Mobile Ad Hoc Networks (MANET)

Abstract

Mobile Ad Hoc Network (MANET) is a wireless network capable of autonomous operation. MANET routing has no fixed base station and hence nodes in the network are mobile and self configuring. A MANET is characterized by multi hop routing so that nodes are not connected to layer 2 but can communicate through layer 3 routing. In MANET every node is a potential router and the topologies are dynamic due to node mobility.

This paper presents a comprehensive study of four MANET routing protocols ADOV (Ad Hoc On demand Distance Vector), DSR (Dynamic Source Routing), OLSR (Optimized Link State Routing) and TORA (Temporally Ordered Routing Algorithm). For experimental purpose six scenarios have been considered. These 6 scenarios are generated with 3 different traffic parameters namely high resolution video, light HTTP and high FTP load. This traffic was passed individually on to each scenario on 2 different node setup (20, 100 nodes). Finally, graphical evaluation of each protocol was based on their performances which are calculated on the basis of performance metrics used which are End to end delay, Network load and Throughput.

Chapter 1: Introduction

This chapter demonstrates a brief overview of the project. Now days in our day to day life we see so many changes around the world accelerating with respect to technology. In this technical world Internet has brought revolution in communication media. Communication is said to the method of information exchange either between two people or between two end users (in terms of computers). Communication is said to be done when a file or a message is passed over the medium among two people. Here the usage of internet comes into existence. Internet can be used as shared medium for information interchange between users around the globe. This information can be of any type:

• Short Message
• Mail
• File transfers
• Video
• Audio etc…

When these types of information are passed among internet, it uses either the wired media (for connecting users locally) or wireless media (for connecting locally or globally). When this information exchange is done the packets carry that information through the protocols available for the respective media to deliver the packet to the destination. This is the process of information exchange. In this paper, similar work is experimented over MANET (Mobile Ad Hoc Networks) [chapter 2] for information exchange using four different protocols. This setup was done internally over a virtual setup of networks using OPNET modeller 14.5. So the investigation of protocols is done based on the performance of each protocol. In the scenarios, each protocol is passed with 3 different types of traffics namely: High FTP load, High Resolution Video and Light HTTP Browsing. This thesis concludes the performance of four routing protocols towards through the end.

Keywords: MANET, AODV, OLSR, TORA, DSR, OPNET Modeler 14.5.

1.1 Organisation of Thesis

This thesis gives an overview of how the project is organised with respect to each chapter. The very first chapter is Introduction of the Thesis, which shows a project idea and the tasks to achieve the goals.

Second chapter Literature Survey which contains the studies and work which has been done previously by other authors on the related topics with the parameters used by them for their work. The scenarios used for my work are related somewhere by overall different results are taken with different parameters used. Again the classification and the background work of MANETs are explained following the comparison of MANET routing protocols.

The parameters and scenarios used for my work are explained briefly with the simulation setup in chapter 3, Implementation.

The results of the simulations and experiments performed are explained in chapter 4, Performance Evaluation and Design. The results are taken graphically which helps in comparing the results for routing protocols against the performance metrics used to investigate the performance. The graphs are explained briefly in chapter 4.

The results gathered with the help of scenarios in chapter 3 are concluded with the future work on the routing protocols in chapter 5, Conclusion and Future Work.

Finally the supported documents for the project which helped in making the simulation and project start up is kept in Appendix A. And finally chapter 6 shows all the sources and references used for the thesis to help in achieving all the supported information and work.

1.2 Objective of Work

The main objective of this work is to study the different routing protocols, which are developed for Mobile Ad hoc networks (MANETs), and to compare the different routing protocols by using simulation tool OPNET Modeler 14.5. As a part of the work four major routing protocols OLSR, AODV, DSR and TORA have been selected and carried out the simulations for comparing the performance of these protocols. Three performance metrics delay; Network Load and Throughput are used to compare the performance of the routing protocols.

Objective 1

To get a clear understanding and functioning of different routing protocols for Mobile Ad hoc networks. This objective could be achieved by reading and understanding the various papers available on routing protocols of Mobile Ad hoc networks.

Objective 2

To do a literature survey of previous work done on MANET Routing Protocols so that something different and efficient simulation could be introduced. For this objective different articles and papers published has been studied thoroughly and analysed from websites, books and all relevant resources available. By going through literature survey on routing protocols, the desired simulation environment and setup have been introduced with different simulation parameters. Literature Survey is explained in chapter 2.

Objective 3

Conducting an experiment and collecting the output data:

This objective has been achieved by designing the appropriate networks with the appropriate simulation parameters and running a simulation for different protocols for different performance metrics. After conducting the simulation the output data have been collected graphically. The results are shown graphically and explained in the project in chapter 4.

Objective 4

Analyzing the output data and ending up with summary and the conclusion:

The results have been studied and explained in chapter 4. After studying the simulation results conclusion has been made with some future work which can be done further. Conclusion of the project and future work is explained in chapter 5.

Chapter 2: Literature Survey

2.1 Introduction

This chapter gives a brief introduction about the work done in previous related papers and reports related to this project. In addition, the introduction to the routing protocols and their classification with respect to routing is demonstrated. This chapter is a short review of the previous work done and the additional objectives regarding routing protocols. The four MANET routing protocols are selected to evaluate the performances; they are OLSR, AODV, DSR and TORA. The further implementations of these protocols are explained later in this thesis (Chapter 3). MANET is an infrastructure less network which provides the freedom to the nodes to be free to move anywhere in the network (Stefano Basagni, 2004).

2.2 Related Work on Routing Protocols:

As we see from (Mahmoud), analysis of two reactive protocols DSR and TORA were done using OPNET Modeller. In their simulation scenario, they used 50 wireless devices in each scenario with constant traffic parameters and protocol specification settings. They have passed FTP traffic of 1000 bytes making it constant throughout the simulation. Evaluation of these protocols was done based on the performance metrics used which are: Delay, Data dropped, Throughput and media access delay. Their simulation results shows that DSR performance was better than TORA in terms of throughput as TORA produces less throughput due to additional overhead used for path creation and path maintenance. At the same time TORA minimizes communication overhead by localization which results in less delay when compared to DSR as there is no such mechanism in DSR.

Similarly, from (Zukarnain, 2009) they have done evaluation of MANET routing protocol AODV in order to establish the connection between the nodes since the mobile node can change their topology frequently. So their study was carried on different mobile node movement pattern which are: Random Waypoint Mobility Model, Random Walk Mobility Model and Random Direction Mobility Model. In order to evaluate the performance of the protocol with these node movement patterns, performance metrics used were Routing overhead, Throughput and Packet delivery ratio. They used 2 simulation scenarios where in the first scenario comparison was made on different mobility model varying different number of nodes 5, 10, 15, 20, 25 with fixed speed of 15m/s and in other case comparison was made to evaluate the protocol on different mobility models with varying speeds 5, 10, 15, 20 m/s and 50 nodes as constant all through. They concluded that Random waypoint model is best for AODV compared to other mobility models as the protocol produces highest throughput than compared to others.

Other related work from (Nyirenda, 2009) gives the similar working of MANET routing protocols AODV, OLSR, DSR and TORA on OPNET 14.5. Their work carried out for study the performance of the four protocols by different performance metrics which are: Network load, Packet delivery ratio, Packet end-to-end delay and Throughput. The simulation setup which was implemented on 6 different scenarios where the traffic passing on to the network was ranging from low to high network load, nodes changing from 5, 20, 50 and with speed ranging from 10 m/s to 28 m/s. Mobility model used in this was Random waypoint model for mobility pattern. They conclusions included in their work showed that OLSR performed better compared to the other protocols as it has a bad routing overhead and hence it is well suited for large and complex networks. So apart from routing overhead OLSR performed better but with routing overhead DSR is better. But when in small network AODV is much better compared to other protocols. So finally, proactive protocols perform well in high capacity links whereas reactive protocols perform better in low capacity networks.

From another paper (Maltz, 2001), we see that working on demand routing protocols in MANET was done where DSR protocol was compared to other on demand protocols like AODV, TORA and DSDV. Simulation setup carried is represented in tables below:

With the above simulation setup the protocols were evaluated based on the performance metrics used which was Packet delivery ratio, Routing overhead, Path optimality and lower speed of nodes. The conclusions showed that DSR performed well when referred to packet loss rate and routing over head is concerned. Of all careful implementation done with all the above parameters tested on the experiment test beds DSR out performed in every case scenario used when compared all the other protocols used in this setup.

2.3 Background Work

A network can be said as association of different systems or organizations where sharing the information can be done collectively. Whereas in computing terms it can be simply defined as a group of computers connected together logically to share information like printing, faxing, etc… The network can be divided into 2 types based on their working behaviour. They are:

Infrastructure network: These networks are used where the topology is said to be limited and there is a fixed point like base station (generally referred as router) to transmit signals and the end points which are connected to base station communicate with each other devices on the network switching from one base station to another. When a node moves out of range within its network, then it is said to be connected to another base station range where this process is referred as handoff. We can often see this type of mechanisms in infrastructure and fixed or wired networks.

