You are on page 1of 14

CHAPTER 4

PERFORMANCE RESULTS FOR RANDOMIZED SIMULATIONS

4.1. Performance Metrics for Ad hoc Network


These metrics are interesting because they can be used to point out what really
happened during the simulation and provide valuable information about the routing
protocol. In the following sections some metrics of this type are described.

4.1.1. Packet Delivery Ratio


The packet delivery ratio presents the ratio between the number of packets
sent from the application layer and the number of packets actually received at the
destination nodes.
It is desirable that a routing protocol keeps this rate at a high level since
efficient bandwidth utilization is important in wireless networks where available
bandwidth is a limiting factor. This is an important metric because it reveals the loss
rate seen by the transport protocols and also characterizes the completeness and
correctness of the routing protocol. The packet delivery ratio is logically the same
with the average throughput. The high packet delivery ratio means low packet drop
and high throughput and traffic ratio.

4.1.2. Routing Overhead


Routing overhead is of course an interesting metric. In some way it reveals
how bandwidth efficient the routing protocol is. The routing overhead metric simply
shows how much of the bandwidth (which often is one of the limiting factors in a
wireless system) that is consumed by the routing messages, i.e., the amount of
bandwidth available to the data packets.
An interesting observation is that for all protocols there is a theoretical limit
where some properties of the scenario effect the data rate down to zero because all the
bandwidth is used for routing messages. The ideal case is naturally no overhead at all
i.e., only data packets traverse the network. An ideal routing protocol can be imple-
53

ented in a simulator but a routing protocol without routing messages is a contradiction


and can not be implemented in a real network. The routing overhead is typically much
larger for a proactive protocol because it periodically floods the network updating
messages.
As mobility in the network increases reactive protocols will of course have to
send more routing messages too. This is where the real strength or weaknesses of the
routing protocol can be revealed. On the other hand in DSR another type of overhead
presents itself even though it is easily overlooked in the previously described packet
delivery ratio metric.
DSR works by finding source routes to the destination on-demand. By storing
information about all intermediate nodes in the packet header as the route discovery
packet traverses the network it knows the full route once the route discovery packet
returns. These source routes cause the packet headers to grow and hence produce
more routing overhead.
Considering this, the traditional metric, packets sent versus packets delivered,
might give the impression that DSR is able to deliver more packets than other
protocols. Looking at the ratio payload bytes sent versus payload bytes received
instead could result in a different performance for DSR. This would be most obvious
in a network with long routes.

4.1.3. Average Throughput


Since the available bandwidth in a network is fairly well known, it is
interesting to see what the actual throughput achieved in a simulation is. If a good
estimation of this value can be extracted it would be possible to see how efficient the
routing protocol is. The higher the average throughput is, the less routing overhead
consuming the bandwidth.

4.1.4. Path Optimality


Traditionally this measurement compares the optimal path usually defined as
the shortest path between two nodes in the simulator at the sending moment with the
length of the path that the packet actually traveled. If the average actual path length is
close to the shortest path, the protocol is said to be good. However, it is hard to know
what the actual optimal path is.
54

4.2. Ad hoc Mobility Models


There are two main issues when doing simulations except for modeling of the
physical environment. The first one is the modeling of how the nodes move in the
mobile Ad hoc network. This is very tricky to know how the nodes will move in the
real world. Ad hoc mobility models differ from cellular mobility models in the
network they model. Cellular mobility models require the used of based stations (BSs)
whereas Ad hoc mobility models require the cooperation of two or more
communicating mobile nodes.
There are a number of suggestions how to do this and some are found in this
section. The other issue is that in order to compare different networks and describe
their behavior needing some kind of metrics. Trying to evaluate different routing
protocols, the mobility may be wanted to know, i.e. how hard it is to handle routing in
the network. It is a popular term and in this case it merely describes the activity of
link-changes caused by external physical interference.
In this section, five Ad hoc mobility models are described in detail:
 Random Mobility Model: A simple mobility model based on random
directions and speeds.
 Random Waypoint Mobility Model: A model that includes pause times
between changes in destination and speed.
 Random Direction Mobility Model: A model that forces mobile nodes to
travel to the edge of the simulation area before changing direction and
speed.
 A Probabilistic Version of the Random Mobility Model: A model that
utilizes a probability matrix to determine the next position of an mobile
node.
 City Area, Area Zone, and Street Unit Mobility Models: Three models
describing simulation areas representing different granular scales of a city.

