may exist some other shortest path between two nodes as selected by
the algorithm.In this method least cost paths from one node (‘source”) to
all other nodes are computed.

Many algorithms are proposed based on this technique,Dijkstra algorithm is

one of the widely popular shortest path routing algorithm. Dijkstra algorithm is
also based on greedy approach as is also known as the least cost path approach.
The working of Dijkstra algorithm is as follows:
1) Select the source node as the start node (S).
2) Mark the direct neighbor nodes as tentative nodes (Initially all the nodes are
considered as tentative nodes).
3) Choose the node from tentative nodes list with lowest cost from the source
node and mark it as permanent nodes and make it as source node.
4) If, the destination node is covered or there is no more nodes in the tentative
list(i.e. no more nodes to be explored) then stop, otherwise go to
step number 2.
Considering the figure below and applying Dijkstra routing algorithm for
finding the best path or the shortest path fromsource node A tothe destination
node E. The steps followed are as follows:

4
 Network Layer

1) Node A is selected as source node.

2) Direct neighbor nodes B and C are marked as tentative node.
(3) Node B has the lowest cost path (of cost 1) from source node A, so it is
marked as permanent node.
3) Node B is not the destination node and node D, E are available nodes in the
tentative nodes lists, so
(4) Make node B as the source node now and explore the direct neighbors of it.
(5) Nodes D and E are the direct neighbors of B,
(6) Node D has minimum cost path from B so, node D is selected and marked
as permanent node and D is not the destination node and also the tentative list
is not empty.
(7) Make D as the source node now and explore its direct neighbors.
(8) Node C and E are the direct neighbors of D.
(9) Node E has the less path cost than C from node D, so E is marked as
permanent node. Node E is destination node, so stop here.
(10) The shortest path from A to E is: A – B – D – E.

2.4 DISTANCE VECTOR ROUTING

In today’s scenario where the number of Internet users increased manifolds’,

static routing is not feasible as the static routing algorithms are not adaptive in
nature. Under such dynamic scenario dynamic routing algorithms performs
very well.
Considering dynamic algorithms: Distance vector routing and link state
routing are most widely known and used dynamic algorithms. Distance Vector
Routing Algorithm has got following properties:
• Distributed – Each node receives some information from one or more
of its direct neighbours and performs path calculation at its own.
• Iterative – iterative in nature means exchange of information among
neighbours continues until no more (new/updated) information
available to exchanged.
Here, in this section distance vector routing algorithm is discussed. Many
timesdistance vector algorithm is known as Bellman-Ford algorithm because
vector algorithm is based on Bellman-Ford Equation.

5
Routing Algorithms

Bellman-Ford Algorithm
Each router also constructs/ maintains routing table known as Distance Vector
table storing the path cost (in terms of distance or the hop count) to reach ALL
feasible destination nodes from itself. The path cost iscalculated using
information received from the neighbour’s distance vectors.
A router maintains following information for Distance Vector table -
 Router ID (each router has an ID)
 Link cost for each link connected to a router
 Intermediate hops
Initially the Distance Vector table is initialized as follows:
 Cost to itself (C = 0
 Cost to all other routers = infinity.

Algorithm: Distance Vector Routing-

1. A router shares its routing table/Distance Vector to its neighbours

(directly connected routers).
2. Each router on receipt of Distance Vector from its neighbours, it saves
the most recently received Distance Vectors.
3. A router updates its routing information/ Distance Vector according to the
Bellman-Ford equation, when:
 A neighbour shares a Distance Vector with routing information
different than before.
 The status of a link to its any of the neighbour changes (up or down).
The Distance Vector is measured while minimizing the cost to each
destination router.
Notations:
Dx[y] = Estimate of minimum distance from node x to node y
C[x,v] = Node x has cost to each neighbour v
Dx= [Dx[y]: y ∈ N ] = Node x records distance vector
Node x also maintains its neighbours' distance vectors
Note:
- For each neighbour v, x maintains D v = [Dv[y]: y ∈N ]
 Each node sends its own routing information/ Distance Vector
information estimate to its neighbours after certain time interval.
 When any node x shared with a new Distance vector information from
any of its neighbor v, it stores the distance vector of node v and updates
its own routing information using Bellman-Ford equation:
Dx[y] = MIN{ C[x,v] + Dv[y], Dx[y] } for each node y ∈ N
Example – Considering the figure below, construct the routing tables for
routers X, Y and Z.

6
 Network Layer

3
1

5
X Z

Step 1: Each node knows the distance to reach to its direct neighbour nodes.
The distance to itself is 0 and the distance to nodes which are known
discovered yet is considered as ∞.
Initially, the DV/ routing table of each node can be as:

X Y Z
X
Y 3 0 1
Z Y

3
1

X Y Z X Y Z
X 0 3 5 5 X
Y X Z Y
Z Z 5 1 0

Step 2: For router X:

Router X will share its distance vector (DV) information to its direct
neighbours and neighbours (Y and Z) will share their DV information/routing
table to X.
The distance from source node X to destination node will be determined
using bellmen- ford equation.
Dx(y) = min { C(x,v) + D v(y)} for each node y ∈ N
When node Y shares its DV information with X, X will recalculate its DV
information using Bellman Ford equation.
Given data:
Clearly, Cv(z) = 1, Cx(z) = 3, Cx(Y) = 3, dx(x) = 0
Now, using the bellman Ford equation and the DV information shared by
node Y to node X.

7
Routing Algorithms
dx(z) = min { [c(x,y) + dv(z)], [c(x,z) + dz(z)]}
= min {[3 + 1], [5 + 0]} = 4

X Y Z
X
Y 3 0 1
Z Y

3
1

X Y Z X Y Z
X 0 3 54 5 X
Y 3 0 1 X Z Y
Z Z 5 1 0

Similarly, when node Z shares its DV with node X:

No updates in X’s DV information. [Note that Z still have a distance 5 to
node X]

X Y Z
X
Y 3 0 1
Z Y

3
1

X Y Z X Y Z
X 0 3 4 5 X
Y 3 0 1 X Z Y
Z 5 1 0 Z 5 1 0

Similarly, when Y shares its DV information with node Z:

dz(x) = min { [c(z,y) + dv(x)], [c(z,x) + dx(x)]}
= min {[1 + 3], [5 + 0]} = 4

X Y Z
X
Y 3 0 1
Z Y

3
1

X Y Z X Y Z
8 X 0 3 4 5 X
Y 3 0 1 X Z Y 3 0 1
Z 5 1 0 Z 54 1 0
 Network Layer
Finally the routing table for all –

X Y Z
X 0 3 4
Y 3 0 1
Z 4 1 0 Y

3
1

X Y Z X Y Z
X 0 3 4 5 X 0 3 4
Y 3 0 1 X Z Y 3 0 1
Z 4 1 0 Z 4 1 0

At the end of convergence process, the DV information of all the nodes are
same until a new change in the topology occurs.

2.4.1 Comparison
The comparison of the two routing algorithms approach should be based on the
new path processing time and the traffic generated by these for the routing
information convergence process.
The evaluation of an algorithm is moreover depends onthe implementation
approach and the specific implementation.
The discussed routing algorithms can be compared on following points:
1. Message complexity
• Link State algorithm:Link state algorithm sends order of O(nE)
messages with n nodes, E links.
• Distance Vector algorithm: Messages areexchange between directly
connected neighbors only
2. Speed of Convergence
• Link State algorithm:It takes order of O(n2) to converge, where n is the
number of routing nodes
• Distance Vector algorithm: Convergence time is not standard with this
and varies due to following situations:
– may be routing loops
– count-to-infinity problem
3. Robustness: Robustness is the confidence of getting the correct result under
any circumstance.
• Link State algorithm:Link State algorithm can face issues like:
• node can advertise incorrect link cost
• each node computes only its own table
• Distance Vector algorithm:Distance Vector algorithm can face issues
like:
• Distance Vector node can advertise incorrect path cost

9
Routing Algorithms
• each node’s table used by others, so if a node shares incorrect
path it could be propagated further to the network on a large
scale.

Further, both can be compared on some more points as follows:

Distance Vector Routing Link State Routing
• Tell neighbors about distance of • Tell about distance to each
all the destination. neighbor to all routers
• Node’s computation depends on • Each router computes its
neighbors. best paths
• Each router constructs a distance
vector table with (dist/cost, exit
interface) tuple for
eachdestination.
• A node shares copy of distance
vector table to all its direct
neighbors.

2.4.2 The Count-to-Infinity Problem

One of the serious drawbacks in Bellman Ford algorithm is Count to Infinity

Problem. It is also known as routing loop problem. In Bellman Ford algorithm
implementation count to infinity is also known as routing loop problem. In
Bellman Ford algorithm, routing loops problem is faced when a link/interface
shutdown/goes off. This issue may also occur, when two routers exchange
routing information to each other at the same time.

Consider the below figure to discuss it in detail.

R1 R2 R3 R4

The routing table for above topology can be (considering the distance of each
link is to be 1 unit). Each cell shows the pair (distance, predecessor node):
R1 R2 R3 R4
R1 0, - 1, R1 2, R2 3, R3
R2 1, R2 0, - 1, R2 2,R3
R3 2, R2 1, R3 0, - 1, R3
R4 3, R2 2, R3 1, R4 0, -

As it is visible in the figure, node R1 is connected to rest of the topology with

only single link i.e R1will be cut off to rest of the topology if the link is down.
Now, say the link between R1 and R2 goes down.
10
 Network Layer
As a result, routers R1 and R2 are the source to know this change in the
topology. Rest all other routers will be able to get this information only when
R1 and R2 share their routing information with them. The router R2 updates
this change in topology in its routing information.
Since R3 is unaware of the link down between R1 and R2, R3 has the
information that R1 is reachable with a cost of 2 via R2 (1 for R3 to R2 and 1
for R2 to R1), as it is not aware of the link break between R1 and R2.
Routers share their routing information after regular time interval, resulting the
router R2 receives router R3’s routing information.
When R2 receives R3’s table, it assumes that R1 is reachable via R3 and it
updates its routing information with infinity to 3 (1 for R2 to R3 and R3 shared
the cost as 2 for reaching to R1).
Further, after certain time interval when routing information is shared again
among routers.
The node R3 receives R2’s routing information, it sees that R2 has updated the
cost to reach to R1 from 1 to 3, so R3 also changes its routing table with cost 4
to R1(as R3 discovered the node R1 very first time from R2 only).
This process continues until all routers reaches to the cost of link to A is
infinity. To stop this loop route poisoning technique is used, in which a
number is considered as the limit of the distance for a path, beyond that cost
the path is considered as unreachable and it is called as Route Poisoning. For
RIP routing protocol infinity is defined as 16, that is beyond the cost value 16
the path is considered to beunreachable.
This situation is shown in table below (each entry shows [distance, predecessor
node])
R2 R3 R4
Sum of cost to R1 after link down ∞, R2 2, R2 3, R3
Sum of cost to R1 after 1st updating 3, R3 2, R2 3, R3
Sum of cost to R1 after 2nd updating 3, R3 4, R2 3, R3
Sum of cost to R1 after 3rd updating 5, R3 4,R2 5,R3
Sum of cost to R1 after 4th updating 5, R3 6, R2 5, R3
Sum of cost to R1 after 5th updating 7, R3 6, R2 7, R3
Sum of cost to R1 after nth updating .... .... ....
∞ ∞ ∞ ∞

In the above table it is understood that the network will not be able to converge
ever. The root cause of this issue is the sharing of routing information with the
node (R2) from which it (R3) first discovered that node (R1).
As discussed above one of the possible resolution of this problem is Rout
Poisoning and another resolution is with Split Horizon.
In Split horizon Rule says that, the information about the path for a
destination (say for R1) is never sent back in the direction from which it was
received i.e. R3 discovered node R1 through R2 so, R3 will not send back the
same path information which was received from R2 about R1 to R2.

11
Routing Algorithms
2.5 LINK STATE ROUTING

The Distance Vector routing algorithm is driven by the sharing of self routing
information with neighbours which leads to routing challenges as the count to
infinity problem. In DV routing algorithm rumors can be spread in a many fold
speed and strength.
For these reasons, a new routing algorithm introduced namely: Link State
Routing algorithm also known as shortest path first.
Link state routing approach is inspired by road navigation map. In contrast to
DV approach, in LS each router has a complete view of thenetwork topology.
In link state protocol each routershares information about itself, its directly
connected links, and the state of thoselinks. Instead of sharing routing
information (routing table containing cost) as in DV, in link state information
regarding the status of the link is shared. On receiving information shared by
other routers, each router keeps a copy of it and further passes it without any
change in it. Each router independently computes the best route (route with
minimum cost) to reach to every possible destination in the topology. That is,
after convergence each router has the map (topology) of the entire network
designated to it. In link state routing each router has the same routing
information. When there is a change in the topology, directly affected router
send the change in routing information to all routers in the topology.
Link state routing protocol maintains three tables namely: neighbor table,
topology table and actual routing table to perform routing.
Link state routing protocols are the most widely used protocols in the Internet.
Some of the widely used link state protocols are: Open Shortest Path First
(OSPF) and Intermediate System to Intermediate System (IS-IS).
Working of the Link State Routing protocol:
Link state routing protocol can be divided into five parts as written by
Tanenbaum. Each router of the topology uses link state routing protocol
performs following actions:
1. Building of Neighbour table:
In link state algorithm a special type message/packet namely: HELLO is
used to discover neighbour nodes in the network. A router sends HELLO
message on each of its connected link. Neighbor routers reply with their
network addresses. The router uses this information and the port on which
it received this information to build up its neighbor table.
2. Path cost measurement Measure to neighbour nodes:

A path cost is determined by sending an ECHO message/packet. A link state

routing node on receipt of this ECHO message (with no payload, a very small
size message) it immediately (with no time delay) sends back the ECHO
message. The node calculates the duration starting from the sending of ECHO
message to receipt of the reply. This duration is called as the round trip time
for the node replied. The path cost for this destination node will be half of the
round trip time with assumption that the delays are symmetric (same delay
from sender to receiver and receiver to sender, which is not always true). The

12
 Network Layer
path cost may be a composite metrics with factors like the end-to-end delay,
throughput, or a combination of these.

