A SYSTEMATIC EVALUATION OF MANET ROUTING PROTOCOLS OVER UDP AND TCP IN WIRELESS MULTI-HOP NETWORK USING NS2 SIMULATOR Essay

Abstract – Mobile ad-hoc network (MANET) routing protocols can generally be grouped into three: proactive reactive and hybrid protocols [1]. This research work focuses on proactive and reactive routing protocols and comparatively analyzes their individual application on both User Datagram Protocol (UDP) and Transport Congestion Protocol (TCP) with end to end delay, throughput, jitter and packet delivery ratio (PDR) as quality of service (QoS) metrics.Forty simulations were run for simple and complex multi-hop network models using NS2 simulation platform. We ran the TCP and UDP on four routing protocols to study and analyze their individual network performances.

The Simulation results are analyzed and graphically presented using NS2 Visual Tracer Analyzer and Microsoft Excel. INTRODUCTIONMANET is a group of wireless movable ad-hoc nodes that establishes a communication network without fixed network infrastructure and administration. It promises good alternative to two or more neighboring mobile nodes to communicate and transfer data. Due to infrastructure-less and decentralized nature of MANETs, it is easily deployable in military exercises, underground mining operation, disaster relief operations and in any other scenario where traditional cellular network is inaccessible.

Extensive researches have been carried out on MANETs and challenges like security, collision avoidance, redundant links, multi-hop routing and synchronization have not been completely solved. For instance, to solve routing problem in proactive routing, every node in the network must send it routing table to other nodes in the network whether the network is active or not. This causes large overhead, excess bandwidth usage and increased power dissipation, so there is trade-off between how frequent routing tables are shared among the nodes and having an up-to-date routing table.II. RELATED WORKSA related research was performed by [2], the author simulated AODV, DSR, DSDV and ZRP routing protocols using only UDP/CBR traffic for throughput, PDR, routing overhead and average delay parameters. He assumed that AODV performed better in PDR and average throughput while ZRP displayed an outstanding performance in routing overhead and average delay. Pankaj et al studied network delay, packet delivery ratio and throughput of AODV, DSR and DSDV with different node numbers and maximum node speed of 15m/s [3]. They established that reactive routing protocols (AODV, DSR) performed better than proactive protocol (DSDV). In addition, MANET routing protocols’ performance for video streaming was studied by authors in [4] and it was concluded that Enhanced Video Streaming in MANET (EVSM) performed better than AODV and AOMDV with the explanation that in EVSM, the routing method (i.e. 60:40 Multipath Routing Design) gave a preferable performance over other two protocols. Han et al studied HYBRID, AODV and NOAH routing protocol using throughput, delay and routing load metrics. It was discovered that the energy consumed by HYBRID is noticeably lower than that of NOAH because NOAH does not have routing load and increase in the number of nodes is proportional to increase in delay [5].In this work, we studied proactive routing protocol – destination sequenced distance vector (DSDV) and reactive routing protocols: dynamic source routing (DSR), ad-hoc on demand distance vector (AODV) and ad-hoc on request multipath distance vector (AOMDV) with different node numbers. The protocols’ functionality and quality of service (QoS) delivery were individually studied through simulation on two upper layer transport protocols: user datagram protocol (UDP) and transmission control protocol (TCP). This complex simulation was performed on NS2 to generate trace file (.tr) and TCL file which were then analyzed. During simulation, file transfer protocol (FTP) traffic generation module was connected to the TCP agent to generate data while UDP data is generated using constant-bit rate (CBR). The study was carried out in both simple and complex network environment with all the intermediate mobile nodes randomly moving at a speed of between 15m/s “70m/s. The source and the destination nodes are fixed.III BRIEF OVERVIEW OF THE ROUTING PROTOCOLS Ad-Hoc on Demand Distance Vector (AODV)AODV is a reactive routing protocol that combines destination sequence number in DSDV and on-demand route discovery in DSR [6]. Using destination sequence numbers, the operation of AODV is loop-free, and it circumvents Bellman-Ford counting to infinity challenge which thus gives it rapid convergence when ad-hoc network topology changes [7]. Unlike DSR, AODV packet size is uniform and it also maintains one route between a source-destination pair. The basic message sets in AODV consists of: RREQ ” Route Request RREP ” Route Reply RERR ” Route Error HELLO ” For link status monitoring.In order to establish a connection, source S will first check it routing table for pathway to destination (D), if no route, then source (S) will broadcast a RREQ message to all active nodes in the network. A multiple RREP messages are sent back to S and it chooses the path with the greatest sequence number and smallest hop count to the D. Source S initiates the route by sending a multicast activation (MACT) message to the next hop of the route. To detect link breakage, neighboring nodes periodically exchange HELLO messages among themselves. Active nodes’ movements and link failures are propagated through Route Error (RERR) messages, which also update destination sequence numbers [8]. Ad-hoc on Request Multipath Distance Vector (AOMDV)AOMDV protocol is an extension to the AODV protocol for computing multiple loop-free and link disjoint paths. It recovers faster from link failure because of its multipath routing