Infrastructure less network: These networks are a typical type of networks where there are no such fixed nodes or topologies on the network and the end pints or devices on the network are free to communicate with each other devices on the network as each device on the network behave themselves as routers and encourages communication process all over the network. By this we can say that wireless technology is a promising technology that can tolerate the information exchange worldwide. In the last decade we can also see the constant increase in the growth wireless technology issues, one of which being mobile devices such as laptops, cell phones, PDA’s etc… Ad hoc networks are also the part of this type of network.

Mobile Ad hoc network is a new technology emerged with the hypothesis of wireless networks. These networks are very typical and do not use any fixed infrastructure for communication process. The nodes connected in these networks are wireless links which are mobile in nature and communicate with each other mobile node in the network with radio transmission and topology is said to establish by the intermediate nodes on the network which are helpful for communication process. As there are no fixed infrastructure and limited topology constraints the nodes on the network are free to join and leave the network and this is possible as the nodes on the network are mobile. Due to this random movement of the nodes in the network the topology of the network changes dynamically. Due to this change protocol must be able to acclimatize with these movements and are also responsible to maintain the routes of the information travelled without disturbing network connectivity.

This concept of ad hoc network allows each node on its network to act like router, resulting in the flow of information exchange with multi hop routing. These types of networks are widely used in military and other rescue applications. (Saadawi, 2003)

2.4 Routing

Routing is the process of moving the data from one place of the network to another. The one end should be the source which intends to transfer the data to the destination (other end). The concept of routing has been there since 1970s but it has caught the move in 1980s. In 1970s networks were simple and today there are large scale networks in existence. To move the data from one host to another at least one router is required in the network. Router has all the information regarding hosts in the network and it can manipulate the best possible route as well. (Javvin, 2004)

Routing takes place at Layer 3 in OSI 7 layer Model. Some of the protocols at layer 3 are IP and IS-IS. These protocols carry the data between source and destination along with their addresses in the data packet. Routing involves two basic activities; Path Determination and Switching. These two activities are capable to determine the optical routing path and to transfer the data packets through internetworking. This is called Switching. Switching can still be easy where as determining the reliable path is difficult.

Path Determination

Routing make use of the different routing protocols to determine the best possible path to deliver the data. Routing protocols use the metric to calculate which the best possible path to accomplish the task is. The metric used could be path bandwidth, path length, delay etc. to determine the optimal path. To determine the path in the network, routing algorithms maintains the routing tables which are used to store the routing information about the network. Routers communicate with each other in order to maintain their routing tables which make communication easier and faster. Routing information can vary according to the routing algorithm used for the process. (Cisco, 2010)

Switching

Apart from path determination, router shows one more activity; Packet Switching. Switching simply means forwarding the required data from one interface to the another in order to reach the destination. In this case, the data to be forwarded is packet. The next interface address is decided by using the destination address present in the packet. Though the nodes know the next hop address, they still need to know how to use it. So for that reason they use routing tables. The routing table throws the packet away when the destination is unknown. But when the destination is known routing table posses all the interface information forming the route to the destination. (2006)

The next hop address could be another host or either it could be a router. According to ISO developed hierarchical terminology, the systems which have the capability to forward the data from the source to the destination are called Intermediate Systems (IS) and the systems which fails to do so are called as End systems (ES). When it is a router it follows the same procedure as this one, and if it is a host it simply forwards the packet. In packet switching, the Layer 3 address of the source and the destination remains the same to authenticate the original sender and the receiver. However the Layer 2 address (MAC) changes from router to router and from router to host which at last determines the destination host. (Cisco, 2010)

2.5 Ad hoc Networks

An ad hoc network is a collection of many autonomous nodes connected together by radio waves and maintain the connectivity in decentralised manner. Wireless ad hoc networks are connected through wireless links so there is no need of any physical medium and hence contend of the medium is always there such as noise, interference and fading. On ad hoc networks each node functions as a node and a router itself. It simply means there is no need of an additional switch or a router to route the data and the control is given to the nodes themselves. (Yang Li, 2010)

2.6 Mobile Ad hoc Networks

A Mobile Ad hoc Network is an autonomous network formed by hundreds or thousands of nodes. These networks don’t need any infrastructure as they can act as a node and router itself. They are free to move anywhere in the network and are connected through radio links. The only limitation with the MANETs is that they can communicate in a particular rage of the radio waves and outside the network they need some additional arrangements to communicate. To overcome this limitation the Intermediate Node or sometimes called as Relays are brought into picture. Intermediate nodes help in forwarding the packets to the destination. In MANET nodes are free to travel anywhere in the network, hence network topology cannot be fixed for such networks and it keeps changing depending on the current location of the nodes. (ANTDS, 2001)

The fundamental differences between the wired networks and MANET are:

• Asymmetric Links
• Redundant Link
• Interference
• Dynamic Topology

2.7 Characteristics of MANETs

There are several MANET routing protocols which should handle the several inherent characteristics of MANETs as follows as mentioned in (Subbarao)(Jain, 2005):

• Dynamic Topologies: Since the MANET is infrastructure less ad hoc networks, the nodes are free to move arbitrarily. The mobility of nodes may be random and so unpredictable. So the links between the nodes may be unidirectional or bidirectional at times.
• Bandwidth Constrained, Variable Capacity Wireless Links: Wireless links generally are bandwidth constrained. Since there is a lower capacity in wireless links as compared to the wired links, the traffic congestion is typical rather than different.
• Energy/Power Constrained Operation: Energy consumption is vital in MANETs as these nodes operate typically off power limited sources. Some or all nodes in MANETs rely on batteries.
• Limited Physical Security: Wireless networks like MANETs are more vulnerable with the security issues available with them. Issues like eavesdropping, jammer attack, spoofing and denial of service attacks must be carefully considered.

2.8 Need of MANET Routing Protocols

A mobile ad hoc routing protocol is used to discover new routes and maintain the existing routes between the nodes in the network. MANET (Mobile Ad Hoc Network) is the collecting of mobile nodes which are present in random manner in the network has the capacity to communicate and exchange information effectively over the network by facilitating the intermediate nodes for their communication. The essential task of a MANET protocol is to create or discover the correct and efficient routes between the nodes so that information can be delivered accurately with respect to time. Route construction should be done with less overhead and minimal bandwidth consumption for effective communication.

2.9 Applications of MANETs

Applications of MANETs can be used in many critical situations today and are increasing widely. There are many applications of MANETs today and some of them are mentioned here. An ideal application is for search and rescue operations. Such kind of operations is characterized by the networks without having an infrastructure present. One of the reasons for this may be because all of the equipment was destroyed or may be the region is too remote. Rescuer must be capable of communicating to use their energy fairly and also they must maintain the security. Carrying the equipments for communication that the rescuers are already using makes the job easier.

The commercial application for MANETs includes computing everywhere which means the mobile devices are free to move anywhere in the network and yet it is possible to be in communication with the devices. The computers are allowed to forward the data to another computer or data networks may be extended far beyond the considered reach. Such networks may be more widely available and much easier in use.

Another important application of MANETs is Sensor Networks. Sensor networks are composed of very large number of small sensors which are able to detect number of properties of the area, for example; temperature, pressure, toxins, pollutions etc. In this case, the capability of sensor network is much limited. So there is always need to forward data to the central computer and for this purpose they have to rely upon others in order to forward data to the central computer. The sensor networks individually are limited in their computing capabilities but together can be very effective and the key to future homeland security. But individual sensors because of their limited computing capabilities can prove prone to failure and loss.

There are many other applications of MANETs like personal area networking where the communicating nodes may be mobile phones, laptops etc. Also this has a potential in military operations with the nodes of the network as soldiers, tanks and airplanes. Many more areas includes civilian environment to diverse taxi cab networks, conference rooms, boats and ships. (Bluetronix, 2006)

2.10 Classification of Routing Protocols

According to (Jain, 2005), ad hoc routing protocols can be classified mainly into two categories:

1. Table driven (proactive) Routing

2.10.1 Proactive (Table-Driven) Routing Protocols

These routing protocols are similar to and come as a natural extension of those for the wired networks. In proactive routing, each node has one or more tables that contain the latest information of the routes to any node in the network. Each row has the next hop for reaching a node/subnet and the cost of this route. Various table-driven protocols differ in the way the information about a change in topology is propagated through all nodes in the network. (LinuxOnly)

In proactive routing protocol each node maintains the up to date routing information of all nodes in the network. Here each node maintains the routing table and broadcast it when there is a change in network topology. As soon as source needs the route to the destination it can select from the routing table. The routing table is exchanged periodically by broadcasting to all nodes in the network to keep track of the new message even though the topology is not changed. Each nodes has the routing information of all nodes in the network though most of it undesired. Advantages of proactive protocols are that the communication experiences a minimal delay and routes are up to date. Disadvantage is that the routes are broken as a result of mobility of nodes. (Chang)

Following are the examples of table driven protocols: GSR and OLSR and some other proactive routing protocols are FSR, HSR, ZHLSR, CGSR and WRP etc.