3. Once the node has discovered the neighbours and their path cost it
constructs a packet called as link state packet (LSP) including the link cost
to these neighbours. The structure of the LSP is shown in table below. This
packet is broadcasted in the network.

Advertiser ID Network ID Cost Neighbour ID

……………
……………… ……………… … ………………
……………… ……………… …………… ……………
…
4. Each node constructs routing information to all possible destination nodes
with the help of routing information received from other nodes of the
topology by applying link state algorithm like Dijkstra’s algorithm.

Problems in Link State Routing

One of the major drawbacks to the link state routing protocols is that the
CPU overhead to recalculate the route due to change in topology is very high.
Another drawback is the amount of memory required to store the routing
information i.e. the neighbor tables, routing table and the full map of the
topology.
If a node advertises wrong neighbor information, the error is propagated to the
whole topology.

2.6 HIERARCHICAL ROUTING

As discussed in link state and distance vector routing algorithms, each router
has to store routing information in the form of routing table. In the routing
table router stores information of path for remote networks i.e. the path cost,
the exit interface.
The amount of routing information to be stored is directly proportional to the
number of routers in the network. That is for a small size network with few
numbers of routers the routing information to be stored by routers can be
handled easily. Whereas, for a network with large number of routers, the size
of routing information to be stored is highly voluminous in size. The purpose
of routers is to route (finding the path) the packet to destination. As a result the
routing tables will become big in size and will consume more space on router
as well as more bandwidth in the network when shared.
To overcome this problem, instead of a flat structure the network can be
designed as a hierarchical structure.
Considering the following example with distance vector routing algorithm
node A has to store 21 entries into its routing table (considering each path
13
Routing Algorithms
having a cost of 1 unit). Exit interface is the interface through which the
destination is connected (here I1 and I2 are considered as the interfaces of A) .

H I
F

A I1
I2
B D E G

C
A

T S O L K

R P
U N M
A Q

Destination Exit Interface Cost

A -- 0
B I1 1
C I2 1
D I1 2
E I1 3
F I1 4
G I1 4
H I1 3
I I1 3
J I1 4
K I1 5
L I1 4
M I1 6
14
 Network Layer
N I1 5
O I2 4
P I2 5
Q I2 6
R I2 5
S I2 3
T I2 2
U I2 3

From table and figure above, it is visible that the traffic due to exchange of
these routing tables will be high.
One possible solution of this can be, if routers are divided into small
groups(called as Regions) in which they have to store routing information of
the routers of the region they belongs to. A router stores only one entry
collectively for all routers of a region.
In the example discussed here, the complete network can be classified into 6
regions as shown below:

Region 2

Region 3
Region 1
H I
F

A I1
I2
B D E G

C
A

Region 6 Region 4
Region 5

T S O L K

R P
U N M
A Q
Let’s again construct the routing table of router A with hierarchical routing
approach:

15
Routing Algorithms
A’s Routing table for Hierarchical routing
Destination Exit Interface Cost
A --- ---
B I1 1
C I2 1
Region 2 I1 2
Region 3 I1 3
Region 4 I1 4
Region 5 I2 3
Region 6 I2 2

If A wants to send packets to any router in region 3 (E, F or G), it sends them
to the interface I1. From the above table it is clear that the routing table size is
reduced leading to improved efficiency due to less overhead of the traffic in
the network.
Hierarchical routing further can be classified into levels. In the example
discussed above a two-level hierarchical routing is implemented. The level of
hierarchical routing is chosen according to the size (number of routers) of the
network. A three or four level hierarchical routing can also be used. In a three-
level hierarchical routing, the network is classified into a number of clusters.
Where, each cluster contains a number of regions, and each region contains a
number of routers. In Internet at a wide scale commonly hierarchical routing is
used.

2.7 THE INTERNET PROTOCOL (IP)

2.7.1 IPV4 addressing

IP addresses are used to uniquely identify and locate any system

connected in the Internet i.e. two networked systems cannot be assigned
identical IP address (although private IP addresses are reusable among
private networks, will be discussed later in this chapter). Internet Protocol
version 4 (IPv4) is the 4thversion of the Internet Protocol (IP). IPv4 uses a 32-
bit address space with total number of 232 unique IP addresses, but from these
large number of IP addreses are reserved for special purpose in networking.
While performing routing, IPv4 addresses are used by routers. Some of the
IPv4 addresses are reserved (as shown in table below) for private networks and
multicast addresses and for future/ scientific purpose.

Address range Reserved as Description

10.0.0.0– These addresses are assigned to systems

Private network
10.255.255.255 connected within a private network.

16
 Network Layer

Address range Reserved as Description

127.0.0.0– Used for local host or the loopback

Host
127.255.255.255 addresses.

Assigned to hosts connected with a link

169.254.0.0– directly, when there is no other device to
Subnet
169.254.255.255 allocate the IP address. Known as link
local address.

172.16.0.0– These addresses are assigned to systems

Private network
172.31.255.255 connected within a private network.

192.168.0.0– Used for communications within

Private network
192.168.255.255 a private network.

224.0.0.0– Used to send message to a group of

Multicast
239.255.255.255 hosts.

Address representations
IPv4 addresses are commonly written in dot-decimal notation. In this 32-bits
are divided into 4 octets separated by periods(.). In dot-decimal notation each
octet is written in decimal format.
For example, IP address 172.16.1.32. For computing purpose, sometimes it is
convenient to use binary notation of IPv4 addresses.
The 32 bits of the IPv4 addresses are divided into two parts: network portion
and host portion. Five classes of IPv4 addresses are defined
Private networks
Private IP addresses are used in private networks managed by single authority.
Private IP addresses are reusable among private networks i.e. private IP
address used in one private network can be reused in another private network.
Private IP addresses are not routable in the public Internet; that is they are not
recognized by public routers. Therefore, hosts with private IP address cannot
communicate with public networks directly, there is a need of network address
translation (NAT) system for this purpose.
IPv4 address classes:

IPv4 address range is classified into five classes; A through E. These classes
are identified by the first octet of the IP address. The details are as shown in
table below:

17
Routing Algorithms
1st Byte 2nd Byte 3rd Byte 4th Byte Description
Class A 0 to 127 (in decimal) Open Open Open The 1st MSB bit
00000000 to 01111111 (can take (can (can take of the 1stoctet is
(in binary) any take any any value always set to 0
value value between 0 (zero) (as shown
between between to 255) in red color)
0 to 255) 0 to 255)
Class B 128 to 191 (in decimal) Open Open Open The first two
10000000 to 10111111 (can take (can (can take MSB bits of the
(in binary) any take any any value 1stoctet are
value value between 0 always set to 10
between between to 255) (as shown in red
0 to 255) 0 to 255) color)
Class C 192 to 223 (in decimal) Open Open Open The 1st three bits
11000000 to 11011111 (can take (can (can take of the 1stoctet are
(in binary) any take any any value always set to 110
value value between 0 (as shown in red
between between to 255) color)
0 to 255) 0 to 255)
Class D 224 to 239 (in decimal) Open Open Open The 1st four bits
11100000 to 11101111 (can take (can (can take of the 1stoctet are
(in binary) any take any any value always set to
value value between 0 1110 (as shown in
between between to 255) red color)
0 to 255) 0 to 255)
Class E 240 to 255 (in decimal) Open Open Open The 1st four bits
11110000 to 11111111 (can take (can (can take of the 1stoctet are
(in binary) any take any any value always set to
value value between 0 1111 (as shown in
between between to 255) red color)
0 to 255) 0 to 255)

2.7.2 Datagram Format

The data unit of IP is named as packet. For each of the IP packet, control
information is added which is used by intermediate nodes to make it delivered
successfully to the destination and also used by end to end nodes for
confirmation of correctness of the message. This control information is added
at the starting of the content called as header of the packet. An IP packet has
two sections: a header section (with control information of IP) and a data
section (payload handed over by upper layer).
Header of IP packet
The header of the IPv4 packet consists of 14 fields, of which first thirteen field
are necessary to be included and the last 14thfield is (options) optional to add.
The header of the IP packet is formed as big endian format (most significant
byte first). The most significant bit is numbered 0. The version of the IP
protocol is the first field of the header (four bits of the 1st byte). The structure
of the IP header is shown in figure below.

18
 Network Layer

Version
The first field of the IP header the protocol version of the Internet Protocol
(IP). This field is of four bits length. For IPv4, as it is the 4 th version of IP so
version field contains 4 (0100).
Internet Header Length (IHL)
As the header of the IPv4 packet is not of the fixed size due to the 14 th field
options, it is necessary to include the size of the header so that the receiver is
able to separate the header fields from the payload of the packet. This field is
of 4 bits. The minimum value for IHL is 5 and the maximum is 15. The value
of this field is calculated by multiplying IHL value with 32 and the result will
be in bit, i.e. IHL field value 5 means: 5 x 32 bits =160 bits = 20 Bytes. The
maximum value of IHL can be 15 (4 bit length), that is the maximum size of
the IPv4 header can be 15 × 32 bits = 480 bits = 60 bytes.
Differentiated Services Code Point (DSCP)
This field is used to specify the type of service (ToS) for the packet in
transmission. At present this field specifies differentiated services (DiffServ).
This field is commonly used by the real-time data streaming applications. An
example is Voice over IP (VoIP), which is used for interactive voice services.
Explicit Congestion Notification (ECN)
This field is used to provide end-to-end congestion control mechanism to avoid
dropping of packets.
Total Length
This field is of the size 16-bits. This field defines the size of the entire packet
in bytes, that is header and data. This field can take a value between 20 bytes
(only header with no data) and 65,535 bytes.
Identification
This field is used for the purpose of identification of the packets for uniquely
identifying the group of fragments (breaking the packet into smaller size due to
constraints of routers or the network links) of a single IP datagram.
Flags
A total of 3 flags are defined each with 1 bit. The purpose of these flag values
is to control or identify fragments. These flags are as follows:

 bit 0: Reserved; must be zero. [most significant bit of flag field]

 bit 1: Donot Fragment (DF)
 bit 2: More Fragments (MF) [least significant bit of flag field]
A packet can be fragmented iff, the DF flag is cleared (assigned a value 0). If
the DF flag is set (assigned a value 1), and the size of this packet is more than

19
Routing Algorithms
than the MTU (Maximum Transferable Unit]) value, that is fragmentation is
required, the packet will be dropped.
The MF field denotes that there are more fragments available after this one of
the original packet. MF flag is 0 (zero) for unfragmented packets, and the last
fragment of a packet. The last fragment of a packet has a non-zero Fragment
Offset field, differentiating it from an unfragmented packet.
Fragment Offset
This filed is of the size 13bits. Fragment offset of a fragment is measured in
units of 8 byte blocks. This field represents the position of a particular
fragment with respect to the beginning of the unfragmented (original) IP
packet. The first fragment has an offset of zero. The maximum value of this
field can be (213 – 1) × 8 = 65,528 bytes, which would exceed the maximum IP
packet length of 65,535 bytes with the header length included (65,528 + 20 =
65,548 bytes).
Time To Live (TTL)
Under some circumstances many of the times packets got stuck in loops in the
network and consumes the bandwidth of the network unnecessarily. The size
of this field is 8 bit. This field restricts packets to enter to live infinitely long
time in the Internet. This value is in seconds, but time intervals less than 1
second are rounded up to 1. The value of this field is set to a number (known
as maximum hop count limit). When the packet arrives at a router, the TTL
field value is decrement by one. The packet is dropped by the router if it
encounters TTL field value to be 0 (zero).
Protocol
This field contains the protocol used at the upper layer (transport layer) in the
data portion of the IP datagram.
Header Checksum
IP supports error-checking of the header. On arrival of the packet at a router,
the checksum is calculated again of the header and compared with the
checksum field of the header. If both values do not match, the router discards
the packet. On each of the intermediate router the TTL field value is decreased
by one so the router must recalculate the checksum value of the header.
Source address
IPv4 address of the sender of the packet is included in this field. The size of
this field is 32 bit. This field is the IPv4 address of the sender of the packet. If
the sender belongs to the private network (having private IP address), this has
to be changed in transit by a network address translation (NAT) device.
Destination address
IPv4 address of the receiver of the packet is included in this field. If the
receiver belongs to the private network (having private IP address). If the
receiver belongs to the private network (having private IP address), this has to
be changed in transit by a network address translation (NAT) device.
Options
In general this field is not used while forming IP packets.

2.7.3 IP Datagram Fragmentation

• Each IP datagram is supposed to be encapsulated within the link-layer

frame for transportation from one router to other. In the Internet
different types of links are used to connect various routers operating
20
 Network Layer
with differentlink layer protocols. Each link layer protocol sets the limit
on length of an IP datagram known as MTU (maximum transferable
unit). MTU: Maximum amount of data that a Link-Layer frame can
carry.