capability. Like AODV, AOMDV routes are established on-demand; however it has more message overheads during route discovery because of its many routes from source to destination. Setting up of multiple paths has assured desirable characteristics in the security of the transmitted data. Using multiple path protocol improves the security of the session by ensuring none of the intermediate nodes have more than a single path passing through them [9]. Also, due to it multipath nature it can detect any form of impersonation attacks like session hijacking, gray-hole attack, black hole attack and IP spoofing by authenticating the real identity of the destination node D through an alternative path. Dynamic Source Routing (DSR)DSR is similar to AODV in that it forms a route on-demand when active node requests a connection. It has route discovery and route maintenance mechanisms that function together to allow the discovery and maintenance of source routes. Each mechanism operates entirely on demand. DSR requires no periodic packets of any kind at any level in the network. [10]. More so, DSR has an advantage of storing multiple routes in their route cache for a valid route before initiating route discovery and if a route is found there is no need for route discovery. Route Caching in DSR is also used when a connection between source S and destination D fails; other route from the local cache can be used if it has the route to D. Otherwise source S initiates a new route discovery by sending RRER message thus caching helps to increase route discovery and ease propagation of route request. Destination Sequenced Distance Vector (DSDV)DSDV is a proactive protocol (table-driven) and routes to all the active nodes are discovered in advance. Whole table is broadcast after a fixed interval of time independent of any route changes. DSDV does not only generate high overheads but also uses more bandwidth because of its proactive nature hence it is not suitable for ad-hoc network with high movement and large nodes. The advantage of DSDV over traditional distance vector routing protocols is that DSDV guarantees loop-free routing.IV SIMULATION MODEL We used a detailed simulation model based on NS-2 in our evaluation. NS2 is an event driven, object oriented network simulator which enables the simulation of a variety of ad-hoc, local and wide area networks. It implements network protocols, traffic sources and queue management mechanisms. NS2 is written in two different languages (C++ and Otcl) in other to separate control and data path implementations. C++ is used for detailed protocol implementation because of its speed while Otcl, though runs slowly is used for configuration and operation because of its flexibility ” it can be modified very fast thus suitable. In our simulation, two fixed source nodes are on transport protocols (UDP & TCP) in the same network to send rescue messages to two fixed destinations. They are both subjected to the same conditions and we made the scenarios very chaotic with intermediate mobile nodes randomly moving at high speeds from a random location to a random destination. We used a zero pause time which means nodes are randomly moving throughout the simulation time (20 secs). In all, we studied the combination of this network transport protocols and ad-hoc routing protocols:Simulation Traffic and Mobility ModelsThe network used for simulations consists of 10, 20, 30, 40, 50, 60, 70, 80, 90, 100 mobile nodes with four nodes in each of the simulation fixed. Two of the fixed nodes were the source nodes for TCP and UDP transport protocols while the other fixed two nodes are the destination nodes. Traffic sources in our simulation are constant bit-rate (CBR) – it is attached to UDP network protocol; and file transfer protocol (FTP) which is attached to TCP. The packet size is set to 1500 byte. Table 1 explains more about our simulation parameters.Table 1: Simulation Parameters for node configurationParameters ValueChannel Size Channel/WirelessChannelNumber of Nodes 10, 20, 30, 40, 50, 60, 70, 80, 90, 100Source Traffics CBR, FTPAntenna Model Antenna/OmniAntennaRadio Propagation Model Propagation/TwoRayGroundNode Speed 0m/s ” 70m/s (random)Number of simulated scenarios 40Interface Queue Type Queue/Droptail/PriQueuePacket Size 1500MAC Type Mac/802.11Buffer Size 50 packetsTransport Layer s UDP, TCPSimulation time 20sV. SIMULATION RESULT AND PERFORMANCE EVALUATIONIn this research work, we performed 40 simulation runs. The simulations were run for 20 seconds with pause time being 0s (i.e. all the intermediate nodes were randomly moving throughout the simulation). For each scenario, UDP and TCP transport protocols were set up to compete for resources. Four routing protocols (AODV, AOMDV, DSR and DSDV) were used. Average end to end delay, packet delivery ratio (PDR), throughput and delay jitter were computed for all the routing protocols. The results are given below with the corresponding graphs. Average end – end delay: this is the time taken by the data packet to be transmitted from the source to the destination node. This includes all possible delays caused during route discovery, queuing at the interface queue, retransmission delays at the MAC and propagation delay.Mathematically, delay time resulting from discovery neighbor nodes in source and intermediate nodes is: D_ND=€‘_(i=1)^n–’–(T_PRHM-T_BHM)— 1where D_ND=delay in route discovering, T_PRHM=processing and returning time of Hello message. T_BHM=broadcast start time of hello message.While delay caused from transmitting and processing control packets including RREQ, RREP and RERR is represented by: D_PD=(T_RREP-T_RREQ ) 2 where D_PD=delay caused by propagation and control packet processing,T_RREP=receiving time of propagation and T_RREQ=RREQ packets transmission timeTherefore, End to end delay D_EE can be represented by: D_EE=D_ND+ D_PD 3From figure 2, AOMDV/UDP