2.10.1.1 OLSR (Optimized Link State Routing)

OLSR Protocol is based on traditional link state algorithm which supports point to point routing strategy. In this strategy the nodes keeps exchanging the information periodically in order to maintain the network topology within the network. OLSR is an optimization version of link state protocols. That means it keeps flooding the topological changes information to all the hosts across the network when happens. OLSR is also proved beneficial for the reason that it minimizes the size of each control message exchanged and also it avoids each node to rebroadcast during the updating of the routes. It uses the Multi Point Replaying (MPR) strategy to do the above task. For MPR strategy each node creates a set of all of its neighbouring nodes, generally called multipoint relays of the node, in the network to transmit the packet once again. Each node in the particular set can receive and process the packet but only cannot retransmit it. To use MRPs, each node has to keep broadcasting periodically to all one hop neighbours using hello messages. Another method to reduce the overhead is to provide the shortest path. When the time interval is reduced for the control messages transmission, it can prove more reactive to the topological changes (Kuosmanen).

The implementation of OLSR protocol basically uses two control messages; Hello message and Topology Control (TC). The hello messages are to be sent only one hop away from the host and are used for finding the link state information and host’s neighbours. MPR selector sets are constructed with Hello messages which explain which host in the network has chosen this host to act as MPR and using the information the host can select its own set of MPRs. The Topology control (TC) messages are to be broadcasted across entire network about the advertised neighbours and it includes at least the MPR selector list. As OLSR is proactive approach and it keeps updating the routing table periodically, the TC messages are also broadcasted periodically and the hosts with MPR selector set can only forward the TC messages.

There can also be MID message which is Multiple Interface Declaration message to declare that the announcing hosts can have multiple OLSR interface addresses. And again MID message are broadcasted throughout entire network and only by MPRs.

2.10.2 Reactive (Source Initiated) Routing Protocols

These protocols take a lazy approach to routing. They do not maintain or constantly update their route tables with the latest route topology. Reactive routing is also known as on-demand routing. The Reactive Routing Protocols are also called as Source initiated Demand Driven protocols. They are called so because the routes are discovered only when needed by source.

Source initiated on demand networks cerates routing only when desired by the source node. When source wants to communicate with destination then it invokes the route discovery mechanism to find the path to the destination. The route discovery process is completed once a route is found or all possible are identified. Once the rout is formed between source and destination it is maintained by a route maintenance procedure until the destination becomes inaccessible or the route is no longer desired. (Chang)

These Examples of reactive routing protocols are dynamic Source Routing (DSR), Ad hoc on-demand distance vector routing AODV, ABR, SSA, CBRP, and RDMAR.

2.10.2.1 AODV (Ad hoc On Demand Distance Vector)

Ad hoc On Demand Distance Vector protocol as the name implies it is an On Demand that is, Reactive Protocol. AODV is capable of both unicast and multicast routing. It is an on demand algorithm, it means that it builds routes between nodes only as desired by source nodes. It uses the concepts of DSR routing for route discovery and route maintenance and DSDV protocol for the concept of sequence number. It uses sequence number concept to ensure the freshness of routes. The ad hoc on-demand Distance vector algorithm facilitates the self-starting, multi hop and dynamic routing between participating nodes to establish and maintain an ad hoc network. AODV algorithm enables the nodes to find the routes for new destinations as and when they are needed and the nodes are not required to maintain the routes to the destination that are not in the part of active communication. It also enables the nodes in the formation of multicast groups and enables the nodes to respond quickly to link breakages and topological changes in the network thus the operation of AODV is loop free and thus avoids the Balham Ford count to infinity problem.

AODV routing protocol is a simple and effective routing protocol for Ad hoc networks. It also uses the concept of hop by hope routing and sequence numbers from DSDV protocol.

The following are the message types defined by AODV they are:

• Route request (RREQ),
• Route replies (RREP),
• Route error (RERR) and group hellos (GRPH) this message types are handled by UDP and IP header.

The route request message format contains the following fields:

• Source address: the address of the node which originates the route request
• Source Sequence number: the current sequence number to be used in deciding the route for the source request
• Destination address: the address of the target node for which the route is initialized.
• Destination Sequence numbers: the sequence number received by source for route towards the destination.
• Broadcast ID: The sequence number by which a RREQ route request can be uniquely identified.
• Hop counts: the number of the hops to be taken from the source node to reach the destination node that handles the Route request.

The Route Reply message format contains the following fields:

• Destination address: the address of the destination node for which the route is abounded
• Destination sequence number: the destination sequence number related to the route
• Source address: the address of the source node that originates the route request
• Lifetime: the time for which nodes receiving the route reply considers the route to be valid.
• Hop counts: the number of hops to be taken from source to destination.

The route error message format contains the following fields:

Unreachable destination address: The address of the destination that has become unreachable due to link failure. (Arbia, 2008)

In AODV the only nodes that take active participation in routing process are the nodes that sit in direct path between source and destination. The nodes which do not lie on active path do not take participate or maintain the routing table, thus AODV minimize the number of control messages sent between two nodes. As long as there is a valid routes between the source and destination for communication, AODV does not play any role and when a new route to a new destination is required and if the route to the destination does not exist only then the source node initialize the route discovery process by broadcasting the RREQ message to find the route to destination. The route request message which contains the number of hops (hop count) to the destination, the source and destination address, the source sequence number and destination sequence number and the broadcast ID the combination of broadcast ID and source address are used to identify the route request. Once the route request is created, the source node broadcast the route request and wait for route reply if the route reply is not received then the node rebroadcast the route request. When the nodes receive the route request, it checks the route request by comparing the identifier (broadcast ID +source ID) against the identifier it has already possessed in its routing table. If the receiving node has the identifier already exist then the node discards the message if this is not the case then node checks if it has route to destination or not, if the node has no route to destination then it process the route request by rebroadcast the route request by using its own address and the node updates its routing table by reverse path to source. If the node is destination node or it has active route to destination with sequence number of destination host greater than the one in route request then it create the route reply message and unicasts it to source node.

When a node receives the route reply it creates a forward route entry to node indicated by destination address field, if the node is not the source nodes then it checks its routing table and determines the next hop for route reply and forwards it to source. If the intermediate node reply to every transmission of a given route request then the destination does not receive any route request so in the situation the destination does not learn route to the originating node this led the destination to initiate the route discovery process in order to see that destinations learn routes to originating node. The originating the node should set the gratuitous flag in route request if destination is likely to route to originating node. If node responses to route request with the flag set the node returns route replies it must unicast a gratuitous route replies to the destination node.

In order to maintain the routes the nodes in AODV periodically broadcast the hello message to the neighbouring nodes if in any case the neighbouring node fails to receive the hello message it is considers that there is a link break. When a link break in active path is detected a route error message is used to inform other nodes about the link break. The error message indicates the destinations which are unreachable due to link break, in order to enable this reporting mechanism each node keeps a precursor list containing the IP address of each neighbouring node that are likely to use as a next hop towards the destination that is now unreachable. In brief when a node ‘A' needs a route to node ‘B' then node ‘A' broadcasts a route request message. The route request message has source address and broadcast identification number which is unique. The nodes which receive the request message and if they do not have the route to node ‘B' then they rebroadcast it. The nodes in the routing path also keep track of the number of hops the message has made and also the originator of the broadcast, if an node has route to node B it replies by unicasting the route replies back to the node from which it has received the request. The reply is then forwarded to node A by unicasting it to the next hop towards node A. A route error message can be used in the network when there is a link breakage. When there is a link breakage a route error message is broadcasted in network the nodes receiving this message remove the route and rebroadcast the error message to all nodes with the information added about the new unreachable destination (Belding-Royar).

Example of AODV

Here a source node ‘A' initiates the route request process by broadcasting the route request targeting the destination node ‘J'. This message is forwarded until it reaches the destination node or else until any intermediate node with fresh route to destination located here. In this case node ‘A' broadcast the route message and it is forwarded by nodes until it reaches the destination node ‘J', node ‘A' forwards message to ‘B' and ‘C' then ‘B' checks its routing table. If there is no route found then ‘B' forwards message to ‘D' and ‘D' forwards to other nodes until the destination node is found. Once the destination node receives the route request, it generates the route reply. The route propagation and route reply are as shown in figure below.

In figure 2a we can see that the node ‘H' moves away from node ‘F' so node ‘F' is not able to communicate with node ‘H' as the connectivity is lost. The node ‘F' then generates the route error message (RERR) to node ‘D' and it marks the route as invalid and unicast the message to node ‘D'. Even node ‘D' does the same thing and unicast the message to node ‘B' and node ‘B' again repeats the same thing and unicast the message to source node ‘A'. When the source node receives the route error message, it initiates the route discovery process for finding the new route to the destination. AODV also uses the hello messages in order to keep the connectivity up to date.

Following are the advantages and disadvantages as mentioned in (uni.lu, 2004).

Advantages

• Loop free Routing
• Optional Multicast
• Reduced Control Overhead

Disadvantages

• Delay Caused by route discovery process.
• Bidirectional connection needed in order to detect a unidirectional link.