When a packet arrives at the router, its destination address is examined to get
the outgoing link on which it has to be forwarded. Once the outgoing link is
identified its MTU is determined. If packet size is more than the MTU value,
and the IP packet header flag field ‘Do not Fragment (DF)’ value is 0 (zero),
then the router fragments the packet into smaller parts and sent one by one on
the link. The allowed maximum size of any fragment can be MTU value
minus the IP header size i.e. it ranges from 20 bytes to 60 bytes).

Considering the following example, the MTU of the exit link is 1500 bytes. A
datagram of the size 4000 bytes with identification number 777 and DF flag bit
set to 0 is received (given that including 20 bytes of transport layer header).

Here, MTU size is 1500 bytes that is the maximum size of the packet
(header (20 bytes) + payload 1480 bytes can be allowed on the link.
So, for the 1stfragment:
– Payload/ data in the packet =1480 bytes
– offset = 0 (data of this packet should be inserted at byte 0 at the
time of reassembling)
– identification number = 777
– MF flag value= 1 (there are more fragments after this fragment
of the original packet)
nd
2 fragment
– Payload/ data in the packet =1480 bytes
– offset = 1480 (data of this packet should be inserted after byte
1480 at the time of reassembling)
– identification number = 777
– MF flag value= 1 (there are more fragments after this fragment
of the original packet)
rd
3 fragment
– Payload/ data in the packet =1020 byte (=3980-1480-1480)
information field
– offset = 2,960 (data of this packet should be inserted after byte
2960 at the time of reassembling)
– identification number = 777
– MF flag = 0 (this is the last fragments of the original packet)

21
Routing Algorithms
Reassembly of the fragmented parts of the packet is performed only at
the end system, routers are not allowed to reassemble the fragments in
transit.

2.7.4 IPv6

Due to increase in users of the Internet, IPv4 addresses are exhausted. Hence,
the size of IP address required to be increased to accommodate all the users of
the Internet. IPv6 is the 6th version of Internet protocol and the size of the IPv6
address is 128 bit. Interoperability between IPv4 and the IPv6 is not provided,
and thus shifting to IPv6 was not easy at all. There was a need of intermediate
system which sits in between these two protocols and acts as a convertor for
both. Several transition mechanisms have been proposed to make the
communication between these two protocols possible.

IPv6 not only provides a large addressing space but also permits hierarchical
address allocation methods that facilitate route aggregation across the Internet,
and thus limit the size of routing tables even in a very large network.

Address Representation

IPv6 addresses are represented in hexadecimal format. 128 bits of the address
are divided into eight groups, separated by colons, each of size 16 bits
represented into 4 hexadecimal digits. Example of IPv6 address is
2001:0db8:0000:0000:0000:8a2e:0370:7334. Further, the IPv6 address can be
shortened by omitting continuous 0(zero) groups and placing double colon (::)
instead. The leading 0 (zeros) in a group can also be omitted. For example,
2001:0db8:0000:0000:0000:8a2e:0370:7334 address can be written as -
2001:db8::8a2e:370:7334.

The host portion of IPv6 address is fixed of the size 64 leaving remaining 64
bits for the subnet size.

IPv6 packets

An IPv6 packet has two parts: a header and payload.

Payload/Data

22
 Network Layer

The header consists of fixed portion and the variable/ optional portion. The
fixed portion is of size 40 octets and consists of eight compulsory fields to
support minimal functionality required for all packets and optional extensions
to implement special features.

Version field is to define the version of the IP protocol, here its value will
always be 0110 (version 6). Traffic class and the flow label fields are used to
provide traffic specific QoS. As the fixed header length is not variable so
header length field is not necessary here. But there is need to identify the last
bit of the packet, hence the field Payload length is added which tells the size of
the payload (total size = header size + payload size). After the header either the
optional header fields will be inserted or the payload is inserted. Next Header
field is used to help the receiver to interpret the data which follows the header.
The "Next Header" field of the last option points to the upper-layer protocol
that is carried in the packet's payload.Without options, at max 64kB of the
payload can be inserted.

In contrast to IPv4 packet, IPv6 packets are not allowed to be fragmented in

transit by routers. Hop Limit field of size 8 bits is used to protect looping of a
packet in the Internet.

2.7.5 Internet control message protocol

Internet control message protocol (ICMP) is a network layer protocol. Since

Internet Protocol(IP) does not have a built-in mechanism for sending error
and control messages, a separate protocol namely: ICMP is used to support
error control. It is responsible to report errors and management queries.
ICMP is a implemented in the networks devices like routers for sending the
error messages and operations information. e.g. host is unreachable. Another
use of the ICMP is in congestion control in the network and flow control
between sender and receiver by sending a request to decrease traffic rate.
ICMP packet is also used in defining the TTL value of a path.

2.7.6 Dynamic host configuration protocol

Dynamic Host Configuration Protocol (DHCP) is a network management

protocol. For a network with large number of IP based systems (Systems
assigned an IP address) it is almost impractical to assign IP addresses
manually. To automate this process Dynamic Host Configuration protocol was
designed. DHCP server dynamically assigns an IP address and other network
configuration parameters (like: default gateway, subnet mask, DNS server IP
etc) to each device on a network.

2.7.7 IP Security

The IP security (IPSec) protocol is used services likeconfidentiality, integrity

and authentication. IPSec protocol can be used to encrypted, decrypted and
23
Routing Algorithms
authenticated packets. IPsec was introduced to provide encryption services for
application layer data. It can also be used to provide security for routing data
generated by transmitted across the public internet. Authentication services are
also provided by IPSec without encryptionlike to authenticate that the data
originates from a known sender. In IPv6 IPSec is made available as optional
header of the IP packet. IPSec also provides facility to establish a circuit using
IPsec tunneling to transmit the data in encrypted form between two endpoints.

2.8 ROUTING WITH INTERNET

2.8.1 Intra Autonomous System Routing in the Internet: RIP & OSPF
An intra-autonomous system routing protocols are responsible to provide
routing capabilities to routers within an autonomous system (An autonomous
system (AS) is a very large network or group of networks with a single routing
policy.). Intra-AS routing protocols are also known as interior gateway
protocols. RIP (the Routing Information Protocol), and OSPF (Open Shortest
Path First) are the most widely used Intra-AS routing protocols.
1. RIP
RIP (Routing Information Protocol) is based on distance vector routing
algorithm. RIP uses Bellman Ford routing algorithm. RIP is a proprietary
routing protocol of Cisco and available with Cisco routers. In RIPv2 is
capable of preventing routing loops in the network. A maximum Hop count
value 15 is used for this purpose. RIPv2 is implemented with mechanism like
split horizon, route poisoning and hold-downto prevent spreading of rumours
and routing loops. RIP is suitable for a small size network. RIP uses UDP as
transport layer protocol making it light weight protocol.
2. Open Shortest Path First (OSPF) :

OSPF is based on link-state routing algorithm. This is an open protocol, i.e

anyone can use it freely. OSPF protocol uses Dijkstra’s algorithm. It finds
shortest path for each source to all destination pair. One of the advantages of
it is that it uses multicast routing in a broadcast domain. Another advantage
of it is that it can handle the error detection by itself.

 Difference Between RIP and OSPF

SR.NO RIP OSPF

RIP is Routing OSPF is Open Shortest

1 Information Protocol. Path First protocol.

Routing Information
Protocol is based on the Open shortest path first
Bellman Ford protocol is based on
2 algorithm. Dijkstra algorithm.

3 RIP is a DV protocol. OSPF is a link state

24
 Network Layer
SR.NO RIP OSPF

Distance or hops count protocol. It uses

are the metric used in bandwidth, congestion
measuring the path cost. metric while identifying
the best path.

This protocol is best

This protocol is suitable suitable for large size
4 for small size networks. network.

A maximum hop count Hop count restriction is

5 of 15 is allowed in RIP. not there in OSPF.

RIP uses User OSPF works for Internet

7 Datagram Protocol. Protocol.

2.8.2 Inter Autonomous System Routing protocol: BGP

An autonomous system (AS) is a network managed and controlled by a single
authority. Inter Autonomous system routing protocols are responsible to
route a packet among various autonomous systems. Border Gateway Protocol
(BGP) is one of the inter autonomous system routing protocol which is used
as the routing protocol in the Internet to exchange routing information among
ISPsmanaged by different Authorities.

It can connect any network of ASirrespective of the topology used. The only
requirement to connect many AS together is that each of the AS should have
at least one router running with BGP. BGP’s is responsible to exchange
network reach ability information with other BGP systems. BGP constructs
an ASs’ graph based on the information exchanged between BGP routers.

2.9 MULTICAST ROUTING

In the Internet at many instances there is a need to send same information to a

group of clients at the same time. In such cases if, the unicast routing is used
the data has to send to individual client and the server has to connect with each
client independently to send the same information leading to overburden the
server as well as the network. Instead of this if broadcasting is used to send the
information, even it is also a wastage of resources computing as well as
networking as not all the clients are interested in sent information or
sometimes the information is not supposed to be disclosed to them. Hence, in
both situations broadcasting approach is not suitable.
Multicast routing algorithm is used to handle such cases (sending a message to
a group of users/ clients).
In multicast routing the most important part is the group management.Under
group management tasks like group creation, deletion and management of
25
Routing Algorithms
membership to the group(like join and leave)are performed. When a host
joins/leaves a group, the information is propagated to its router. A router may
be a member of one group, many groups or not a member of any group. While
performing routing to send a message to a group, router requires the
information about the members of the group. Hence, the multicast routing
algorithm has to maintain this information of group membership. This
information can be maintained and propagated in two ways: either the host
informs to router about their membership, or routers send a query to their hosts
periodically. Each router periodically shares their group management
information to theirneighbors, and like this the information propagated through
the subnet.
In addition to group management, a logical spanning tree is constructed of the
topology to perform the multicast routing. Once the spanning tree is
constructed pruning is performed for each of the group. For there may exist
more than one spanning tree of a graph, so each node will construct its own
spanning tree.
Once the spanning tree is constructed it is then pruned for a group. Pruningis a
technique to preserve links connecting hosts that are member of the group
only.
One of the methods of pruning the spanning tree can be: starting from the last
node of the path and moving towards the root, remove all routers that do not
belong to the group under consideration.
Let’s consider an example to understand the working of multicast routing.
Here, there are two groups 1 and 2. Some of the hosts belong to group 1 and
some to group 2 and some belongs to both 1 and 2 simultaneously. Here to
perform multicast routing, each router constructs a spanning tree covering all
otherrouters. Figure below shows one of the spanning trees for the router R1.

1 R1

1,2 2 2

R2 1 R3 1,2

26
 Network Layer
Figure below shows the pruned spanning tree for group 1 for the spanning tree
constructed for R1 above.
1 R1

1 2 2

R2 1 R3 1

Similarly, the pruned spanning tree for the group 2 of spanning tree of router
R1 is as shown in figure below:

1 R1

2 2 2

R2 1 R3 2

After pruning is completed, the multicast packets are forwarded only along the
corresponding spanning tree. As the basic requirement of this algorithm is to
store separate pruned spanning tree for each member of every group. Hence
this method is not suitable for large networks.

27
Routing Algorithms
2.10 MOBILE IP

Increasing the number of mobile devices with Internet access leads to invent
the new modified protocol for these devices namely: Mobile IP. This
protocol is designed by extending standard Internet Protocol. It is designed
by keeping in mind the mobility of the devices and it provides the ability to
users that they canswitch to another network with the same IP address
without dropping out the connection.
This protocol allows location-independent routing of IP packets throughout the
Internet. A mobile device is always recognized by the home address assigned
irrespective of its current location in the Internet.

2.11 SUMMARY

In this unit we have learnt about the routing of a packet. That, how a packet
reaches to destination from the source following the best route. Shortest path
routing is simple to understand and implement hop count based routing
approach. It is static in nature, means a graph is constructed by the source node
for all the destination nodes in the network. Dijkstra’s algorithm is one of the
widely used shortest path approach based routing algorithm. It is also known
as the greedy approach based routing algorithm. Distance Vector routing
algorithm is another solution for routing in the network. DV approach is
dynamic in nature that individual node constructs a complete map of the
network topology. Each node constructs the routing table with cost and the
vector component for each of the remote network. Distance vector routing
algorithm is based on the Bellman-Ford equation. Distance Vector based
approach face the count to infinity problem which is addressed by
implementing split horizon and route poisoning together. Above both methods
of routing are not suitable in a large network due to huge traffic generated by
routers to exchange routing information. Hierarchical routing is one of the
possible solutions to perform routing in large networks with huge number of
routers. Complete network is divided into smaller sub-networks in the form of
hierarchical network. Further,we have learnt about Internet protocol and its
two versions: IPv4 and ipv6. The address space of IPv4 is of the size 32 bits
and that is of IPv6 is of 128 bits. IPv6 also include optional headers fields. One
of the features provided by IPv6 as optional header is the IPSec. ICMP and
DHCP are major network management protocols. ICMP is used for many
services like, congestion control, flow control, network diagnosis etc. DHCP is
responsible for assigning IP address, subnet mask, and gateway and DNS
information to the clients in a network. It is very difficult to manage this
information manually so managed efficiently by inserting a DHCP server.

28
 Network Layer
2.12 SOLUTIONS/ANSWERS

Review Questions:

1) Which of the following statements are True or False.

A Distance vector routing is a static routing algorithm. T F
B Dijkstra’s algorithms can be run locally in link state routing to
T F
constructthe shortest path.
C Flooding discovers only the optimal routes. T F
D Flooding generates lots of redundant packets. T F
E In hierarchical routing, each router has no information about routers
in other regions. T F

F A spanning tree is a subset of a graph that includes some of the

nodes of that graph. T F

G Sending a message to a group of users in a network is known as

broadcasting. T F

2) What are the problems with distance vector routing algorithm?