4.2.1. Random Mobility Model


The Random Mobility Model for Ad hoc networks is the Random Walk
Mobility Model for cellular networks. In the Random Mobility Model, the current
speed and direction of a mobile node is independent of its past speed and direction.
Thus, an unrealistic generation of movements is encountered such as sudden stopping,
55

sharp turning, and completely random wandering. In order to avoid these problems,
many authors modify the Random Mobility Model by changing the calculation of
speed, direction, or both.

4.2.2. Random Waypoint Mobility Model


The Random Waypoint Mobility Model includes pause times between changes
in direction and/or speed. A mobile node begins by staying in one location for a
certain period (i.e., a pause time). Once this time expires, the mobile node chooses a
random destination as well as a speed that is uniformly distributed between [0,
maxspeed]. It then travels towards the newly chosen destination at the selected speed.
Upon arrival, the mobile node takes another break before starting the process again.
Many authors have adopted this model in their simulation studies. Being noted that
the movement pattern of an mobile node using the Random Waypoint Mobility Model
is similar to the Random Walk Mobility Model if pause time is zero and [0,
maxspeed] = [speedmin, speedmax].

4.2.3. Random Direction Mobility Model


The Random Direction Mobility Model was created in order to overcome a
flaw discovered in the Random Waypoint Mobility Model. Mobile nodes using the
Random Waypoint Mobility Model often choose new destinations, and the probability
of choosing a new destination that is located in the center of the simulation area, or
requires travel through the middle of the simulation area, is high.

Figure 4.1. Traveling Pattern of a Mobile Node using the Random Waypoint Mobility
Model on 500m×500m Area [98Ton]
56

This trend is illustrated in Figure 4.5, being stated that mobile node moving
with the Random Waypoint Mobility Model appear to converge, disperse, converge
again, etc. In order to alleviate this type of behavior and promote a semi-constant
number of neighbors, the Random Direction Mobility Model was developed.

4.3. Simulation Setup


This section will be described how the simulations were done. Two different
types of simulations have been done:
 Preliminary simulations: A randomized mobility model was built to test
how the proposed protocols work and to find an intelligent way to decide
which simulation parameters to vary and which to keep constant.
 Realistic simulations: A new realistic mobility model was modeled and
then the proposed routing protocols were simulated based on this realistic
scenario. This simulation does not provide a general view of the protocols,
but instead tests certain characteristics of the protocols.

In all simulations, except the realistic scenario simulations, a randomized


scenario has been used. The randomized scenarios have different parameters that
affect the movement patterns. The parameters that can be changed are:
 Maximum speed: Every time a speed is going to be randomized, it is
randomized in the interval [0, maximum speed].
 Number of nodes: This was constant during the simulations.
 Environment size: Determines the size of the environment.
 Simulation time: The time for which the simulations will be run at.
 Pause time: Pause time is the time for which a node stands still before
randomizing a new destination and the speed will be used to reach this
destination.
Each node selects a random destination, a waypoint somewhere in the
environment space. Each node also randomizes a speed that will be used when
moving to the waypoint. This speed is randomized uniformly in the interval 0 to
maximum speed. Every time a node reaches a waypoint, this procedure will be
repeated. A factor that has not been taken into consideration with the scenarios is the
fact that a real person is not likely to stand on the same place if the connection goes
down. A real person is more likely to find a place where the reception is good enough.
57

The metrics has to be measured that describes the characteristics behavior of Ad hoc
network. There are many simulation parameters that need to be varied in order to
perform exhaustive simulations. In this chapter the simulation parameters used to
produce the simulation suite for this work are presented and explained. Table 4.1
shows a summary of the parameters for randomized simulations

Table 4.1. Simulation Parameters for Randomized Simulations


Parameter Value
Routing Protocols AODV, DSR
Interface queue length 50 packets
Simulation time 150s
Number of nodes 3 ,5,7 and 10
Channel Type Wireless Channel
Antenna model Omni Direction
Mobility model Random way point
Simulation Area 500mx400m

4.4. Simulation Experiments


Network Simulator (NS) has been tested to build and validate under Windows
XP using Cygwin. Cygwin provides a Linux-like environment under Windows
because the primary NS build platforms are various flavors of Unix. The screenshots
of the network simulator are shown as following Figures. Cygwin can also be used for
generating sent packets, received packets and forward packets for all simulations.
These data are very useful for performance evaluation of routing protocols.