2.10.2.2 DSR (Dynamic Source Routing)

Dynamic Source Routing is a simple and very effective routing protocol specifically designed for the multi hop wireless ad hoc network. The DSR protocol makes the network self-organize and self-configure without the need of existing infrastructure (Johnson, 2007). The nodes in the network cooperate with each other in multi hop communication by forwarding the packets to the nodes which are not within the direct transmission range of each other. As nodes in the network changes, some nodes may join or leave the network. In such cases all the routing is automatically determined and maintained by the DSR routing protocol. As the nodes move in the network, the number of hops needed to reach destination changes which results in changing the network topology. This protocol makes the nodes to discover the source route to any destination in the network by using multiple network hops. Each packet in its header contain the total information like the list of nodes the packet has to pass through thus making the packet routing loop free and thus by avoiding the need for maintaining the up- to- date routing information by intermediate nodes through which packet is forwarded. By means of including the source route in header of every packet the nodes, which are forwarding are overhearing this packets may cache this routing information for future use.

It uses the concept of source routing. Dynamic source routing protocol does not use periodic advertisement (Johnson). Dynamic Source Routing protocol computes the route whenever it is necessary and maintains it in this protocol source node, that is, the sender of the packet determines the complete sequence of the hops the packet has to pass through. Here the sender mentions or lists the route in the packet header identifying the hop by address of the next node to which the packet has to be transmitted on its way to the destination. Dynamic Source Routing protocol allows the packet to travel form source to destination rather than from destination to source so that each source can choose the optimal path to reach the destination. In DSR as route is part of the packet itself routing loops either short or long cannot be formed as they can be detected and eliminated that leads to optimization.

There are two stages involved in DSR and they are Route Discovery and Route Maintenance.

Route discovery is the process by which the source finds or obtains the route to destination when it wishes to communicate with particular destination. The process of route discovery is used only when source attempts to communicate with destination and when no route exists to destination.

Route maintenance is the process by which the source using a particular source route to a particular destination can detect the changes in the network topology. The changes in the network topology may be due to link break in the source route and it cannot use the source route to destination. When route maintenance indicates that source route is broken or the source can no longer be used for the source route then source attempts to use the alternative route to destination or else it invokes the route discovery process to find the new route.

Both route discovery and route maintenance operate entirely on demand and are used by source only when it wants communicate with particular destination. Dynamic Source Routing protocol does not use the concept of periodic advertisement, the level of overhead packets caused by DSR tend to zero. This is because of on demand behaviour and lack of periodic advertisement characteristics of DSR. When the nodes in the network are stationary, all the routes required for communication are discovered. During the process of route discovery the nodes tend to know the multiple routes to any destination and store them. By doing this the nodes can react rapidly to any change in routing and they can use the alternative routes to accomplish the task. Thus avoiding overhead of performing the process of route discovery each time whenever the route breaks. Due to differing antenna or propagating patterns or due to source interfaces in wireless networks, the links or routes between the two nodes may not work equally good in both directions. Dynamic Source Routing Protocol uses such unidirectional links whenever it is necessary and thus by improving the overall performance and connectivity in the system Dynamic Source Routing protocol supports the internetworking between different types of wireless network (Broch, 1999). The nodes in Mobile ad hoc network may have different radio ranges; some may have short ranges while the others may have long radio ranges. All these combinations of nodes are considered as a single network.

Dynamic Source Routing has an advantage by virtue of source routing as the route is a part of the packet itself and it contains the total information like number of hops to be taken on its way to destination. Thus by avoiding the formation of loops, any of such formation can be detected easily and eliminated so it provides the loop free path. (Johnson)

2.10.2.3 TORA (Temporally Ordered Routing Algorithm)

TORA algorithm is presented by Park and Corson which belongs to the family of link reversal routing algorithm. It follows the idea of GB (Gafni-Bertsekas) and LMR (Lightweight Mobile Routing) algorithms. TORA is Temporally Ordered Routing Algorithm which is the one algorithm from basic link reversal algorithm. Basic functionality of TORA can be divided into three main processes, 1) Creating Routes, 2) Maintaining Routes and 3) Erasing Route. Erasing route is one process to erase the other routes from the network and is done by clear (CLR) messages (Sun, 2000).

Routes can be created by assigning the directions to the links for whole network or sometimes to the part of the network as desired. For route creation process, it builds a Directed Acyclic Graph (DAG) graph. Route creation process involves route discovery first which can be discovered using Query (QRY) and Update (UPD) messages. When a node needs to find the route to the destination, it broadcast the QRY packet in the network. Route creation associates with the height (nodes distance) of the mobile nodes in the network and the data or messages flows downstream from the node with height to the nodes having lower height. However sometimes it may be possible that multiple routes are discovered to the destination, so the route with the short height, that is, distance is selected. The QRY packet broadcasted travels the network until it has reached the destination itself or at least finds the route to the destination. The corresponding node will again broadcast UPD packet to the network and will contain the node height. The packet will then travel to multiple nodes in the network and the nodes receiving the UPD packet will set their height to the larger than indicated in the packet and will broadcast its UPD packet for other nodes. This process continues and finally the number of directed links can be obtained from the originator of QRY packet to the destination. (Uni.lu, 2005)

The second process which is maintaining routes refers to the topological changes which are always expected in the mobility environment where MANETs are worked on. By this, it means that when there is a topological change, all the routes to the destination have to be re-established and redefined within a finite time. And most importantly whenever there is a partition in the network, the links to the other side of the partition should be erased and undirected and for this TORA has the concept of easing routes. Using CLR message it can clear such undirected routes. (Maltz, 2001)

2.11 Comparison of MANET Routing Protocols

The following table shows the short comparison between the routing protocols. All the available routing protocols with MANETs provide loop-free and shortest path routing. Each characteristic has been mentioned separately for the routing protocols.

Chapter 3: Implementation

This chapter looks into the different parameters used in the simulation of the four routing protocols (From chapter 2). Also it provides an overview of some popular network simulators which could have been used for the project.

3.1 Network Simulators

The Popular Network Simulators

This section gives a brief overview of each simulator. Its aim is to summarize the different implementation approaches of each simulator. The way a new algorithm is integrated can be pretty different from one simulator to another. The various simulators commonly available are:

• OPNET Modeler
• NS -2
• GloMoSim/Parsec

3.1.1 OPNET Modeler

OPNET Modeler is a powerful network simulator developed by OPNET. It can simulate all kinds of wired networks, and an 802.11 compliant MAC layer implementation is also provided. Although OPNET is rather intended for companies to diagnose or reorganize their network, it is possible to implement one's own algorithm by reusing a lot of existing components. Most part of the deployment is made through a hierarchical graphic user interface. Basically, the deployment process goes through the following phases.

First step it to choose and configure the node models (i.e. types) that is to be used in the simulations -for example a wireless node, a workstation, a firewall, a router, a web server, etc. Then build and organize the network by connecting the different entities. The last step consists in selecting the statistics needed to collect during the simulations.

The difficulty with OPNET Modeler is to build this state machine for each level of the protocol stack. It can be difficult to abstract such a state machine starting from a pseudo-coded algorithm. But anyway, state machines are the most practical input for discrete simulators. In summary, it is possible to reuse a lot of existing components (MAC layer, transceivers, links, etc.) improving the deployment process. But on the other hand, any new feature must be described as a finite state machine which can be difficult to debug, extend and validate. (Opnet, 2009)

3.1.2 NS -2

NS-2 is a discrete event network simulator that has begun in 1989 as a variant of the REAL network simulator. Initially intended for wired networks, the Monarch Group at CMU have extended NS-2 to support wireless networking such as MANET and wireless LANs as well. Most MANET routing protocols are available for NS-2, as well as an 802.11 MAC layer implementation. NS-2's code source is split between C++ for its core engine and OTcl, an object oriented version of TCL for configuration and simulation scripts. The combination of the two languages offers an interesting compromise between performance and ease of use.

Implementation and simulation under NS-2 consists of 4 steps:

• Implementing the protocol by adding a combination of C++ and OTcl code to NS-2's source base;
• Describing the simulation in an OTcl script;
• Running the simulation and
• Analyzing the generated trace files.

Implementing a new protocol in NS-2 typically requires adding C++ code for the protocol's functionality, as well as updating key NS-2 OTcl configuration files in order for NS -2 to recognize the new protocol and its default parameters. The C++ code also describes which parameters and methods are to be made available for OTcl scripting. The NS-2 architecture follows closely the OSI model. We have adapted the implementation of flooding provided in NS-2 in the context of diffusion in sensor networks. (Neglia)

An agent in NS-2 terminology represents an endpoint where network packets are constructed, processed or consumed. Such an Agent was implemented at the Application layer for the broadcast source, and the simulation trace was collected at the MAC layer. Some disadvantages of NS-2 stem from its open source nature. First, documentation is often limited and out of date with the current release of the simulator. Fortunately, most problems may be solved by consulting the highly dynamic newsgroups and browsing the source code. Then code consistency is lacking at times in the code base and across releases. Finally, there is a lack of tools to describe simulation scenarios and analyze or visualize simulation trace files. These tools are often written with scripting languages. The lack of generalized analysis tools may lead to different people measuring different values for the same metric names. The learning curve for NS-2 is steep and debugging is difficult due to the dual C++/OTcl nature of the simulator. A more troublesome limitation of NS-2 is its large memory footprint and its lack of scalability as soon as simulations of a few hundred to a few thousand of nodes are undertaken.