3) Explain the spanning tree.

Solution:

1)

A. False, B. True, C. False, D. True, E. True, F. True, G. False

2) Distance vector routing algorithm faces the count to infinity problem. The
convergence is slow. Routing information isexchange among direct neighbors
only. Chances of rumors about false routing information are always there in
distance vector routing. Due to which a packet may enter into routing loop.

3) A spanning tree is a tree with no cycles and constructed such that all the
vertices are covered with minimum possible number of edges.

2.13 FURTHER READINGS

1. Computer Network, S. Tanenbaum, 4th edition, Prentice Hall of India,

New Delhi 2002.

2.Data Network, Drnitri Berteskas and Robert Galleger, Second edition,

Prentice Hall of India, 1997, New Delhi.

3. Data and Computer Communication, William Stalling, Pearson Education,

2nd Edition, Delhi.

29
 Network Layer
UNIT 3 CONGESTION CONTROL
ALGORITHMS

3.0 Introduction
3.1 Objectives
3.2 Reasons For Congestion In The Network
3.3 Congestion Control Vs. Flow Control
3.4 Congestion Prevention Mechanism
3.5 General Principles Of Congestion Control
3.6 Open Loop Control
3.6.1 Admission Control
3.6.2 Traffic Policing And Its Implementation
3.6.3 Traffic Shaping And Its Implementation
3.6.3.1 Leaky Bucket Shaper
3.6.3.2 Token Bucket Shaper
3.6.4 Difference Between Leaky Bucket Traffic Shaper And Token Bucket Traffic
3.7 Congestion Control In Packet-Switched Networks
3.8 Summary
3.9 Solutions/Answers
3.10 Further Readings

3.0 INTRODUCTION

In the Internet nodes acting as transmitting nodes are inserting packets into the
Internet and nodes acting as receiving nodes consume the packets from the Internet.
Internet has a capacity to handle the traffic load (packets). When the rate of insertion
of packets into the Internet is higher than the rate of consumption of the packets from
the Internet at last Internet is unable to handle the traffic and the performance of the
resources of the Internet is degraded. This situation is termed as congestion.

Hence, the goal of congestion control algorithms is torefrain the transmitter from
inserting packets in the network more than the handling capacity of the Internet.

In this unit section 3.3 discusses about the reasons for congestion in the network.
Section 3.4 differentiates congestion control from flow control. Congestion prevention
mechanisms are covered in section 3.5. In section 3.6 general principles of congestion
control are elaborated. Further in section 3.7 congestion control technique namely:
Open loop control is discussed. Section 3.8 is about the congestion control in packet-
switched networks. Section 3.9 summarizes the unit. Problems and their solutions
covering the entire unit are discussed in the section 3.10. Section 3.11 enlists further
readings.

3.1 OBJECTVIES

After completing this unit, one should be able to:

 Identify the reasons of the congestion in the Internet;

 Differentiate congestion control and the flow control;
 Devise the preventive measures of the congestion;
1
Congestion Control
Algorithms  understand the congestion control techniques;
 understand the close loop and open loop congestion control techniques,
and
 understand congestion control in packet-switched networks.

3.2 REASONS FOR CONGESTION IN THE

NETWORK

In the Internet there can be several reasons to occur the congestion. When many
transmitters insert data packets on to input lines at a time and are to be sent on the
same output line, assuming that the capacity of the output line is much less than that
of the packets received then a long queue will be build up for that output line. In this
situation if the buffer memory if not big enough to hold all these packets, then extra
packets will be dropped. To stop dropping of the packets, if the memory available is
made infinitely large even then the congestion may be reduced but the overall quality
of the service of the traffic will be worse; because by the time packets reach to the
output line to get dispatched, there TTL (time to live) value gets expired and their
duplicate packets been already inserted into the network. If all the packets carried to
the final destination, these duplicate packets will increase the traffic load only in the
Internet and will be discarded, due to time out. So, it will be good for the Internet to
drop these packets as soon as their TTL value gets expired.

Another reason of the congestion in the Internet is the slugging performance of the
processors of the intermediate devices. If any of the intermediate router’s CPU is
performing slower than expected speed, their jobs (i.e.Queuing buffers, routing
packets, updating tables, reporting any exceptions etc.), will be slowed down. The
arrival rate of the packets at input line is greater than the processing and removal of
the packet from output line. This again creates a situation of congestion.

Another point of issue is the LowBandwidth. Due to low bandwidth capacity of the
lines amount of the traffic increases in the network causing congestion.

Resolution of any one of the issues discussed above will not handle the
congestion;instead it will just shift the bottleneck to some other point. The root cause
of the real problem is the mismatch of the capacity (computing or the carrying) of
various components of the system. Once congestion happens in the network the
routers respond to overloading by simply dropping the packets.

The bursty nature of traffic is one of the major causes of congestion. This could be
controlled by restricting the insertion of the traffic at a uniform rate.

3.3 CONGESTION CONTROL VS. FLOW

CONTROL

Congestion control and flow control are two different things, which are mixed up at
times. As discussed earlier congestion control is the entity of the network whereas
flow control is about regulating the transmission of data between devices on the
connection/link between them, and not what is happening in devices between them.
2
 Network Layer
Flow control is about point-to-point traffic control between sender and receiver for a
specific transmission to avoid packet drop at receiver. If the incoming traffic rate is
higher than the processing rate at receiver, the receiver is flooded and packet will be
dropped. To overcome this situation the receiver should send some kind of feedback
to sender to inform the sender about the drop of packets and to slow down the sending
speed. This is called flow control between sender and receiver and is handled at
transport layer responsible for end-to-end data delivery. Congestion is a situation
when the traffic in the network is higher than the handling capacity of the network.

3.4 CONGESTION PREVENTION MECHANISM

Congestion control is to restrict the traffic load below the handling capacity of the
network. As shown in Figure below, the congestion control techniques can be broadly
classified into two categories:

• Open loop: Methods to prevent or avoid congestion are classified as open loop
techniques. Open loop methods ensure that congestion state never exists in the
network. Open loop policies are applied in the network to prevent congestion before it
happens. The congestion control policies are applied either at the source or the
destination.

• Close loop: these methods acts to treat or alleviate the congestion once it happens.
Once the system enters to congestion state, closed loop techniques used to detect it,
and then take action to bring the system out of it.

Open Loop solutions are static in nature. These policies are not adaptive in nature and
do not change according to the present state of the system. These methods take
decisions about when to accept packets, when to dropthem etc. These methods make
decision without taking into consideration the present state of the system. The open
loop congestion control methods are further classified on the basis of whether these
are applied on source or on destination.

Close loop congestion control techniques are based on the concept of feedback. These
techniques are dynamic in nature and actions are taken during transmission. Some
system parameters are continuously measured in the network and whenever a
congestion state is observed, feedback system is used to take action to reduce the
congestion. Open loop techniques work as per the following 3 steps:

Step 1: Continuous monitoring of the network to detect the congestion state,

the actual location of the congestion and devices involved.

Step 2: Sending thefeedbackabout congestion state to the devices where

actions can be taken

3
Congestion Control
Step 3: Take the necessary actions to remove the located congestion.
Algorithms

Some of congestion control algorithms based on these techniques are discussed in

following sections.

3.5 GENERAL PRINCIPLES OF CONGESTION

CONTROL

Congestion in the network can be measured in terms of various Metrics like: the
average queue length, timed-out packets, delay, packets dropped due to unavailability
of buffer space, etc.

As discussed previously, congestion occurs in the network when senders insert

packets more than the handling capacity of the network. The responsible entities for
the congestion in the network are: the sender: sending packets without considering the
status of the network, the intermediate resources: speed of the router, bandwidth of the
bottleneck link, control messages generated by intermediate device etc. The
congestion in the network can be controlled by either dropping the excess packets or
to restrict the sender by inserting packets into the network at a lower speed. In general
TCP provides mechanism to control the congestion. Internet protocol header also
provides ECN field to notify the sender about the congestion in the network. The
congestion happens in the network and happened due to sender, hence the network
entities are required to notify the sender about the congestion and accordingly the
sender takes necessary steps to control it.

3.6 OPEN LOOP CONTROL

As discussed in previous section, open loop methods are preventive measure for
congestion control.

Some of the open loopmethod based policies of congestion control are discussed
here:–

Retransmission Policy :

A packet transmitted with reliable data delivery protocol, if it’s TTL value expired
before it reaches to destination, gets dropped. The sender has to retransmit such
packets until get delivered successfully. More the congestion in the network leads to
more packet drops which leads to retransmission of these packets leading to more
traffic in the network. This retransmission leads to congestion in the system.

To prevent congestion due to this issue, value of timers used by retransmission policy
must be set such that state of congestion in the network is prevented and also able to
optimize the efficiency of the network.

Window Policy:

In Go-Back-N window if a packet is lost/received out of order, several packets are

resent, although some packets may be received successfully at the receiver side. This
4
 Network Layer
may increase the congestion in the network. Therefore, selective repeat window
should be preferred instead of Go-Back-N. In selective repeat window packets
dropped arethat may have been lost.

Discarding Policy :

Routers has to adapt a good discarding policy such that congestion in the network is
prevented at the same time a router must attempt to discard corrupted or packets of
unreliable services (i.e. UDP) by maintaining the quality of the messages with reduced
number of retransmission of dropped packets.

Packets transmitted with UDP services may be discarded before the packets
transmitted with reliable services i.e. TCP. The video streaming over the internet may
tolerate some loss of packets while text messages may not tolerate loss of any packet.

Acknowledgment Policy :

Acknowledgment is sent by receiver to notify the sender about receipt of the packet or
not. Even though the size of acknowledgement packets issmall in comparison to the
data packets but still they also offer traffic load in network. In order to reduce the
number of acknowledgement packets sent, the receiver should wait for the next
incoming packet and if it is in sequence with the previous packet instead of sending
the acknowledgement of individual packet a cumulative acknowledgement is sent for
both the packets. That is sending a cumulative acknowledgement of packets received
in sequence can save the bandwidth of the network.

Admission Policy :

Admission policy is applied in a network to prevent the congestion. Before allowing

the traffic to enter into the network, switches first check whether the resources
required are available or not. The traffic is only admitted if the available resources are
more than the requirement of the traffic. The requested virtual circuit is established iff
there is no chance of congestionafter reserving the resources for this.

The policies discussed above can be used to prevent congestion before it happens in
the network.

3.6.1 Admission Control

In this section, how the congestion is handled in a virtualcircuit network is discussed.

Admission control is a closed-loop technique, in which action istaken once network
enters into a congestion state.Admission control policy is the first and very simple rule
to prevent the congestion in network. It says that admit the traffic on the condition that
it will not be responsible for congestion in the network. If there is congestion in the
network, the first attempt to recover the network from congestion state could be to
stop new packets to be inserted into the network. The thumb rule to prevent the traffic
is “admit traffic only when the network can handle it without congestion”.Admission
control policy is feasible and can be applied successfully in the circuit-switched
datagram networks. This is not easy to identify the source of congestion hence could
not be applied in packet-switched networks.

Some of the admission control systems are as follows:

5
Congestion Control
Algorithms Admission Control Methods:

The goal of admission control methods is to estimate the expected bandwidth

requirement for the incoming traffic and determine whether this traffic can be
allocated the needed bandwidth, such that congestion state does not occur. Admission
control methods are widely used for real time application sensitive towards delay and
jitter. Many admission control methods are available. Some admission control
methods are based on mathematical calculations and statistical indicators, and others
are based on measuring traffic state.

Admission control methods are generally classified into two categories: parameter
based and measurement based admission control. Parameter based methods are based
on the characteristics of the active traffic and do not consider new incoming traffic,
hence are not optimal. Measurement based methods consider the real time network
conditions by serving new incoming traffic, hence a higher network utilization may be
achieved.
Each admission control method follows the principle that, allocatethe
availablebandwidth tothe incoming traffic flows only in case of not exceeding the
capacity of the line. For a node to implement admission control policy, it should have
access to QoS parameters i.e delay, packet drop rate etc.By doing so the traffic can
achieve the QoS as desired, but should be independent of the type of traffic
underwent.
Similarly, in case of virtual circuit subnets, no more new virtual circuits are accepted
once congestion state in the network is identified.
3.6.2 Traffic Policing and its Implementation
Traffic policing is to monitor the traffic flow in the network. If the traffic flow rate is
greater than the specified rate, traffic policing methods simply discards the overflow
packets.Traffic policing can be used to control both inbound and outbound traffic.
Traffic policing methods maintain a constant flow (pre-defined) of traffic.

Traffic policing does not hold packets received above the allowed flow rate, hence
does not require buffer. It is easy to implement traffic policing in comparison to traffic
shaping as it does not require maintaining packet buffers. Traffic policing does not
cause delay, and queuing, rather it simply discards the packets.
Components of an implementation of traffic policing system are as follows:

6
 Network Layer

 Meter: this component measures the traffic and provides the measurement
result to the next component (marker) for further action.
 Marker: marker assigns colorsto packets out of green, yellow, or red based
on the measurement result provided by the meter. Marker provides this
coloring information to the next component namely: Action.
 Action: this component performs actions based on packet coloring results
received from the marker. This component performs following actions in
accordance to the pre-defined rules:
 Pass: a packet will be forwarded further if it meets network
requirements.
 Re-mark + pass: local priority of the packets not meets the network
requirements are changed and forwarded.
 Discard: packets not meeting network requirements are dropped.
If the rate of traffic is below the threshold value, packets are marked with green and
yellow color and forwarded, whereas if the rate of traffic exceeds the threshold value,
packets are either marked with yellow, lowers the priority and forwarded or marked
with red color and dropped according to the traffic policing configuration.
3.6.3 Traffic Shaping and its Implementation

In contrast to traffic policing, traffic shaping tries to adjust the rate of outgoing traffic
instead of dropping the packets to ensure an even transmission rate. Traffic shaping
makes use of a buffer to hold bursty traffic for a while to control the traffic. Packets
are delayed if the system is unable to forward all of them at a time and will be
forwarded as the link is found free. It is a congestion control technique which delays
some packets to remove the congestion state. Traffic shaping is not practically
applicable to traffic of real time applications. Traffic shaping can be used to control
the outbound traffic only.