Figure 4.2. Screenshot of NS-2.28


58

Figure 4.3. Screenshot of NS to Trace Analysis

4.5. Simulation Results


Network Simulator (NS-2) is simply an event driven simulation tool that has
proved useful in studying the dynamic nature of communication network. A
simulation study was carried out to evaluate the performance of Ad hoc network
routing protocols such as AODV and DSR changing the network scenarios in number
of nodes three to ten. It based on the metrics packet delivery ratio, routing overhead
and average throughput.

4.5.1. Packet Delivery Ratio


The packet delivery ratio of routing protocols AODV and DSR does not
depend on number of nodes. The results for packet delivery ratio of randomized
scenario are shown in following Figure 4.4 and Figure 4.5 respectively.
100

99

98
PacketDeliveryRatio (%)

97

96

95
AODV
94
DSR
93

92

91

90
50 100 150 200 250 300 350 400

Simulation
Simulation TimeTime(second)
Figure 4.4. Packet Delivery Ratio for Three Nodes
59

100
AODV
DSR
99.8

99.6

Packet Delivery Ratio (%)


99.4

99.2

99

98.8

98.6
50 100 150 200 250 300 350 400

Simulation
Simulation TimeTime
(second)
Figure 4.5. Packet Delivery Ratio for Ten Nodes

When the simulation starts to run, every node moves with very low speed that
means low mobility and then gradually speed up that is high mobility. It shows how
many of the sent data packets are actually received and why have the dropped packets
been dropped. The receive packets for the AODV version is large even for high
mobility. The value is almost constant and very close to one. Therefore, considering
the metric delivery ratio, AODV and DSR performs similarly. Because the optimal
value of delivery ratio is one, which means no packet loss for AODV and DSR.
Although these protocols may dropped packets for these simulation, their network
communication scenarios are too large to simulate. The main reasons for dropping
packets are that the protocols are sending packets on a broken route that packets in the
buffers are dropped because of congestion and timeouts. As packet delivery ratio is
logically same with average throughput, the performance of the realistic simulations is
evaluated by means of average throughput.

100

99.8
Packet Delivery Ratio (%)

99.6

99.4

99.2

3 nodes
99
5 nodes
7 nodes
98.8
10 nodes

98.6
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.6. Packet Delivery Ratio in AODV
60

100

99

98

Packet Delivery Ratio (%)


97

96

95

94

93
3 nodes
92
5 nodes
7 nodes
91
10 nodes
90
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.7. Packet Delivery Ratio in DSR

4.5.2. Routing Overhead


In this section, the routing overhead (byte) means that the total number of
routing packets divided by total number of delivered data packets. So, the average
number of routing packets in bytes needed to deliver a single data packet. This is
needed because the size of routing packets may vary.
Because the routing protocol need to send control information to achieve the
task of finding routes, it is interesting to see how much control information that is
actually sent for each protocol. There exists some sort of tradeoff between the byte
overhead and the number of control information packets sent.
The simulation shows that the byte overhead of AODV is much lager than
DSR even at low data rates. The reason for larger byte overhead for DSR is of course
the source route in each packet in taking the time to route destination of large
scenarios.
4
x 10
3
Packet no:

AODV
2.5
DSR

2
Pkts number
Routing Overhead

1.5

0.5

0
Overhead

Figure 4.8. Routing Overhead


61

In following Figures, they represent each routing protocols researching in


number of nodes for overhead. DSR is varying from minimum number of nodes to
maximum number of nodes good performing in routing overhead. Moreover, the
routing overhead increases dramatically after speeding of node speed. DSR
maintained the lowest routing overhead at that speed. If that speed is over 10m/s DSR
produces higher overhead than AODV. The reason behind this dramatic increase is
that the route cache property is useless when mobile nodes are moving at higher
speeds and links are lost more frequently.
4
x 10
4