3.1.3 GloMoSim/ PARSEC

GloMoSim is a scalable simulation environment for wireless and wired networks systems developed initially at UCLA Computing Laboratory. It is designed using the parallel discrete-event simulation capability provided by a C-based parallel simulation language, Parsec. GloMoSim currently supports protocols for purely wireless networks. It is built using a layered approach. Standard APIs are used between the different layers. This allows the rapid integration of models developed at different layers by users. The protocols with the current library have been shown against the layers here:

To simulate the network in GloMoSim, one need to the latest compiler which is called Parsec Compiler. Parsec codes are used widely in GloMoSim kernel.

3.2 Simulation with OPNET Modeler 14.5

By looking at (Wolf, 2009), OPNET Modeler is a leading tool for the network research and development in the industry today. It also allows designing as well as studying the different protocols, communication networks, devices and applications. Different protocols include the protocols for the wired networks, wireless networks and MANETs (Mobile Ad hoc Networks). It provides a platform to make new ideas, test them and also find the appropriate solutions for them at the low cost. To make our ideas, it provides us a graphical editor interface to make various entity models from physical layer modulator to application process. All models made in OPNET Modeler are modeled in object oriented approach which in terms provides the easy mapping to the real systems. OPNET is a simulator built on the top of a discrete event system. Once the system is made, it simulates the system behaviour and also processes it by the processes defined by the user. Hierarchy strategies are used for organizing all the models to build whole network. OPNET also provides us the programming tool to design any kind of packet format we want to make in the routing protocols.

Programming in OPNET has following major tasks:

• Define the packet format
• Define the state transition machine for running the protocols
• Define process modules and
• Transceiver modules needed for each device node
• Define the network model by connecting the device nodes together using user-defined link models.

Finally we can learn many things with the OPNET Modeler tool. As stated above, OPNET Modeler is a leading R&D tool for network industry today. So by building up our own network and simulating them we can gain experience of making a communication network in the real world. Also it can help us to gain the knowledge of all the layering techniques and protocol automatons learnt in the text. We can build our own router and can check how it performs in OPNET is a real good experience (Prof. Suda).

3.3 Software Environment

The software used in the project for investigating the performance of MANET routing protocols is OPNET Modeler 14.5. OPNET Modeler is a powerful network simulator tool developed for the large and small networks by OPNET Technologies Inc. OPNET Modeler gives rise to the Research and Development process for the analysis and the design of the communication networks, devices, protocols and applications. It provides users the platform where they can analyse the simulated networks for comparing the impact of different technology designs. The standard model library of OPNET Modeler includes the hundreds of vendor specific and generic device models which include different routers, switches, workstations and packet generators.

 As far (Floyd, 2009), OPNET Modeler is a leading commercial tool OPNET Modeler through its R&D tools provides the possible and useful solutions that can help in many academic researches today in the following areas such as; Evaluation and enhancement of wireless technologies; for example, WIMAX, Wi-Fi, UMTS etc., evaluation and design of MANET protocols, analysis of optical network designs, enhancement in the core network technologies such as VoIP, OSPFv3, IPv6, TCP, MPLS and power management scheme in sensor networks (Opnet, 2009).

3.4 System Requirements

3.5 Simulation Scenarios

OPNET Modeler 14.5 has been used for the simulation of routing protocols and the scenarios can be created and performed in four easy steps as follows:

• Simulation Setup
• Choosing Statistics
• Running the simulation
• Viewing the results

3.5.1 Simulation Setup

This is the very first step for any project to start in OPNET Modeler. This can be done by creating a new project and empty scenario and setting up the different models in the workspace. There are two ways to design any network in workspace, either automatically or manually. For configuring the network automatically, Rapid Configuration option can be used or the network can be designed manually in our own way. For manual designing the object pallet has to be used and by dragging and dropping the different models from object palette to the project editor workspace, the network can be designed. If the scenarios have already been defined, it can be imported if required.

We have used OPNET Modeler 14.5 for this project and the performance for different routing protocols has been studied. For each scenario, the performances of the routing protocols have been compared for three performance metrics, that is, End-to-End Delay, Network Load and Throughput. Performance Metrics are explained in the section 4.1 in this paper.

The simulation has been conducted for small-scale network and the large-scale network, that is, for 20 nodes and 100 nodes. This provides a clear understanding for the comparison of the performance and so becomes easy to study. The traffics used for the network are High FTP Load, Light HTTP Browsing and High Resolution Video. The A step-by-step simulation and design can be seen in Appendix-A. Appendix-A explains the key parameters chosen for the simulation.

The following scenarios have been considered for the simulation:

• Scenario I: High Resolution Video for 20 mobile nodes.
• Scenario II: High Resolution Video for 100 mobile nodes.
• Scenario III: Low HTTP Browsing for 20 mobile nodes.
• Scenario IV: Low HTTP Browsing for 100 mobile nodes.
• Scenario V: High FTP Load for 20 mobile nodes.
• Scenario VI: High FTP Load for 100 mobile nodes.

Many others scenarios are also simulated and the best scenarios has been considered for the project. Under each scenario, the behaviour of the four routing protocols AODV, OLSR, DSR and TORA is simulated. The goal of the project was to study the performance of the routing protocols for different traffic across the network.

All of the above simulated scenarios are considered with the node mobility of constant 10 m/s. The campus network of the area 1000m×1000m was selected for modelling the network. For this project three kind of traffic was chosen, High FTP Load, Light HTTP Browsing and High Resolution Video. The reason for these kinds of traffics is explained in the next section. The mobile nodes and the server are spread randomly in the workspace. Then the Application Configuration, Profile Configuration and Mobility Configuration have been selected for defining the application and provide the mobility to the nodes in the network.

For the mobility purpose the Random Waypoint Mobility Model has been chosen which is by default in the OPNET Modeler 14.5 and the most commonly used model for providing the mobility in ad hoc networks. When the nodes move in the network for a particular time, it changes its direction. The pause time before changing the direction has been kept default for the project. Random Waypoint Model has been explained in the next section.

The different applications have been defined to generate the traffic in the network using the Application Configuration and the profiles for the particular application using the Profile Configuration. The attributes for Application Configuration and Profile Configuration are explained later in this chapter. The traffic is been deployed to the network after Application Definition and Profile Definition.

A. WLAN Parameters:

The mobile nodes selected were WLAN mobile nodes and has been setup for the data rate of 11Mbps and power of 0.005 Watt.

We setup the Wireless LAN parameters for each node according to the need of the project.

B. Traffic Generator Parameters:

Traffic is been generated in the network configuring the following process models and the server. The configuration for these models is explained below:

B.1. Application Configuration:

The traffic generated in the network was both heavy and light traffic; High Video Resolution and FTP Load and for Light HTTP Browsing. The parameters were defined individually for each kind of traffic generated and named applications as video_app, ftp_app and http_app. The applications have been defined in the application definition as High Resolution Video for the heavy video traffic, Light HTTP Browsing for light HTTP traffic and High FTP Load for heavy FTP Load.

We kept the application parameters by default for all the applications.

B.2. Profile Configuration:

For deploying the traffic in the network, profile has to be defined for using and deploying the traffic. We defined a profile for the application used as ‘app_pro' which supports the applications video_app, ftp_app and http_app defined in the Application Configuration.

In addition to the application profile parameters, the application start time was set to uniform(5,10) and the duration for the application is set to the end of profile. Moreover, the repeatability of the application is set to unlimited during the profile.

C. Routing Protocol Parameters:

The appropriate routing protocols are chosen with Ad hoc Routing Protocols and the parameters for the particular protocol used are defined. The four kinds of protocols are been used in the project which is AODV, OLSR, DSR and TORA. The parameters of the protocols are not really changed in this case and are kept default.

D. Trajectory:

Trajectory is the path on which the mobile node moves in the network. The trajectory used in this case is mobile_server1. The speed of the mobile nodes can be kept constant or other, as desired by configuring the mobility configuration. The mobile node moves with the constant speed of 10 m/s in this case and follow the trajectory.

Reasons for the Scenarios

Following are the reasons for selecting the particular scenarios for this project:

Many papers have been referred on the performance evolution of routing protocols which explains the scenarios used for their projects.

• Many papers (Mahmoud), (Maltz, 2001), (Nyirenda, 2009), (Zukarnain, 2009) on routing protocols has been studied and read and then these scenarios are considered to get something different and possibly efficient results.
• There is always need to have MANET performing better for High video conferencing and high FTP load when mobility is considered and so also light internet browsing is needed as per as HTTP application is considered.
• A considerable difference between the small and the large network will give the clear and noticeable outcome.

Random Waypoint (RWP) Mobility Model

Random Waypoint Mobility model, as its name suggest let the nodes move randomly in the network. Usually it allows the node mobility in zigzag pattern from one waypoint to the other. The waypoints are the points for the nodes to change the direction and are distributed uniformly over the network. Before changing the direction the nodes pause for a particular time and then start moving to the next waypoint. If the pause time expires, the mobile node chooses the random destination as well as speed for the next destination. Each node reaches a particular waypoint with a selected speed and the pattern and changes the direction after pausing at the point for a particular time. The RWP parameters can be kept default or can be configured manually if desired. RWP is an elementary model describing the independent pattern for node mobility by simple factors. (Jin, 2005)

RWP model can be used for the reasons:

• It's often used for ad hoc networks.
• It's easy to implement for the process simulation.