Further, Traffic shapers can be classified into two categories based on their
capabilities; simple traffic shaper and advanced or more sophisticated traffic shapers.
Simple traffic shapers shape all traffic uniformly. Whereas advanced traffic shaper
can classify the traffic and can be used as a technique to provide Quality of Service
7
Congestion Control
Algorithms
(QoS) to a traffic category by delayingother category of trafficto bring them into
compliance with a desired traffic profile.

Two of the widely known traffic-shaping algorithms are leaky bucket and token
bucket, discussed in next section in detail.

3.6.3.1 Leaky Bucket Shaper

Leaky bucket shaper as its name says, is based on the way aleaky bucket functions.
It sends out the traffic at a fix rate even if the incoming traffic is bursty in nature.
Bursty traffic could not be sent out at a time,, will be stored in the buffer (called the
leaky bucket) and will be sent out once the outgoing line is free.

In the figure above, it is considered that the capacity of the network to carry the
traffic is of 3 Mbps. The leaky bucket traffic shaper will not send traffic above 3
mbps in the network. Here, the host inserts a burst of data at a rate of 8 Mbps for 2
sec, and sends 16Mbits of data. Further, it does not send any data for next2 sec and
then sends data at a rate of 2 Mbps for 4 sec, by sending 8Mbits of data. The host
inserts a total data of 24Mbits in a duration of a 8 sec. After applying the leaky
bucket traffic shaping policy the traffic is sent out at a rate of 3 Mbps for the
duration of 8 sec. Here, traffic shaping policy smooth the traffic in the network.
There are not data in the duration 2 to 3 sec, and a burst of data during the interval
0 to 2 forcing the network to congestion state. Leaky bucket policy can transmit
this whole data in a smooth manner without any congestion in the network.

3.6.3.2 Token Bucket Shaper

8
 Network Layer
The leaky bucket traffic shaping policy discussed in previous section, does not
consider the input traffic pattern. It shapes the traffic with a fixed defined rate.

Token bucket traffic shaping policy considers the input traffic bursts and allows
sending the traffic on a higher rate also to prevent the drop of packets.

Token bucket policy uses the leaky bucket which holds tokens generated at regular
intervals (one policy to add tokens is to generate a token per clock tick). Token bucket
policy works as follows:

 Token are generated at regular intervals and placed into the bucket.
 The bucket has a maximum capacity of holding the tokens.
 A packet can be sent to the output line only if a token is available in
the bucket.
 Once a packet is sent on output line, is removed from the bucket.
 As many tokens are removed from the buckets as number of packets
are sent from the bucket.
 If there is no token available in the bucket, the packet cannot be sent.

Token bucket policy shapes the bursty traffic by allowing bursty traffic on output line
but to a limit of available number of tokens. Figure below shown the working of the
token bucket traffic shaping mechanism. In the figure host A sent 3 packets and there
are only 2 tokens available in the bucket, hence only 2 of these packets are transmitted
on the output line and 1 is hold back and will be sent once a token is placed in the
bucket.

Host A Host A

Unregulated
Flow 1 Packet left (no
more tokens
2 tokens
available No tokens
left in bucket

Regulated
2 tokens
Flow

3.6.4 Difference between Leaky Bucket Traffic Shaper and token Bucket Traffic

Shaper

Difference between Leaky and Token buckets –

Leaky Bucket Token Bucket

Host places the packet to Tokens are generated in fix
be sent in the bucket. intervals and placed into
the bucket.
The traffic is sent onto the Output traffic rate is
output link at a constant regulated by the number of
9
Congestion Control
Algorithms rate tokens available in the
bucket.
Leaky bucket shapes the The output traffic may be
bursty traffic into uniform bursty (to a limit of tokens
traffic. available in the bucket).
The bucket can hold a finite A packet can be sent only
number of packets in a when a token is available in
queue and outputs at finite bucket.
rate

3.7 CONGESTION CONTROL IN PACKET-

SWITCHED NETWORKS

Congestion control is one of the very important parts to be considered while

designing any packet-switching networks due to rapidly growing network and
bandwidth intensive network applications. Various methods are proposed for
congestion control.

Source
1 10-M
bps
Ethe
rnet Router
Destination
1.5-Mbps T1 link
I
FDD
bps
Source 1 00-M
2

In the figure above, source 1 and 2 inserts traffic at a rate of 10 and 100 Mbps
respectively. The router can transmit the traffic on output link limited to 1.5 Mbps.
The packets will start dropping at router once the buffer is full and the state is kown as
the congestion state. Congestion control mechanism in packet switched network can
be applied on either transport layer or network layer. Flow is a sequence of packets
flowing between a source/destination pair and following the same route through the
network.TCP provides connection oriented reliable service at transport layer while
Internet protocol (IP) provides connectionless packet delivery service.Routers does
not maintain any state of the flow for connectionless service whereas state of the flow
is maintained at routers for the connection oriented service.The Internet
Protocol (IP) provides the basis for packet delivery and the Transmission Control
Protocol (TCP) provides a best-effort delivery mechanism.Best-
effort delivery service is the basic packet deliveryserviceswithout guarantee of
delivering it. The best efforts are made to deliver packets to the destination, but there
is no mechanism to recover lost packets. At transport layer a TCP window is used to
control the transmission rate according to feedback received from the sub network. As
the congestion is a network layer issue and happens in the network, the routers play
crucial role in handling the congestion state. Each router is installed with certain
10
 Network Layer
buffers to hold the incoming packets could not be sent at the moment due to
congestion. Many policies are applied to these incoming packets to handle them in the
buffer queuing. Some of the possible choices in queuing algorithms are:FIFO also
called Drop-Tail, Fair Queuing (FQ), Weighted Fair Queuing (WFQ), Random Early
Detection (RED) etc. Routers also send a special type of packet namely: choke packet
for the purpose of congestion handling. Routers monitor the utilization of their output
line and send choke packets back to hosts using output lines whose utilization has
exceeded some warning level. Another solution frequently used to control the
congestion state is Explicit Congestion Notification (ECN) used by routers at network
layer to notify the sender about the congestion state.An ECN-aware router sets a field
in the header of the IP instead of dropping a packet to signal about the congestion.
The receiver of the packet notifies about the congestion to the sender, which reduces
its transmission rate.

3.8 SUMMARY

In this section we have discussed about the congestion control state in the network. A
network is congested when traffic in the network is more than its capacity to handle it.
The congestion occurs when the number of packets into the network is more than its
handling capacity. The bursty nature of traffic is the root cause of congestion.When
part of the network no longer can cope with a sudden increase of traffic, congestion
builds upon. Other factors, such as lack of bandwidth, ill-configuration and slow
routers can also bring up congestion.

Flow control is an issue of data link layer whereas congestion control is an issue of
network layer. Flow control is meant to prevent a fast sender from crushing a slow
receiver. Flow control can be helpful at reducing congestion, but it can't really solve
the congestion problem. Many congestion control techniques are applied in the
network to avoid the congestion state in the network. Open loop and closed loop
congestion control techniques are the broad categories of the congestion control
algorithms. Traffic policing and traffic shaping are the main techniques of open loop
congestion control. Traffic policing is, sending the traffic at a fix rate irrespective of
the incoming traffic pattern. In contrast to traffic policing, traffic shaping tries to
adjust the rate of outgoing traffic instead of dropping the packets to ensure an even
transmission rate.

3.9 SOLUTIONS/ANSWERS

Q1. What is congestion?

Ans :In the Internet nodes acting as transmitting nodes are inserting packets into the
Internet and nodes acting as receiving nodes consume the packets from the Internet.
Internet has a capacity to handle the traffic load (packets). When the rate of insertion
of packets into the Internet is higher than the rate of consumption of the packets from
the Internet at last Internet is unable to handle the traffic and the performance of the
resources of the Internet is degraded. This situation is termed as congestion.

Q2. Why congestion occurs?

11
Congestion Control
Algorithms
Ans: In a packet switched network, every intermediate device maintains buffers/
queues to hold packets while processing them to forward further. Under the situations
of receipt of the bursty traffic these buffers gets full and packets are dropped. As a
result as per the quality of service of the dropped packets, they may require to be
retransmitted, further increasing the traffic in the network. At last the system enters to
congestion state.

Q3. What are the two basic mechanisms of congestion control?

Ans :Congestion in the network can be addressed in two ways: preventive method and
recovery method. In preventive method,actions are taken such that congestion doesn’t
occur and recovery method allows the system to enter in congestion state and then it
tries to remove it.

Q4. How congestion control is performed by leaky bucket algorithm?

Ans : In leaky bucket algorithm, packets are inserted into the bucket. In case of bucket
overflow, packets are dropped. The packets are exited from the bucket at a constant
rate allowing bursty incoming traffic into the network at a constant rate.

Q4. In what way token bucket algorithm is superior to leaky bucket algorithm?

Ans : The leaky bucket algorithm is very conservative in nature in the sense that it is
not adaptive to the incoming traffic. Token bucket algorithm is made sensitive
towards incoming traffic. The output rate is not dependent on the predefined upper
limit rather, it depends on the availability of the tokens in the bucket. In the starting if
the tokens are available in enough quantity the rate can be more and once there are no
tokens available after that the output is limited by the rate of token generation.

Q5. Differentiate traffic policing and traffic shaping.

Ans. Difference between Traffic Policing and Traffic Shaping:

S.NO. Traffic Policing Traffic Shaping

Traffic policing is a mechanism Traffic Shaping is a congestion

which monitors the traffic in any control mechanism that brings
1. network. delays in packets.

The packets with rates that are It buffers the packets with rates
greater than the traffic policing that are greater than the traffic
2. rate are discarded. shaping rate.

Traffic policing doesn’t cause Traffic shaping causes delay of

3. delay. packets.

The token values are calculated in The token values are calculated in
4. bytes per second. bits per second.

5. In traffic policing queuing of Queuing of traffic is not

12
 Network Layer
S.NO. Traffic Policing Traffic Shaping

traffic is not performed. performed in traffic shaping.

Traffic policing supports traffic Traffic shaping doesn’t supports

6. remarking. traffic remarking.

Traffic policing can be used to

control outbound or inbound Traffic policing can used to
7. traffic. control outbound traffic only.

3.10 FURTHER READINGS

Computer Network, S. Tanenbaum, 4th edition, Prentice Hall of India,New Delhi 2002.
Data Network, DrnitriBerteskas and Robert Galleger, Second edition,Prentice Hall of
India, 1997, New Delhi.
Data and Computer Communication, William Stalling, Pearson Education,2nd Edition,
Delhi.

13
 Network Layer
UNIT 4 EMERGING NETWORKING
TECHNOLOGY

4.0 Mobile Ad Hoc Networks (MANETs)

4.1 MANET Characteristics
4.2 MANET Limitations:
4.3 Advantages of MANET
4.4 Applications of MANET
4.5 MANET Architecture:
4.6 MANET-Operating Principle
4.7 Challenges facing ad hoc networks
4.8 Wireless Sensor Networks (WSN)
4.9 Applications of Wireless Sensor Networks (WSN)
4.10 Structure of WSN
4.11 Classification of WSNs
4.12 WSN Topologies
4.13 IoT introduction
4.14 Characteristics and Applications of IoT
4.14.1 IoT in Home Automation: Intrusion Detection Use Case
4.14.2 IoT in Smart Cities: Smart Lightening Use Case
4.14.3 IoT in Environment: Air Pollution Monitoring Use Case
4.14.4 IoT in Health: Wearable Devices Use Case
4.15 IoT System Architecture
4.16 Summary

4.0 MOBILE AD HOC NETWORKS (MANETS)

Today, we live in a digital age and are constantly surrounded by digital devices, such
as laptops, mobiles, cameras, music players, etc. These devices are connected through
various communication networks like WiFi, Bluetooth, Cellular Networks etc. In the
earlier times, communication infrastructure had a fixed wired backbone with
stationary cell towers and access points, as shown in figure 1. Such infrastructure
networks, for example, cellular networks, were suitable for locations where access
points were easy to install.
Infrastructure

AP: Access Point

AP
AP

Figure 1: Typical communication between Infra. and APs.

1
Emerging Networking
Technology However, today, the nodes (laptops, mobiles, etc.) are not always fixed at one point
but can move in the system shown in figures 2, they connect to the fixed
communication infrastructure of cell towers and communication access points
(routers) further connected to the Internet backbone. However, networks can be
configured without APs too. A collection of at least two or more electronic devices
equipped with wireless communication and networking capability forming
infrastructure-less networks without base stations is called a mobile ad-hoc network
(MANET). MANET is a self-organizing and self-configuring multi-hop wireless
network with a dynamically changing topology. The devices (nodes) in the network
not only act as hosts, receiving data but also as routers that send data to other nodes in
the network. MANET supports peer-to-peer communications and peer-to-remote
communications with significantly reduced administrative costs. Therefore, the
MANET can be defined as an infrastructure-less, fully distributed network in which
each node acts as a transmitter, receiver, and router. The way communication is
accomplished between the components present in the network is the main difference
between the wireless and wired networks.