3.5

3 nodes
Routing Overhead

2.5
5 nodes
2
7 node
10 nodes
1.5

0.5

0
1

AODV Routing Protocol

Figure 4.9. Overhead in AODV


4
x 10
3

2.5
3 nodes
5 nodes
7 nodes
Routing Overhead

10 nodes
1.5

0.5

0
1

DSR Routing Protocol

Figure 4.10. Overhead in DSR

4.5.3. Average Throughput


The average throughput curves for the different protocols with a packet size of
512 bytes (or larger) are shown in Figure 4.11 and Figure 4.12. It must be mentioned
that the curves in this case are only interesting from relative view, as a comparison
62

between the protocols. It has been tried to determine the relative different in average
throughput for these two protocols with respect to the changing number of nodes and
specific load that having used.
4
x 10
6

AODV
5
DSR

Average Throughput (%)


4

0
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.11. Throughput Comparison of Three Nodes
4
x 10
2.5
AODV
DSR
2
Average Throughput (%)

1.5

0.5

0
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.12. Throughput Comparison of Ten Nodes

The average throughput curves for all protocols are very similar to the packet
delivery ratio curves. This is logical because large packet drops will of course mean
lower throughput. DSR has higher throughput in less node density. But as the longer
the simulation time or the higher paths to route, AODV has more route. And then
AODV has performed communication of realistic large area. A large network often
means longer routes and longer routes mean that the source overhead in each packet
grows.
4
x 10
2.5
A v erage T hroug hput (% )

1.5 3 nodes
5 nodes
7 nodes
1 10 nodes

0.5

0
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.13. Throughput Comparison of Nodes for AODV
63

4
x 10
6

5
3 nodes

A verage Throughput (% )
5 nodes
4 7 nodes
10 nodes

0
50 100 150 200 250 300 350 400

Simulation Time
Simulation Time (second)
Figure 4.14. Throughput Comparison of Nodes for DSR

4.5.4. Traffic Ratio


The comparison of sent packets and received packets for these protocols are
shown in Figure 4.15 and Figure 4.16.
Traffic Data
12000
AODV
10000
DSR
Number of Packet

8000

6000

4000

2000

0
Generate Recive

Figure 4.15. Traffic Data (for three nodes)

The above Figure shows the generated packet and receive of packets in
number of three nodes. As the same traffic pattern was used in all simulations, the
sending packets of three protocols are the same in some simulations and some are
different according to the simulation time. But received packets and forward packets
are different for all simulations according to the behavior of the protocols.
Traffic Data
15000
AODV
12500 DSR
10000
Packet Number

7500

5000

2500

0
Generate Receive

Figure 4.16. Traffic Data (for ten nodes)


64

As the average throughputs of these protocols, DSR has high throughput of


188.51% and AODV has 55.22 % in node numbers ten setting. As Table.4.2, the two
protocols that have the high traffic send/receive ratio are DSDV and DSR that entirely
relies on topology and traffic load. Although AODV is rather less throughput than
DSR, it has to be seamed in this random mobility model of scenario. And then,
although DSR has a little good in average throughput, it must however be noted that
the traffic load in this scenario is very high and most of the communication is taking
place in a large area, thus leading to a very low average throughput and traffic ratio.
In summarizing with these protocols, these numeric numbers represent that 1 for best,
2 for good and 3 for bad.

Table 4.2. Numerical Comparison of two Routing Protocols


Metrics AODV DSR
Delivery ratio 2 2
Overhead 3 2
Throughput 3 1
Generate Packets 2 1
Receive Packets 2 1

4.6. Summary
The Random mobility scenario shows how protocols behave in certain
circumstance of a situation. For this reason, these scenarios were designed and
simulated with different versions of the protocols DSR and AODV. DSR show the
best performance results overall. And another good candidate is AODV with having a
slightly higher packet overhead in different network density.
The DSR protocol is however based on source routing, which means that the
byte overhead in each packet can affect the total byte overhead in the network quite
drastically when the offered load to the network and the size of the network increases.
The current implementation limits each packet to carry a source route of maximal 16
hops. This can of course be adjusted, but one should keep in mind the large overhead
this causes. Then this implementation would not manage to route to all nodes.
This protocol AODV needs better link breakage detection. Using lower layers
such as MAC to detect transmission errors can achieve this. This protocol is a definite
choice for highly mobile networks. This protocol is highly optimized and also shows
good results in all simulations. As Table 4.2, AODV has less throughput, but the
65

packet delivery ratio is in optimization one. However, AODV protocol may good
performance in a realistic life scenario. This protocol could definitely be used in
highly mobile or dynamic networks as well as static networks.

You might also like