3.5.2 Choosing Statistics

The different scenarios for the performance comparison of different routing protocols have been designed and the results have been collected. There are three kinds of statistics which can be considered for individual statistics. They are Global Statistics, Node statistics and Link Statistics and can be selected from DES (Discrete Event Simulation) menu. The desired performance metrics, we want to see the performance for, has been selected from the particular statistics. For this project, End-to-end delay, Network Load and Throughput have been considered from the global statistics to see the network performance. After selecting the performance metrics, the simulation has been run to view the results. The results are shown graphically for the selected performance metrics.

3.5.3 Running the simulation

The DES simulation can be run for desired simulation time in seconds, minutes, hours etc. The simulations in this project have been run for 600 seconds for the results for the better and efficient results. The simulations were performed repeatedly for the reliable results and the results were consistence. The seed value is considered as 128. Simulation kernel was kept as optimized as it gives the better values for the simulation as compared to the other two. The DES simulation is run 100 values per statistics and update interval of 500000 events. Running the simulation compiles and debugs our network and search for any errors or warning in the designing.

3.5.4 Viewing the Results

The average values for the results have been taken to compare the results with the overlaid view which makes it clear to distinguish amongst all. All the results for different parameters as well as different scenarios for small and large scale networks are taken separately. For each kind of traffic supplied, we could get six different graphs. For High FTP Load we got three graphs for 20 nodes and three for 100 nodes. The three kinds of graphs show End-to-end Delay, Network Load and Throughput. Similarly, the graphs have been obtained for Light HTTP Browsing and High Resolution Video. The results and the different graphs obtained are shown in the next chapter 4.

Chapter 4: Performance Evaluation and Analysis

In this chapter, the various metrics considered for the performance evaluation and the design parameters of the routing protocols are described. The performance metrics has been explained first following the results of the simulation along the performance metrics.

4.1 Performance Metrics

Different performance metrics represents different characteristics in the performance evaluation of routing protocols. And they represent the characteristics of the overall network performance. This report explains the three basic performance metrics used in the comparison to study their effects on the network performance for different protocols. The three metrics are end-to-end delay (seconds), Network Load and Throughput.

4.1.1 End-to-end Delay

This is the time taken by the packets to be transmitted for the entire network from the source node to the destination node. Basically delay for the packet means the time taken between the packet creations at the source node and the packet destruction at the corresponding destination. This time is measured in seconds.

There are various applications in MANETs which requires different levels of packet delays. Some applications are delay sensitive applications which need to have a low average delay such as voice applications. There are some applications like FTP which may require the delay up to certain level in the network. The delay in the network is caused by many factors such as node mobility, packet retransmission due to weak signal strength and connection tearing and making. Therefore we can say that a delay is a result of adaptation of various constraints in the network by the routing protocols. It represents the reliability of the routing protocols. The three kinds of traffic have been generated across the network for the performance evaluation which are shows and analysed as follows.

The first graphs give the output for the high resolution video application for 20 and 100 nodes followed by the similar results for light HTTP browsing and high FTP load. The graphs have been taken as overlaid view to compare the performance and the average value has been considered.

For High Resolution Video

The graphs shown below are the result of the routing protocols AODV, OLSR, DSR and TORA under the simulation setup described in the section3.4 of chapter3. The following graphs show the average delay of AODV protocol (in seconds) when high video traffic has been passed through the network. Graph A shows the average delay for a small scale network for 20 nodes and graph B shows the result of corresponding large scale network for 100 nodes. From the graphs, we can say that the protocols OLSR and TORA perform better than AODV and far better than DSR.

Scenario 1

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High Resolution Video

Area: 1000m×1000m

Graph 1: Average Delay for High Resolution Video

The overlaid view for the routing protocols has been taken for 20 nodes to compare the performance. From the graph, we can say that DSR protocol shows the highest end-to-end delay as compared to the rest of three protocols and so is not very good in delivering the packets in the network as the rest of the three are. TORA shows less delay in delivering the packets to the destination for high resolution video as it has been made for reducing the impact of mobility in the ad hoc networks.

The average delay for 100 nodes does not show the significant difference as compared to the 20 nodes graph except that after 200 seconds simulation DSR protocol starts increasing the performance. The AODV protocol shows the second highest delay for 100 nodes followed by OLSR and TORA. TORA has been seen performing better than all protocols in both the cases; for 20 nodes as well as 100 nodes.

For Light HTTP Browsing

The graphs have been taken for two scale networks for light http traffic. From the graphs, we can see that both protocols DSR and AODV start performing after 100 seconds of the simulation time.

Scenario 2

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: Light HTTP browsing

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 2: Average delay for Light HTTP Browsing

The average delay for light http browsing is been shown for small network of 20 nodes which shows some noticeable difference in average delay between the protocols. In this case, TORA takes more time for delivering the packets initially as compared to rest of the three protocols. And again DSR shows the second highest delay followed by AODV and OLSR. For light traffic and mobility OLSR has less network load and therefore less delay is caused by OLSR.

The average delay for 100 nodes again shows the highest average delay for packet delivery for TORA. The delay for TORA keeps on increasing as the simulation time increases whereas AODV, OLSR and DSR shows consistence performance as compared to TORA. We can say clearly by the graphs that TORA performs worst than the remaining protocols for light internet browsing. TORA is generating large control overhead in the network rather than in certain routing area which would have decreased the delay generated. This increased the route discovery and initialisation packets process causing the higher delay in the network.

For High FTP Load

The graphs show the network result for small and large networks for high FTP load.

Scenario 3

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High FTP Load

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 3: Average Delay for High FTP Load

The graph A demonstrates the average delay for the high FTP load for 20 nodes. It can be observed in the graph that OLSR performs better than the rest with the minimum average delay, followed by AODV. DSR shows the comparatively high and increasing delay with increasing simulation time. Also TORA in this case has the significant delay with some peak values at the starting of the simulation and a consistence values after 300 seconds. The high peak values for the delay were because of the loss of routing information.

The graph B shows the high FTP load for 100 nodes which shows the average delay increasing with the simulation time for TORA. The rest of the protocols are seen performing better than TORA with the high FTP traffic across the network. When we look at the performance of AODV, OLSR and DSR in the graph C, OLSR is much better in delivering the packets to the destination with the minimum delay. Both the reactive protocols shows delay at the same simulation time with the high peak values at 100 to 110 seconds. Reactive routing protocols can incur much less overhead than that of small networks. On the other hand, OLSR can be seen performing better than all with the stable level throughout the simulation due to its table driven approach.

4.1.2 Network Load

Network load is basically a large impact of network protocols on the network. Besides network load is the result of bandwidth used, buffer availability and the processing time between the intermediate nodes and is responsible for the traffic delay.

For High Resolution Video

The following two graphs are the simulation results of network load when high video traffic passed to 20 and 100 nodes network. The protocols perform better with the less network load during the simulation and vice versa.

Scenario 4

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High Resolution Video

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 4: Average Network Load for High Resolution Video

From the graph below, it is obvious that the reactive protocol (DSR) has less network load during simulation as compared to other three protocols which has highest network load with the peak value of 3400000 bits/sec. TORA showed gradual increase with the curve till the end of simulation as it performs route creation, route maintenance and erasing the routes processes which consumes much network bandwidth and so the network load increases while AODV and OLSR curve starts falling after reaching high peak. OLSR shows high peak value for the network load during the initial stage of simulation and drops. This is because of the constant mobility of the nodes which causes the frequent changes in its link state routing and also results in changing the MPR nodes in the network. Also OLSR is a table driven approach so it generates high routing overhead which increases the network load.

On the other hand AODV also shows the increasing network load initially. This is because AODV keeps transmitting RREQ and RREP packets through the network which causes much communication overhead. Due to increase in communication overhead, bandwidth consumption increases in the network causing the high network load.

The graph for 100 nodes doesn't much differ from the result of 20 nodes except the performance of AODV which is better for large network. AODV shows the better performance with less network load while DSR shows higher network load than AODV. OLSR and TORA shows high network load because of the link state algorithm and high routing overhead. TORA was expected to perform well for the network load. TORA limits the communication overhead to the node area and saves the utilisation of network bandwidth which increases the performance of TORA but TORA in this case outperforms.

For Light HTTP Browsing

The graphs explain the network load for the 20 and 100 node scenarios when light HTTP traffic is supplied through them.

Scenario 5

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: Light HTTP browsing

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 5: Average Network Load for Light HTTP Browsing

For light load internet surfing, DSR and AODV performance is always better than TORA. As AODV shows higher routing overhead than DSR with RREQ, RERR and RREP packets dominating, DSR's overhead. TORA in this case shows high network load at the peak value of 25000 which then falls slowly with the simulation. This is because of the link reversal algorithm employed in TORA so the link failure is limited to a certain area of the network and doesn't affect the whole topology. This results in improving the network performance limiting the network overhead which can be seen by graph falling down for TORA.