The term "Ad Hoc" network means a network that is temporary or implemented
immediately when needed for a particular purpose. With the technological
advancements in recent times, it is possible to use mobile nodes and mobile routers (a
node can act as a router) to create a communication network where everything is
mobile and nothing is fixed. In such a network, the topology of the network changes
continuously as the neighbours of any given node are likely to change frequently. A
few nodes will join the network at any point in time, while a few will disengage from
the network, making the environment ad-hoc. With minimal or no dependence on
infrastructure, a MANET is easy to deploy to support communication and computing
anywhere, anytime.
ad-hoc network

Homogeneous network
Figure 2: Ad-hoc and Homogenous network

In heterogeneous networks, the devices have varying capabilities, while all the nodes
have identical capabilities and responsibilities in homogenous networks.

4.1 MANET CHARACTERISTICS

 A MANET does not have a centralised infrastructure, unlike the traditional

mobile wireless networks that need base stations, access points and servers to
make it operational, as illustrated in Figure 1. Figure 2 shows a decentralized
ad hoc network with all interconnected mobile nodes receiving data while
functioning as routers simultaneously. One crucial characteristic of MANET
is that it is a self-configuring network in which the network activities such as
discovering the dynamic topology, delivery of messages and receiving data
packets are executed by the nodes themselves.

 MANET has a dynamic topology in which node scan freely move in and out
of the network arbitrarily. It causes the network topology to change rapidly
and unpredictably over time. For nodes to communicate in such dynamic
2
 topologies, alternative paths are automatically discovered and data packets are Network Layer
forwarded across these multi-hop paths. Various route discovering
mechanisms are used in MANETs to accomplish this.

 Another characteristic of a MANET is that it operates on lower bitrates with

varying capacity wireless links. Compared to wired networks, a MANET has
relatively low bandwidth, high bit error rate, and unstable and asymmetric
links. Due to these factors, MANETs encounter higher delays and traffic
congestion very frequently in comparison to fixed networks.

 A MANET is a network of nodes that are often hand-held devices with small
battery backups. The battery-powered devices run out of battery support quite
frequently. Thus MANETs are characterised by energy-constrained operations
in which power conservation is essential. As a power conservation method,
the network uses nodes in a manner so that they radiate power as little as
possible and transmit data only when essentially required. Nodes can be
configured to go into sleep mode.

 Compared to wired networks, MANETs have lesser physical security. As a

result, they are vulnerable to various attacks such as eavesdropping,
interception, denial-of-service and routing attacks. Therefore, different cyber
security techniques are deployed to make MANETs more secure and robust
against attacks.

4.2 MANET LIMITATIONS

The present and future need for dynamic ad-hoc networking technology is
tremendous. This highly adaptive networking technology, however, still faces various
limitations.

Throughput Drops with More Hops

Nodes in the transmission range of each other can send the data packets directly.
However, a sequence of multiple hops is required when the node needs to send data
packets to a non-neighbouring node. The data packets in such a case are forwarded to
the destination node with the help of intermediate nodes acting as routers. With the
increase in the number of intermediate hops, the throughput decreases rapidly.

Throughput Drops with Increasing Mobility

Highly mobile nodes result in increased overheads due to frequent topology
variations. Suppose, in a situation of route failure, the need to discover new routes
emerges and it results in enhanced demand for data packet retransmission. For
example, in the proactive routing protocol, the routing table is used in which each
node keeps a list of all available destinations and the number of hops required to reach
each destination. The routing table also needs to change as the network's topology
keeps changing. Any changes in routing table are then relayed to all the other nodes in
the network, which significantly increases overheads in data transfer. Larger overhead
leads to a lower percentage of the data packets reaching their destinations.

Delay Time
The time taken by a data packet to arrive at its destination after leaving the source
node is referred to as delay. Nodes are busy with the transmission and receiving the
packets to increase the network's throughput. Because of it, the queue at each node is
always not empty or small which can lead to a longer delay.

History of Ad-hoc Networks:

3
Emerging Networking
Technology The first generation of ad hoc network, called Packet Radio Network (PRNET), can
be traced back to the 1970s when the Defence Advanced Research Project Agency
(DARPA) launched packet-switched radio communication to provide reliable
communication. PRNet provided an efficient means of sharing broadcast radio
channels among many radios. PRNET, with low throughput (2 kbps per subscriber
approximately), was not entirely infrastructure-less as it required a static station for
routing. The PRNET then evolved into the smaller, cheaper, energy-efficient and
cyber-resilient Survivable Adaptive Radio Network (SURAN) in the early 1980s. The
United States Department of Defence (DOD) developed Globe Mobile Information
System (GloMo) and Near Term Digital Radio (NTDR),providing a self-organizing
and self-healing network. The NTDR, which uses clustering and link-state routing,
was a widely used ad hoc network. The Internet Engineering Task Force (IETF)
adopted IEEE 802.11 as the standard routing protocols for MANET for using mobile
devices like PDA's, palmtops, notebooks, etc.

4.3 ADVANTAGES OF MANET

Compared to traditional networks, MANETs are cost-effective solutions and easy to

set up, especially in geographical locations where fixed infrastructure is not feasible,
as the nodes are not tethered. The device-to-device communication is used to create,
implement and manage the MANET without any need for fixed hardware
infrastructure. The ability to connect to the internet without any wireless router is the
main advantages of MANET, making it more affordable than a traditional network.
Their ability to handle connection failures is also created as routing and transmission
protocols are designed to manage these situations. MANETs are extensively chosen
due to their other distinct advantages, such as the ability to auto-configure, adaptive,
self-diagnosing capability and cost-effectiveness.

4.4 APPLICATIONS OF MANET

The applications of MANET are not limited to a single area. There are many
applications of MANET in industry, defence, medical and environmental
sciences. A few of the applications include:

Tactical networks Military Communication automated Battlefields

Sensor Network Remote weathers for sensors, earth activities
Emergency Services Disaster recovery, earthquakes, crowd control and
commando operations
Educational Setup virtual class & conference rooms
Applications
Entertainment Multi-user games, robotics pets
Location-Aware Automatic Call forwarding, advertise location-
Services specific services, Location–dependent travel guide.

4.5 MANET ARCHITECTURE

Depending on the application and the area to be covered, the MANET

architecture may use body area networks (BAN), personal area networks
(PAN), local area networks (LAN), metropolitan area networks (MAN) and
wide (WAN) area networks as enabling technologies.

4
BAN coverage extends to about 2 m and is usually deployed using wearable Network Layer
devices. When the communication requirement is of the order of 10-15m, PAN
is used. MANETs using LAN easily cover an area ranging up to 500m and are
widely deployed for communicating within residential colonies, markets, or a
cluster of buildings. WAN can also be used for even larger areas, although
large networks have limitations in addressing, routing, and security
management.

The MANET architecture layers, shown in figure 3.

 Link Layer
 Network Layer
 Transport Layer
 Application Layer

It is assumed that the appropriate physical layer is available.

Figure 3: Reference MANET layer architecture

Link Layer:
The joint design of the link and physical layer can significantly enhance a MANET's
efficiency and reliability. The link layer identifies priority packets and schedules
packet delivery according to priority levels. It is accomplished through a Media access
control (MAC) layer with a MAC discriminator and priority classifier. MAC is a
protocol that enforces a methodology to allow multiple devices access to a shared
media network.
The queue management section schedules the packets according to the priority levels.
Data for real-time applications like live movies, video conferencing, etc., get higher
priority than the packets for applications such as FTP and Email. When the network is
congested, some packets may be dropped. Assigning higher priority to real-time
packets ensures that they are not dropped in case of network congestion. Queue
management thus helps in the timely delivery of data packets in real-time applications
and improves the packet delivery ratio.
The MAC discriminator in the layer differentiates data packets that arrive from the
wireless channel and sends them to the network layer. The address resolution protocol
(ARP) packets go to the queue directly. The MAC packets used in IEEE 802.11 stay

5
Emerging Networking
Technology in the MAC layer. The bandwidth estimation control packets are sent to the bandwidth
estimation module for use in the routing layer's adaptive scheme.
The priority classifier classifies the data packets that arrive from the packet queue,
whether required for real-time or non-real-time applications. Post classification, they
are sent to the packet scheduler to schedule the packet delivery according to priority
levels. This helps the distributed ad hoc network offer both real-time and non-real-
time applications by granting real-time packets higher priority to capture the channel.

Network Layer: As described earlier, MANETs are self-configuring and have

dynamic topologies with several addressing and routing issues. In large networks, the
communication between nodes becomes unstable and the network itself becomes
prone to cyber-attacks. For reliable, error-free data transfers, various protocols
connect to enabling technologies to construct an end-to-end delivery mechanism
between a sender and one or more receivers. The routing protocol should have
provisions such as call admission or adaptive feedback to provide quality service. The
network layer provides IP addresses to hosts and establishes routes between senders
(sources) and receivers (destinations).For this, the network must map the dynamic
topology and overcome the challenging addressing and routing issues. The network
layer obtains the network resource information from lower layers and sends the
network status to the applications layer. The main objectives of the network layer are
to enhance the efficiency of the network by minimising overheads in routing, stability
and reliability in computing routes with a high convergence rate, Congestion control,
a better quality of service (QoS) and scalability by providing quality service in a large
network with a large number of devices. The important parameters are delay and
throughput.

Transport layer: Transmission control protocol (TCP) and user datagram protocol
(UDP) are two transport layer protocols widely used in wired networks. In TCP, a
connection must be established between communicating devices before the
transmission of data. After transmitting the data, the connection must close. On the
other hand, in UDP, there are no overheads for opening, maintaining and terminating
a connection. In TCP, the delivery of data at the destination is guaranteed and it has
extensive error checking mechanisms. UDP has a primary error control mechanism
and the delivery of data to the destination is not guaranteed. UDP is faster, simpler
and preferred for broadcast and has no congestion control mechanism to react to
network congestion. Due to this, the applications using UDP as the transport protocol
to transmit packets in MANETs can easily overwhelm the network with data. Such a
scenario may result in considerable power wastage and limited use of available
bandwidth in transmitting packets. TCP is comparatively slower and doesn't support
broadcasting but has an inherent congestion control mechanism (not clogging the
network and overloading the capacity in the routers). The standard TCP assumes that
the end-to-end congestion is due to increased flow in the transport pipe and/or reduced
available bandwidth. In wireless situation, congestion in TCP might be shown due to
frequent breaks in the end-to-end TCP connection. Hence, the standard TCP needs to
be modified and such modifications are available.

Applications Layer: The flexibility, ease of deployment and cost-effectiveness of

MANETs make them a highly attractive option for several application scenarios.
Based on the sensitivity to transmission delay, the applications are classified into real-
time and non-real-time. Online live movies, live sports, and video conferencing are a
few examples of real-time applications. On the other hand, applications such as Email
and FTP are non-real-time applications.
Applications need to be designed to handle frequent disconnection and reconnection,
time-varying delays and data packet losses. Initially deployed for the military
operation, MANETs today are widely used in disaster management, educational,
environmental, entertainment, law and order management and several location-aware
applications. Extensively developed new communication technologies such as WiFi,
6
Bluetooth, IEEE 802.11, WiMAX and HyperLAN have accelerated the deployment of Network Layer
MANETs across several application domains. In addition, various applications
developed for specific purposes form the application layer in MANET architecture.

4.6 MANET OPERATING PRINCIPLE

The nodes in an ad hoc network are initially unaware of the network's topology. They
eventually have to discover the topology. During the discovery process, every node
will learn about its neighbor nodes and the distance between them. This way, it also
lets the other nodes know about its existence in the network. For efficient routing,
routing tables are initialized and maintained by the routers using routing protocols and
these tables are stored in their memory. The routers use the routing protocols to decide
the path of the packet from the source node to the destination. Finding stable routes
decrease the route-related overhead and finding the shortest routes are the main goals
of the routing protocols.
 Fig. depicts a peer-to-peer multi-hop ad hoc network
 Mobile node A communicates directly with B (single hop) when a channel is
available
 If a channel is not available, then multi-hop communication is necessary, e.g.,
A->D->B
 For multi-hop communication to work, the intermediate nodes should route
the packet i.e., they should act as a router
 Example: For communication between A-C, B, or D & E, should serve as
routers

B C

A D E

Figure 4: Example of an Ad Hoc Network

4.7 CHALLENGES FACING AD HOC

NETWORKS

Regardless of the variety of mobile ad hoc network applications, there are still some
issues and design challenges that we have to overcome.

MANET being a wireless network, it inherits the traditional problem of wireless

networking:
 The channel is unprotected from the outside interfering signal.
 The wireless media is unreliable as compared to the wired media.
 The channel has time-varying and asymmetric propagation properties.

With these problems, there are some other challenges and complexities:

7
Emerging Networking
Technology Bandwidth Constraints: The bandwidth of wireless links is much lower than its
wired counterparts. For example, while several Gbps are possible in wired LAN, the
wireless LANs typically work around 2 Mbps (nowadays up to 50 Mbps).
Energy constraints: The power of the batteries in the devices is the limiting factor
that defines the operative time for the nodes.
High Latency: Nodes sleep or are idle when not in use to conserve energy. In data
exchanges involving sleeping nodes, the delay might be higher if the routing
algorithm needs to wake them up.
Transmission Errors: Attenuation and interferences are the effects of the wireless
link that increase the error rate.
Scalability Concerns: Scalability is a crucial aspect of MANET, mainly when used in
military communications. MANETs, when expanded, according to the need, each
node must be capable of handling the expansion or intensification of the network.
Fault Detection and Management: MANET is an infrastructure-less network with
decentralised administration. Faults become challenging to detect and manage as
every node can communicate with every other node. In addition, dynamic topology,
mobility and ever-changing routes add to the number of faults and data packet losses.
Security concerns: MANETs have more significant threats than the wired networks.