There were no such noticeable changes for 100 nodes as compared to 20 node network. All the protocols were performing better than that of 20 nodes as the mobile hosts are situated near and bandwidth required will be less and so the network load will be less. Reactive protocols were performing after a certain amount of time of simulation because they find the route when it is required and so the bandwidth required is always less than that of proactive and hybrid routing protocols. When there is less bandwidth consumed in the network, the performance measured is always better with the less network load as in the graph.

For High FTP Load

The results for high FTP load traffic are shown below for both 20 and 100 nodes.

Scenario 6

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High FTP Load

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 6: Average Network Load for High FTP Load

When the high TCP application traffic is passed through the network the load increases on the network routing. TORA and OLSR are found out to have high load on routing while AODV and DSR have network traffic load of 70000 which in turns decreases to 30000 till the end of the simulation.

On the other graph for 100 nodes, high TCP application traffic creates high network load on the routing protocols as that of 20 nodes. And TORA has high network load than the remaining protocols and at the same time AODV and DSR show better performance with less network load. The reason is that reactive routing protocols like AODV and DSR incurs less routing overhead as compared to proactive protocols as per as mobility is concerned.

4.1.3 Throughput

It is the average value of ratio of total number of packets delivered to the destination for till the time taken for the last packet received. The value is measured in bits per seconds or bytes per seconds. It is measured in bits/seconds in this case.

Following are the results of the simulation of the high resolution video, light HTTP browsing and high FTP load for 20 and 100 nodes. The protocols showing higher value of throughput is supposed to have the better performance amongst others.

For High Resolution Video

The following graphs show the performance of the routing protocols for 20 and 100 nodes network.

Scenario 7

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High Resolution Video

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 7: Average Throughput for High Resolution Video

The graph shows the average throughput for each protocol. It is clear from the graph that AODV, OLSR and TORA show the better network performance. AODV and OLSR show high throughput generated during the mid of the simulation and goes down as AODV updates its links very often by sending periodic updates which causes the network to be overloaded with broadcasting. And on the other hand OLSR follows the concept of link state algorithm which updates its tables by sending control messages to their MPR very often which degrades the performance as the simulation time increases with respect to the network nodes. TORA shows its performance increasing as the simulation proceeds. AODV generates its highest peak performance at 3600000 bits/secs followed by OLSR on 3400000 bits/secs. DSR protocol shows comparatively less throughput than others.

The average throughput for each protocol for 100 nodes can be seen in the graph. For high resolution video, OLSR and TORA can be seen performing better with the peak values at 5000000 bits/secs during ending of simulation and shows a slight drop at the end. This is because TORA is good in maintaining overhead for the communication between nodes. So as the nodes increases TORA performs better at the constant node mobility.

For Average Throughput for Light HTTP Browsing

The following graphs are the results of simulation of the network with passing light HTTP traffic. The traffic parameters for HTTP application are kept default and the simulation is carried out. For both the scenarios it is seen that OLSR performance is best throughout the simulation with the significant AODV performance.

Scenario 8

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: Light HTTP Browsing

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 8: Average Throughput for Light HTTP Browsing

Figure shows the average throughput for the routing protocols with light HTTP browsing and it is seen that proactive protocol OLSR performs far better than rest of the three. It shows the great performance right through the beginning of the simulation and maintained a stable level throughout the simulation.

As it is very clear from the graph, OLSR is performing much better for light HTTP traffic for both the scenarios. For graph B, it can be seen performing at a constant value of 18000000 bits/secs throughout the simulation. While DSR and TORA did not show significant throughput for the large network as AODV did. AODV starts performing on 100 seconds the simulation start and shows an average raise in the performance.

For High FTP Load

The graphs below show the performance of the protocols for the small as well as large networks. And it is clear from both the graphs that the proactive protocol OLSR has been seen performing better than all in both scenarios as for light HTTP traffic.

Scenario 9

No. of Nodes: 20 and 100 nodes

Mobility: constant 10m/s

Traffic: High FTP Load

Area: 1000m×1000m

Simulation Time: 600 seconds (10 minutes)

Graph 9: Average Throughput for High FTP Load

The average throughput is shown in graph A for a small network of 20 nodes. It is observed that OLSR generates the best throughput for high FTP load across the network regardless of routing overhead and end to end delay. The peak value OLSR show is on 300000 bits/secs and shows consistency with the simulation progress. The same kinds of graphs are seen for AODV, DSR and TORA but with the less significant value as compared to OLSR. The highest peak for them is around 100000 bits/secs at the initial stage and is been consistent for the rest of simulation. The reason that TORA did not perform well is because TORA erases routing information when it's not in use, which ultimately resulted in less throughput generated.

This graph shows the simulation of 100 nodes for the network with high FTP traffic. As mentioned earlier, OLSR's performance is noticeably better. For large network, TORA and DSR could not show the significant performance for high FTP load. AODV still can be seen giving better throughput as compared to TORA and DSR as the simulation time increases. This is due to the reason that AODV has multiple route information available which produces the higher throughput.

Chapter 5: Conclusion and Future Work

In this report, the performance of four MANET routing protocols AODV, DSR, OLSR and TORA have been investigated using the tool OPNET Modeler 14.5. The performance was evaluated using random mobility and three kinds of traffic generated through the network. The performance was seen for the three performance metrics. Those are end-to-end delay, network load and throughput. This chapter contains the conclusion of the project and the possible future work which could be done ahead.

Conclusions and Findings

From the results shown in chapter 4, for high UDP application (video resolution), TORA and OLSR performs better than with the minimum delay shown whereas AODV is not very good for the packet delivery as compared to TORA and OLSR because of the mobility. DSR on the other hand shows increase in delay for small network while the delay time decreases with graph coming down for large network. This is because of the fact that DSR update routing table when it is necessary which decreases the network load but increases the delay for delivery of packets to the destination. Hence DSR showed higher delay in both the cases. When we consider light TCP application for HTTP traffic and high FTP load, TORA outperforms than that of video traffic graph. As far as mobility is concerned OLSR performs well with less network load resulting in the less end to end delay. For large network, due to increase in route discovery process and re-initialisation packets caused the higher delay for TORA. OLSR as a table driven approach shows least end delay than for both small and large scale networks as the routing information is frequently updated in the routing able for each node and no need to discover new routes to the destination. There is no concept of route erasing for reactive protocols consuming less network bandwidth as compared to TORA causing much efficiency in the network than TORA.

OLSR and AODV had increased network load as can be seen in the graphs. As OLSR is link state protocol and uses the concept of table driven approach it causes much routing overhead which increases the bandwidth consumption and so network load increases. On the other hand AODV keeps transmitting RREQ and RREP packets resulting in higher communication overhead. TORA showed highest and at the same time gradually increasing graph for network load for the high video traffic. Constant mobility for the mobile hosts in the network generates high routing overhead for OLSR causing the frequent changes link state routing. After OLSR, AODV causes high network load due to transmission of frequent RREQ and RREP packets for route discovery causing much communication overhead and bandwidth consumption. As we considered the large network, DSR experienced the increasing network load with increasing the number of mobile nodes. TORA limits the network overhead to a node area which saves the network bandwidth consumption. In this case, TORA experienced high network load giving the worst performance for the network.

With higher routing overhead in AODV, performance is always better than TORA. TORA initially showed high network load and then keeps performing better as the simulation time increases because of the limited routing overhead in the limited nodes are and not in the whole network topology. Both the routing protocols, AODV and DSR showed their consistence performance throughout the simulation with no much communication overhead and so network load. With high TCP application traffic, TORA and OLSR are found showing high network load initially and then a decrease in graph as per other routing protocols. When mobility is concerned, the reactive routing protocols acquire less routing overhead than that of proactive approach. Mobility had high impact on TORA increasing the network load in both the cases.

The protocols performing better are always counted having high throughput and vice versa. From the results gathered after simulation and the parameters used, all the routing protocols are performing better except DSR. DSR has worst performance for the small networks while increasing the number of nodes in the network increases the throughput for DSR. AODV keeps broadcasting and updating the routing table and performs poor as the nodes in the network increases. For small networks it has shown better throughput. OLSR is always a better and consistent performer when throughput is considered for the large and small networks. The remaining protocols due to their mechanism to discover the routes and update has less considerable throughput. They have more routing overhead and delay and AODV has advantage of multiple route information raise the performance as the number of nodes increases.

Future Work

As far as we have seen the performance of four protocols evaluated using OPNET Modeler a further study can be made on these protocols using different performance metrics. Wide scope of study can be done these same four routing protocols using IPV6 as we used IPV4 in this paper. IPV6 is the latest version of IP addressing and method of routing with that differs from IPV4. Extent study can be done on MANET routing protocols by considering other protocols and to evaluate them with different performance metrics. Conducting experiments on the study of mobility of nodes is an interesting aspect as we can only see node failures every now and then in the experiment. If we can introduce the artificial link failures into the nodes the work can be more advantage and interesting as well. Same work done in this thesis can also be done by changing the parameter of simulation and passing them with other topologies like Random waypoint model, poison distribution model etc... Overall the work on this area of MANET is always time consuming yet interesting.