 Limited physical security: The physical sizes of the devices involved in these
networks are so small that they can easily be victims of theft.
 Cyber security: They are vulnerable to cyber-attacks and can be compromised
easily.
 Each node acts as a router and a receiver in the network, so each node plays
two roles becoming significantly more vulnerable.

Check Your Progress -1

1.1 Military vehicles on a battlefield with no existing infrastructure will deploy

............... network.

(A) MANET (B) Cell Network

(C) LAN (D) Wi-Fi

1.2 What is the type of network in which the topology change from time to time?

(A) Wi-Fi (B) Cell Network

(C) LAN (D) MANET

2. What is an ad hoc network?

--------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------
--------------------------

3. What are the challenging issues in ad hoc network maintenance ?

--------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------

8
4.Differentiate between cellular and ad hoc network in terms of cost Network Layer

effectiveness and bandwidth usage.

4.8 WIRELESS SENSOR NETWORKS (WSN)

As the name suggests ‘wireless sensor network’ is a network of sensor nodes. Now a
question may come to your mind ‘what are sensors’? Any hardware equipment which
has the capability of sensing the physical characteristics of its surroundings may be
called a sensor. To understand sensors, let’s compare these with our body sensors,
through which we sense our surroundings, smell by our nose, feel heat or cold by our
skin, taste, etc. Similarly, if these senses are measured by any hardware device, we
can say it is a sensor.
You must have seen the Fire Alarms installed in the buildings or in houses. Fire
alarms have a sensor that continuously measures smoke or temperature around them
and if the surrounding smoke or temperature increases by a certain limit, it starts
beeping/ alarming. Generally, these fire alarms have wired connections, or they can
have a wireless connection. So, can we put this fire alarm into the category of wireless
sensor network? The answer is ‘No’ because generally, the fire alarms are a
standalone system of sensors. It is not a network of sensors, and sensors are not
talking to each other. However, we can create a wireless sensor network for the
application of fire alarms in forests.
In the last decade, wireless sensor networks have gained popularity because of their
applications and benefits. Some of the popular applications of wireless sensor
networks are in Military, Agriculture, Security, and Healthcare. Let’s discuss some
examples of how wireless sensor networks have been changing the world around us.

4.9 APPLICATIONS OF WIRELESS SENSOR

NETWORKS (WSN)

There are various applications of wireless sensor networks that we can see happening
around us and many applications which are coming soon. In this section we will
discuss some of these applications.
Agriculture: The deployment of wireless sensor networks in the field of agriculture
has shown a lot of enthusiasm among the farmers. In agriculture, these networks are
highly useful in monitoring and controlling the temperature, humidity, soil
parameters, etc. So that farmers are getting better information and they can improve
their production. In livestock management, wireless sensor networks have shown
great relief to the farmers in monitoring the location and controlling the livestock and
also their biological parameters.
Military: At present, we can see many wireless sensor network applications in the
military. One of the most popular applications which many countries are currently
using is protecting their border from any illegal trespassing or any intrusion attack. To
do so, a network of sensors continuously observes the vibrations, sound, visuals,
positions and many other physical characteristics in the border areas. If these sensors
detect any suspicious activity, they can give this information to the nearby station or
9
Emerging Networking
Technology to the military officers. Similarly, the military can do effective surveillance, intrusion
detection, and collect intelligence with minimum risk on the battlefield in a better way
by using wireless sensor networks.
Environment: There are a large variety of applications of wireless sensor networks in
controlling and managing our environment.
● Air or Water Pollution: One application is monitoring the air pollution or
monitoring the water pollution. We can have a network of sensors monitoring
the pollution level in the air. If certain chemicals/air particles are increased
more than a limit in the air it can provide us the information that pollution has
increased in a particular location. Similarly, pollution management can be
effectively handled in the rivers and oceans by sensing appropriate
parameters.
● Flood and Landslide Warning: wireless sensor networks can be deployed on
the bank side of rivers or in possible flood areas. These networks can give us
a warning whenever the water level increases to a certain limit or water
crosses into certain areas. Based on these warnings residents can be informed
in a timely manner to evacuate the place. Similarly in the hilly areas, landslide
warnings can be issued so that we can help residents to take the appropriate
safety measures.
● Forest fire: As we discussed in the beginning regarding the fire alarm, we can
create a wireless sensor network in a forest that continuously measures the
temperature and the smoke level. Whenever there is a fire in some forest area
the forest authorities can be informed about the fire, and they can take the
appropriate measures to stop or the spread of fire.
Health: Wireless sensor networks are also useful in the health sector for effectively
monitoring the patients’ clinical data. We can provide various kinds of sensors to the
patients to continuously measure their health parameters like a heartbeat, ECG,
temperature, blood pressure, etc. These medical sensors keep recording the
information from the patient’s bodies and provide this information to a server. Server
can analyse the information and assist doctors to make suitable decisions. These
wireless sensor networks can be installed within the hospital where patients are
admitted, or wearable sensors can be used so that patients can do their daily routine
things and their clinical data can be recorded continuously.
Traffic: Another application of wireless sensor networks is in controlling and
managing traffic. Various kinds of sensors can be installed in vehicles that
continuously provide information to the base station so that the authorities can
observe the number of vehicles in a particular area. Whenever there is a possibility of
congestion in some location other vehicles can be informed or diverted to the other
routes to avoid congestion. These sensor networks are also used to manage the
parking lots and whenever there is any free parking area the vehicles can be directed
to that location. Similarly, these sensors can also communicate from vehicle to vehicle
so that any possibility of accidents is avoided.
Industries: Wireless sensor networks can be used within the industries to monitor the
functioning of machines and if there is any change in the parameters such as
temperature, pressure, or increase of chemicals then appropriate decisions can be
taken for the effective working of machines.
There are applications of WSN in structural monitoring, like monitoring bridges and
dams to reduce the human supervision and reduce the cost of monitoring, as well as
improve the efficiency of the processes. As wireless sensor networks will grow; we
will be able to see many more applications in the future.
10
 Network Layer
4.10 STRUCTURE OF WSN

In this section, we will discuss a typical structure of wireless sensor networks. This
structure will be helpful for you to understand the overall picture of wireless sensor
networks and their components. You will be able to understand how a wireless sensor
network is interconnected with the Internet or with the local network and how they are
able to perform some desired actions.

Figure 5: An example of WSN having two different sensing regions

The above figure 5 is showing an example of a wireless sensor network where we

have two sensing regions. Each sensing region has a network of wireless sensors
deployed in the region. All sensors in a sensing region are connected with each other
through wireless connections. These sensor nodes communicate with each and share
the data with others and also communicate and share the data with Base Node or Base
Station.
The purpose of base node is to give command to the sensor nodes to perform certain
tasks of sensing, based on the command given by the base node all sensors collaborate
with each other to send back the required data. Base node is also responsible for
accumulating the data from the sensors and passing it to the Internet or a specific
network. Further, this network or Internet is connected with different computing units,
which store, process and analyse the data coming from the base nodes and perform the
necessary action/decision.
Sensor Node
A sensor node is a basic component of wireless sensor networks. Each sensor node
has a power unit, sensing unit, communication unit and processing unit. These sensors
are radio frequency-based sensors which can transmit / receive data to/from a short

11
Emerging Networking
Technology distance. Hence, each sensor node also works as a relay to pass-on the data to other
nodes. Let’s see the structure iinside
nside a sensor in the following diagram.

Figure 6: An example of a Sensor node structure

There are different types of sensors available at present which can easily sense
physical characteristics around it like temperature, pressure, vibrations, sound, flow,
humidity, radiations, motion, position, light, etc. Also, many other types of sensors are
under development for various purposes which will be available for the desired
implementations.
As depicted in the abo
above
ve diagram, a sensor has a sensing unit which transfers its data
to the processing unit, this data is processed and stored by the processing unit and
further it can be communicated to the base station and/or to the nearby sensors. The
power unit is generally a battery-based power unit which provides the required power
to all these above units of a sensor.

4.11 CLASSIFICAITON OF WSNS

Wireless sensor networks are broadly classified in following types:

● PAN, LAN, MAN, WAN wireless sensor networks
● Stationary and Mobile wireless sensor networks (WSN)
● Single base and Multi-base Station
● Single Hop and Multi-Hop WSN
● Self-Organised and Non-Self-organised
● Homogeneous and Heterogeneous WSN

Let’s understand the differences and similarities between these WSNs

PAN, LAN, MAN, WAN wireless sensor networks
This classification of wireless sensor networks is based on the size of wireless sensor
network. You might have heard about these terminologies before. Let's refresh the
12
meaning of PAN, LAN, MAN and WAN. PAN refers to a personal area network, Network Layer
some typical examples are Bluetooth, and wireless USB. Any wireless sensor network
which is working in a very short area within a personal area network is a PAN
wireless sensor network. Similarly, depending on the size of our WSN we can classify
these WSN into LAN (local area network), MAN (metropolitan area network) and
WAN (wide area network) WSNs.
Stationary and Mobile WSN
Stationary sensor network is a network of stationary sensors placed in a particular
region and they communicate with others and pass the information to the nearby base
station from a fixed physical location. However, in Mobile wireless sensor networks
the sensors are not stationary and keep changing their physical location e.g. sensors
installed in the vehicles.
Single base and Multi-base Station
As we have seen earlier in our example of a wireless sensor network with two regions
that had a single base station for each sensing region which is giving command to
sensors and communicating with the network. However, in muti-base station scenario,
we are having two or more base stations to give commands to sensors in a sensing
region and to collect the information from these sensors.
Single Hop and Multi-Hop WSN
In a single hop wireless sensor network, all sensors are directly connected with the
base station and sensors are directly receiving command from the base station and
sending back the required information. However, when we have very large sensing
regions that time this one-to-one connection between base station and sensors is not
very effective and it may consume high energy. That's why, each sensing region is
further divided into different clusters and all sensors are also divided into clusters.
One sensor in a cluster will be treated as a cluster head or a supervisor sensor. Now, in
this arrangement the base station will communicate with the cluster heads and sensors
in a cluster will communicate with the cluster head. All cluster heads will send the
required information to the base station, not the sensor nodes. This kind of
arrangement of sensors with cluster heads to Base Station is called multi-hop WSN.
Self-Organised and Non-Self-organised
A non-self-organised wireless sensor network is a type of system where sensors are
not capable of organising themselves in a network and needs a control unit to
accomplish a task given by a base station. However, nowadays most of the wireless
sensor networks are self-organised networks, they can organise themselves in a
network and collaborate with each other to accomplish the task given by the base
station.
Homogeneous and Heterogeneous WSN
In our previous example where we discussed the components within a sensor. We
have seen that each sensor has sensing, processing, storage and communication
capabilities. If all sensors are having the same type of sensing, processing, storage and
communication capabilities we can say that as Homogeneous WSN. However, if all
sensors are not having the same sensing, processing, storage and communication
capabilities the WSN is heterogeneous. Some sensors have high capabilities, and some
are low. In that case, we will call that Heterogeneous WSN.

13
Emerging Networking
Technology
4.12 WSN TOPOLOGIES

In this section we will discuss different topologies of wireless sensor networks, which
means how these radio frequency sensors and base station(s) can be arranged as a
wireless network.
Star Topology WSN
This is one of the most common kinds of wireless sensor network, where all the
sensors are connected to a base station and this base station is commanding all the
sensors and receiving the data provided by the sensors.

Figure 7: Star Topology of WSN

one advantage of sta

starr topology is its simplicity to implement, second this arrangement
won’t add much power consumption on the sensors, third the communication delay
between sensors and base station is minimal. However, in terms of demerits- the base
station should be in the radio frequency range of all the sensors so that sensors can
send and receive data.
Mesh Topology WSN
In mesh topology all sensor nodes can communicate with each other, this kind of
topology is useful in multi-hop WSN. If one node wants to transmit a data to another
node which is out of its radio frequency range then one note can transfer the data to
the nearby node and that note can transfer data to the destination node, this way, we
can achieve a multi hop wireless sensor network. The main advantage of the system
it's that it can be deployed in a very large geographical location, furthermore, mesh
topology is more reliable because if some node is failed to transfer any information,
then other nodes can be used as a relay within a network. The main disadvantage of
this arrangement is the increased power consumption by the sensor nodes.

14
 Network Layer

Figure 8: Mesh Topology

Also, Mesh topology can be integrated with star topology and form a hybrid topology,
which can have advantage of both the topologies in Wireless sensor network.

Check Your Progress -2

1. What is a Wireless sensor network?

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
2. Write short notes on Sensor node.
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------

3. Differentiate between single hop and multi hop WSN.

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------

4. Name different topologies of WSN.

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------

5. Give an example of a heterogeneous WSN.

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------

4.13 IOT INTRODUCTION

IoT is a network of physical objects, called “Things”, embedded with hardware like -
sensors or actuators or software, for exchanging data with other devices over the
internet. With the help of this technology, it is possible to connect any kind of devices
15
Emerging Networking
Technology like simple household objects for example- kitchen appliances, baby monitors, ACs,
TVs, etc to other objects like- cars, traffic lights, web camera, etc. Connecting these
objects to the internet through embedded devices, allows seamless communication
between things, processes or people.