Chapter 6: References

Arbia Anis Ben, Hamdi Hedi, Youssef Habib, “Wireless Routing Protocol Based on Trust Evaluation”, Third International Conference on Systems and Networks Communications, 2008. Available from the website:- http://doi.ieeecomputersociety.org/10.1109/ICSNC.2008.74

Bagrodia Rajeve, “GloMoSim Global Mobile Information System Library”, Mobile System Library, University of California, Los Angeles, 2009. Available from the link:- http://pcl.cs.ucla.edu/projects/glomosim/

Basagni Stefano, Conti Marco, Giordano Silvia, Stojmenovic Ivan, “Mobile Ad hoc Networking”, IEEE Press, Willey Interscience, A John Wiley & Sons Inc., Publications, 2004. Available from the link:- http://books.google.co.uk/books?hl=en&lr=&id=GnkcHEsxAigC&oi=fnd&pg=PR7&dq=%22Basagni%22+%22Mobile+Ad+Hoc+Networking:+Edited+by+Stefano+...%22+&ots=iYe86QDub5&sig=X5RoqPjwWNbIGbGbgk-VwKO9Nzo#v=onepage&q=%22Basagni%22%20%22Mobile%20Ad%20Hoc%20Networking%3A%20Edited%20by%20Stefano%20...%22&f=false

Belding-Royar Elizabeth M., Department of Computer Science, University of California, Santa Barbara, CA 93106, Perkins Charles E., Communications System Laboratory, Santa Barbara Research Centre, Mountain View, CA 94043. Available from the link:- http://209.85.229.132/search?q=cache:2oMuX2X-_OMJ:www.cs.ucsb.edu/~ebelding/txt/aodv_evol.ps.gz+http://www.cs.ucsb.edu/~ebelding/txt/aodv_evol.ps.gz&cd=1&hl=en&ct=clnk&gl=uk

Bheemalingaiah M., Naidu M. M., D. Rao Sreenivasa, Varaprasad G., “Energy Aware Node Disjoint Multipath Routing in Mobile Ad hoc Network”, Journal of Theoretical and Applied information Technology, 2005. Available from the website:- www.jatit.org/volumes/research-papers/Vol5No4/7Vol5No4.pdf

Bluetronix, Bluetronix Inc. 2006. Available from the site:- http://www.bluetronix.net/mobile_ad_hoc_networks.htm

cadfamily, 2009 available from the link:- http://www.cadfamily.com/Download.asp?ID=290069

Chang Zhijiang, Gaydadjiev Georgi, Vassiliadis Stamatis, Computer Engineering Laboratory, EEMCS, Delft University of Technology, Mekelweg 4, 2628 CD Delft, the Netherlands. Available from the source:- http://ce.et.tudelft.nl/publicationfiles/1104_536_chang_TUD_IST_Summit_2005.pdf

Floyd Sally, “Network Simulators”, National Science Foundation, April 2009, from website:- http://www.icir.org/models/simulators.html

Jain Nitesh, Routing Protocols in Mobile Ad hoc Networks, School of Information Technology Indian Institute of Technology Kharagpur, Kharagpur, 2005. Available from the link:- http://sit.iitkgp.ernet.in/research/aut05seminar2/report8.doc

Javvin, Networking TCP/IP, Network Dictionary, 2004. Available from the link:- http://www.networkdictionary.com/networking/routing.php

Jin Shudong, “Mobility Models for Wireless Ad hoc Network Research”, EECS 600 Advanced Network Research, March 28, 2005. Available from the source:- http://vorlon.case.edu/~sxj63/EECS600-S05/Lecture0328.ppt

Johnson D., Hu Y., Maltz D., “The Dynamic Source Routing Protocol (DSR) for Mobile Ad hoc Networks”, February 2007,. Available from the link:- http://mailman.rfc-editor.org/pipermail/rfc-dist/2007-February/001533.html

Johnson David B., Maltz David A., Broch Josh, “DSR: The Dynamic Source Routing Protocol for Multi-Hop Wireless Ad hoc Networks”, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA 15213-3891. Available from the link:- http://www.monarch.cs.rice.edu/monarch-papers/dsr-chapter00.pdf

Kaosar Golam Md., Hafiz M. Asif, SheltamiTarek R., Hasan Mahmoud Ashraf S., “Simulation-Based Comparative Study of On Demand Routing Protocols for MANET”, Department of Computer Engineering, King Fahd Univrsity of Petroleum & Minerals, Dhahran 31261, KSA. Available from the link:- http://www.lancs.ac.uk/postgrad/asifh1/final%20version%20of%20india%20paper.pdf

Linux Only, “Proactive Routing Protocols”. Available from the Website:- http://www.linuxonly.nl/docs/1/2_Proactive_routing_protocols.html

Maltz David A., May 2001, “On-Demand Routing in Multi-hop Wireless Mobile Ad Hoc Networks”, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA 15213. Available from the link:- http://www.monarch.cs.rice.edu/monarch-papers/dmaltz-thesis.pdf

Maltz David A., Broch Josh, Jetchava Jorjeta, Johnson David B., “Effect of On-demand Bahaviour in Routing Protocols for Multi Hop Wireless Ad hoc Networks”, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA 15213, from the website:- http://209.85.229.132/search?q=cache:2mMWOyO987wJ:www.monarch.cs.rice.edu/monarch-papers/jsac99.ps+Both+route+discovery+and+route+maintenance+operate+entirely+on+demand+and+are+used+by+source+only+when+it+wants+communicate+with+particular+destination&cd=1&hl=en&ct=clnk&gl=uk

Mohd Saad Mohd Izuan and Zukarnain Zuriati Ahmad, “Performance Analysis of Random-Based Mobility Models in MANET Routing Protocol”, Dept. of Communication Technology and Networks, Faculty of Computer Science and Information Technology, University Putra Malaysia, EuroJournals Publishing, Inc. 2009.from the website:- http://www.eurojournals.com/ejsr_32_4_01.pdf

Neglia Giovanni and Ibrahim Mouhamad, “Introduction to Network Simulator”, Maestro Team, INRIA Sophia-Antipolis- France. Available from the source:- http://www-sop.inria.fr/members/Giovanni.Neglia/ns_course/slides/lec1.pdf

Nyirenda Bained and Mwanza Jason, “Performance Evaluation of Routing Protocols in Mobile Ad hoc Networks (MANETs)”, Blekinge Institute of Technology January 2009. Available from the link:- http://www.bth.se/fou/cuppsats.nsf/all/247e26e7f470cf9fc1257543005b1bfc/$file/MANETs_Thesis_Report.pdf

OPNET Technologies Inc., 2009, from website:- http://www.opnet.com/solutions/system_requirements/modeler/14.5.html

OPNET Technologies Inc., 2009, from the website:- http://www.opnet.com/solutions/network_rd/modeler.html

Pettery Kuosmanen, “Classification of Ad hoc Routing Protocols”, Finnish Defence Forces Naval Academy, P.O. Box No.5, FIN-00191Helsinki, Finland. Available from the link:- http://www.netlab.tkk.fi/opetus/s38030/k02/Papers/12-Petteri.pdf

Professor Suda Tatsuya, “Introduction to OPNET Simulator”, Computer Network Research, School of Information and Computer Science, University of California, Irvina. CA 92697-3425. From the link:- http://bolero.ics.uci.edu/~ypan/OPNET/Introduction%20to%20OPNET%20simulator.pdf

Saadawi Tarek, Franklin Lakes, NJ (US); Hussein Osama, Bayonee, NJ (US), Prior Publication Data, “Routing Method For Mobile Infrastructureless Network”, Sept. 6, 2005. Available from the link:- http://www.freepatentsonline.com/6940832.pdf

Subbarao Madhavi W., “Performance of Routing Protocols for Mobile Ad hoc Networks”, Wireless Communication Technologies Group, National Institute of Standards and Technology, 100 Bureau Drive Stop 8920, Gaithersburg, MD, USA 20899-8920. Available from the link:- http://w3.antd.nist.gov/wctg/manet/docs/perf_routing_protocols.pdf

Sun Allen C., “Design and Implementation of Fisheye Routing Protocol for Mobile Ad hoc Networks”, Department of Electrical and Computer Science, May 14, 2000. Available from the Source:- http://web.mit.edu/profit/PDFS/SunA.pdf

Tonnesen Andreas, “Implementation of The Optimized Link State Routing Protocol”, Available from the link:- www.olsr.org/docs/olsr_unik_05.pdf

Uni.lu, “Temporally-Ordered Routing Algorithm”, University of Luxembourg, SECAN Lab, 2005. Availbale from the link:- http://wiki.uni.lu/secan-lab/Temporally-Ordered+Routing+Algorithm.html

Uni.lu, “Ad hoc On-demand Distance Vector Routing Protocol”, University of Luxembourg, SECAN Lab, 2005. Available from the link:- http://wiki.uni.lu/secan-lab/Ad+Hoc+On-Demand+Distance+Vector+Routing+Protocol.html

Wolf Prof. Dr. -Ing. Larg, “Simulation-based Study of TCP's Behaviours in Wireless Networks with OPNET Modeler”, August 2009. Available from the website:- http://www.ibr.cs.tu-bs.de/theses/xiaogu/wtcp4opnet.html