Some of the applications of IoT devices are – smart home voice assistant Alexa, smart
traffic light system. IoT devices when connected to cloud platforms can provide a
huge and wide variety of industrial or business applications. As the number of IoT
devices are increasing, the problem of storing, accessing and processing is also
emerging. IoT when used with Cloud technology provides solutions to these problems
due to huge infrastructure provided by the cloud providers.

Internet of Things (IoT), is also seen as an extension of the Machine to machine

(M2M) communication with some more capabilities like sensing, acting and to some
extent decision making. Here the large network of small devices communicates via
cloud or other networking platforms. It enables the users to communicate/ control a
large number of devices in a minimum time and to the maximum extent. Hence IoT as
a system comprises of smart objects having components like network, sensors,
power, processing power, data and services around it.

4.14 CHARACTERISTICS AND APPLICATIONS

OF IOT

Characteristics and Applications of IoT

The participating devices in IoT are required to have few characteristics like
communication, power, sensing & actuating (not necessarily) and data processing (to
some extent). Many of the devices also have some processing and decision making
capabilities and it depends on the requirement of the system. These all are
characterized in four different classes of fundamental characteristics namely
heterogeneity, dynamic, interconnected and scaling.

 Heterogeneity: Different type of device participates in the network, here hardware

platforms and communication networks also differ but at the end they all provide
a seamless connectivity of the thing.
 Dynamic: The smart things under monitoring change their states, orientations
very frequently. Sometimes the number of participating/ enabling devices is also
dynamic (with self configuring capability) and hence the entire network changes
on the go.
 Interconnection: The aim of the IoT is to make things interconnected to the outer/
digital world and make things accessible/ controllable from the remote location as
well. With IoT, theoretically, you can connect anything and everything which is
of some use for you.
 Scaling: As number of things around us is enormous in their connection and other
related devices would be more in number. This number increases with the
increase in the things of our interest. Hence the IoT devices/ network have to be
scalable in nature.

As we are now aware about the IoT as an intelligent connected network of smart
things/devices, we can see the following application of the IoT in our real life.

- Monitoring industrial processes, equipment health and product quality.

- Monitoring the health of smart buildings, critical infrastructure and vehicles.
16
 - Environment monitoring and data collection in extreme climatic conditions. Network Layer
- Patient monitoring at homes as well as in emergency vehicles, and OTs in
hospitals.
- In many other industries like automotive, aviation, retail, agriculture, defense
and space explorations.

Apart from the above mentioned scenarios, there are several other application of IoT
where a smart device with some power, sensing capability and communication
capability is required.

Now you have a basic understanding of the characteristics of the IoT devices and their
purpose of use. Let us now discuss few of the use cases where IoT can help users in
many ways. We have a number of use cases in various sectors of which a few are
mentioned here as examples.

4.14.1 IOT IN HOME AUTOMATION: INTRUSION

DETECTION USE CASE

Security of the home is a major a matter of concern to everyone especially when

nobody is at home. It is also important to ensure the security of children and elderly
members of the house. A lot of security solutions are available today from access
control to intrusion detection when an unauthorized person tries to enter in the house.
IoT has made these systems work smartly and effectively in a real time mode.

A typical home intrusion detection system makes use of smart cameras with night
vision and motion detection capabilities. Some of the intrusion detection systems also
use Passive Infrared (PIR) and door sensors to detect the motion and inform the users.
Many of such systems also send short clips or images of the intruders to the users in
very short/ real time. These systems can also be controlled using cloud based
platforms and also possess on-system memory to store some vital information. The
below given figure 9 represents the typical setup of the intrusion detection system for
a house.

Figure 9: Home automation, intrusion detection systems

Here in the figure, the door sensors or motion detection sensors are connected to the
gateway wirelessly. Sensors come with wireless connectivity like Bluetooth, WiFi or
17
Emerging Networking
Technology ZigBee. The smart camera system often comes either with WiFi or Ethernet
connectivity. The gateway shares the collected information to the cloud based
analytical engine which processes the real time information and makes relevant
decisions. The users get appropriate notifications/ alerts about the decision made by
the analytical engine. These engines can be trained/ configured for the user specific
requirements and hence provide security to the home.

4.14.2 IOT IN SMART CITIES: SMART LIGHTING

USE CASE

Lighting in homes, buildings, roads and parks is one of the major areas of energy
consumption. Majorly the electricity is produced using non renewable energy sources
worldwide. These include petroleum products like oil and gas, coal and other
resources. These resources are not in abundance and also contribute to environmental
pollution in many ways. The more demand of electricity increases the consumption of
these resources. Apart from some essential sectors, many sectors can contribute to
energy saving strategy to save electricity. IoT enabled devices and smart monitoring
of consumption helps in saving electricity. Today, we are focused on developing more
energy efficient green buildings to save our environment and lower carbon footprint.
The concept of green building itself incorporates the energy efficient and environment
friendly ecosystem. IoT enabled devices helps in scheduling the lighting and
controlling light source intensities depending upon the ambient conditions and
requirements and hence reduce energy wastage in these buildings. These devices can
also be controlled at the remote locations by the authorized operator/ persons. Sensors
on these devices also gather environmental data and device parameters. This helps in
better monitoring of the working of devices in these environments and improves the
device life. These devices also monitor the building’s structural health and monitor
indoor environment health. Smart street lighting, park lighting and condition
dependent lighting (like in fog or in storm) are some examples of IoT use case for
smart lighting.

4.14.3 IOT IN ENVIRONMENT MONITORING:

AIR POLLUTION USE CASE

Small size of devices and sensors with minimum battery/ power requirements are the
key advantages for any air quality monitoring sensor/ device. These can be placed at
desired places with ease and can monitor the air quality. A typical air pollution
monitoring IoT devices consists of a few sensors for monitoring CO, CO2, NO, SO2
and other poisonous gas levels in environment. These are also equipped with GPS
(Global Positioning System) system for location and wireless transmission of data to
the cloud. In the cloud, analysis of the information received is carried out and
outcomes are reported to the users/ authorities. Every year during the winter season
many Indian states face severe air pollution condition. Various factors contribute
towards it and regulatory policies are implemented. IoT devices have contributed it a
lot for air pollution monitoring. Using IoT devices, government collect all the
required data at one go from different places and governments can issue guideline for
the citizens accordingly. The constituents of the pollution can also indicate the
industry sector which is responsible for majority of the pollution at a particular place.
The authorities can take better and fast decisions to tackle the situation. On the other
hand, various air purifiers are being made to address the issue in a closed door
environment. This contribution of IoT devices can make authorities swiftly issue

18
guidelines for residents if higher levels of air pollutants are reported in the nearby Network Layer
areas.

4.14.4 IOT IN HEALTH: WEARABLE DEVICES

USE CASE

Health monitoring is major area where IoT is contributing. Many smart gadgets are
now available in market which monitor vital body parameters and logs them in the
personal devices. We are recording our pulse rate, oxygen saturation, ECG,
temperature and many other body parameters which are majorly non invasive in
nature using IoT devices. These devices are small, and battery powered. These devices
collect body parameters and store the data either at the local storage or in the cloud
where Artificial Intelligence based algorithms extract useful information which can be
shared with the doctors for seeking any medical advice. Various health devices are
now capable of making emergency communications to family members or health
officials/ doctors in any emergency situation. These devices are also useful in tracking
the school going children or elderly people who require extra safety and attention.

4.15 IOT SYSTEM ARCHITECTURE

IoT connects physical things to the network and hence various network devices are
involved. The physical things are made smart by placing some sensors on them to
sense their state, orientation or environment. These sensors require power modules to
operate and communication modules to send the sensed data. Communication can be
either in wired mode or in wireless mode. As an example we are here considering the
wireless mode only. This includes WiFi, Bluetooth, ZigBee, LORA and many other
technologies. A typical system architecture of IoT for physical world network is given
below in Fig.

Figure10: Typical IoT communication

The given figure is self explanatory and you can make out the communication of data
from the flow of information signals shown in the figure 10. The figure also shows the
communication via gateway, without gateway and device to device direct
communication.

19
Emerging Networking
Technology
There are many protocols that IoT devices use for communication and message
sharing on the network. A few of the IoT protocols are shown in the figure 11 given
below. These protocols work at different layers of the protocol stack like Application,
Transport, Network and Link layer. The devices which connect real world things to
the network are small and battery powered (most of the devices). These devices
communicate with each other or other networking infrastructure for data exchange
using various communication protocols. The below given figure shows the various
protocols used by IoT devices at different layers.

MQTT
XMPP

AMQP
HTTP

CoAP
DDS
NTP
Application Layer

TLS DTLS
ZigBee
Transport Layer
TCP ICMP UDP

IPSec
Network Layer
RPL DSR MPL OLSR AOD

6LoWPan
Physical/Data Link Layer RFID/ NFC Wi-Fi
IEEE 802.15.4 BLE

Figure11: IoT protocols

NOTE: The details of these protocols are given in a later course MCS-221. You are
advised to refer MCS-221 for further details.

Check Your Progress -3

1. What are things in IoT? What are characteristics of IoT ?.

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------

2. How IoTs communication differs from M2M communication?

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------

3. Write short notes in ZigBee, NFC and Bluetooth.

----------------------------------------------------------------------------------------------
----------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------

20
 Network Layer
4.16 Summary

In this unit we have learnt about different emerging technologies in networking

namely MANET, WSN and IOT. Details about the architectures, usage,
advantages and disadvantages are given. We have also learned about the
various scenarios of using these technologies in day to day life. In MANET,
characteristics and operating principles are described in detail. The structure of
wireless sensor networks with its types and classification is also mentioned.
WSN topologies like star and mesh are discussed in detail. AT the end, IoT
and its network architecture, Things and smart objects are mentioned as
integral part of the IoT network.

Solution to Problems

Answers to CYP-1

1.1 Answer: A
1.2 Answer: D

2. An ad hoc network is a multi-hop, infrastructure less network which has no centralized

server to control the communication between the nodes and resources cannot be reserved
beforehand. It is used in battlefields and military applications.

3. Following are the challenging issues in ad hoc network are

a) Medium access scheme
b) Routing for low delay
c) Multicast routing for one to many data transfer
d) Transport layer protocol for end to end reliable data transfer
e) Pricing Schemes
f) Quality of Service Provisioning
g) Self-Organization
h) Security
i) Addressing and Service Discovery
j) Energy Management
k) Scalability
l) Deployment considerations

4. The difference between cellular and ad hoc network is:-

21
Emerging Networking
Technology Criteria Cellular Network Ad Hoc Network
Cost high network Low network maintenance
Effectiveness maintenance cost cost due to self-organisation
in network
Bandwidth easy bandwidth Access needs complex
Usage reservation access control protocol
Bandwidth can be Shared radio channel hence
guaranteed variable bandwidth

Answers to CYP-2

1. A wireless sensor network is network of communicating devices called

nodes over wireless channels and having some sensing capability.
These nodes collect the information like temperature, humidity,
pressure, noise level etc. and then share it with the base node which
then transmits this information to the gateway.
2. A small inexpensive monitoring device with some sensing and
computing capability and limited battery to monitor a target
environment. The node collects the target information and shares with
other nodes wirelessly. In general a node consists of a transducer which
converts physical quantities to the voltage levels which are
communicated to other nodes/ devices to a relatively short distance.

3. The differences are as following

Single hop Multi hop

Directly connected with base station
Intermediate devices exists in
connection
Uses less energy (only when the Requires more energy
geographical distance is small)
Used in small sensing areas Used in large sensing regions

4. Different topologies in WSNs are Mesh, Star, Ring, Circular, Bus and
Grid topology.

Answers to CYP-3

1. Things in IoT are any physical thing which is of the interest or use to
the user. This may include a chair, a toaster, a water bottle etc. These
can be converted to smart objects by placing a sensor on them which
enables them to communicate and share data with other devices. The
characteristics of IoT are Heterogeneity, Dynamic, Interconnected and
scaling.

22
 Network Layer

2. Difference of IoT and M2M communication is as following

M2M – Machine 2 Machine IoT – Internet of Things

Point to point connection Devices are connected through the
establishment network and also supports
connecting to global cloud
networks.
Limited amount of intelligence Decision making is enabled
Makes use of internet protocols Makes use of traditional
like- HTTP, FTP, etc. communication protocols
Generally may not rely on Generally Rely on internet
internet connection connection
Less scalable Highly scalable

3. Short notes on the given topics.

ZigBee
It is a wireless technology based on IEEE 802.15.4 used to address needs
of low-power and low-cost IoT devices. It is used to create low cost, low
power, low data rate wireless ad-hoc networks. It is resistant to
unauthorized reading and communication errors but provides low
throughput. It is easy to install, implement and supports a large number of
nodes to be connected. It can be used for short range communications
only.

NFC
Near Field Communication (NFC) is a protocol used for short distance
communication between devices. It is based on RFID technology but has a
lower transmission range (of about 10 cm). It is used for identification of
documents or objects. It allows contact less transmission of data. It has
shorter setup time than Bluetooth and provides better security.

Bluetooth
It is one of the widely used types of wireless PAN used for short range
transmission of data. It makes use of short range radio frequency. It
provides data rate of appx 2.1 Mbps and operates at 2.45GHz. It is capable
of low cost and low power transmission for short distances. Its initial
version 1.0 supported upto 732kpbs speed. Its latest version is 5.2 which
can work upto 400m range with 2 Mbps data rate.

***

23