performed most badly compared to other protocols in end to end delay. This is due to the fact that if a link breakage occurs in the network, AOMDV being a multipath routing protocol tries to establish an alternative path from among the backup routes which thus generates additional delay to the packet delivery time and more so UDP which was attached to CBR continuously generate packets non-adaptively this also caused delay due to queuing at the interface therefore UDP transport protocol has high delay for all the routing protocols. AODV/TCP was able to maintain a constant delay till having a sharp delay increase when the number of nodes increased to 50 and it later normalized. This is because AODV repairs link failure by choosing a new route and speedily reestablish connection. DSR/TCP performance is the best.More so, due to its proactive nature, DSDV does not generate any traffic thus making it unsuitable for network with short connection time, large packet size, high mobility and node density. DSDV could not establish both UDP and TCP connections because the network is rapidly changing. Figure 3 shows the route messages being generated by the active nodes (scrNode) but couldn’t be transmitted successfully to the destinations (dstNode) because of the high mobility of the nodes. DSDV does not initiate route discovery mechanism therefore packets will be dropped if route does not exist. Destination = -1 which means timeout. (figure 3) Throughput: This is defined as the total number of packets a destination node D receives from the source node S divided by the time taken by D to receive the last packet. It is a function of other factors like congestion, collision, packet loss etc. It is measured in bits per second. Mathematically, T= (N_out –. L—_F . 8)/T_t 4where,Tis throughput,N_out=amount of frame to send,– L—_F=frame length, T_t=testing time and 8 is the conversion from byte per second to bit per second.TCP has the worst performance compared to UDP in throughput (figure 4), this is because the default parameter of TCP has been consciously designed to sacrifice throughput in exchange for fair bandwidth sharing on congested networks [11]. From our simulation result, AODV/UDP and AOMDV/UDP performed better than other protocols and AOMDV/UDP was able to maintain a constant throughput for node 60 to 90. More so, we observe increase in the number of nodes leads to decrease in the throughput. This is because increase in nodes creates high congestion thus the increase in propagation and processing time for the data to be sent to the destination. Packet delivery ratio (PDR): It is the ratio of data packets received by the destination node to those generated at the application layer of the source node.Mathematically, PDR is_PDR= €‘_(i=1)^n–’P_D/P_S X 100 5where P_D is the packets received by the destination node, P_S is the packets generated by the source node and n=number of active nodes In simple network, PDR for DSR/UDP is better than the other protocols but as the number of nodes increases its performance decreases. For AODV/UDP and AOMDV/UDP the packet delivery ratio increases as the network increases. In general, UDP/CBR transport protocol performs better than TCP/FTP in all of the routing protocols for packet delivery ratio. This is because connection-oriented TCP has a retransmission and control mechanisms that makes it wait for earlier packets to be delivered before transmitting new packets unlike UDP that is connectionless and transmit packets in a fire and forget manner thus having higher delivery ratio. AOMDV performs better than other routing protocols because its intermediate nodes have alternate paths thus can easily reconfigure themselves when there is connection breakage- reducing route discovery and increasing packet delivery. Delay Jitter: This is the magnitude of delay jitter. Network state is always changing and dataflow sometimes is large and sometimes is small. If the flow is large some packets will have to queue in the node to be delivered when the network state is small, so the sending time for transmitted packets from source node to the destination node will be different. The difference is referred to as Delay Jitter or Delay Variation. The higher the jitter the more unsteady the network and the more the loss of quality of service delivery (QoS).From our simulation result (Figure 6), all the routing protocols maintained a low level of delay jitter between 0 to 0.6secs except AODV/TCP that has a sharp rise at node number 50; this reflected in the end to end delay of the network at node 50. AOMDV/UDP maintained almost a constant delay jitter as the number of nodes increases. AOMDV/TCP has the highest deviation. It was observed that at node number 80, TCP did not generate any delay jitter for all the routing protocols. This is because the network did not establish RREQ connection in AOMDV (dstNode = -1 (timeout)) and RREP connection in AODV (Figure 7).UDP and TCP transport protocols’ performances were carefully studied on four ad-hoc routing protocols: AODV, DSR, DSDV and AOMDV. Forty simulation scenarios of different number of nodes were ran using ns2 and the generated tr. file and TCL files were analyzed using NAM, Microsoft Excel and NS2 Visual Tracer Analyzer 0.2.72 software. Throughput, end-end delay, PDR and jitter were assessed under both low and heavy traffic to determine the combination of routing protocol and transport protocol that delivers best QoS.Simulation results show UDP performed poorly in end to end delay with all the routing protocols but outperformed TCP in PDR and throughput. This is because UDP has no flow control mechanism so packets are delivered in a fire and forget style thus giving it a high PDR and throughput. TCP delay is short because the transmission window is minimal. More so, congestion controls (congestion window and slow-start threshold) give TCP low PDR and throughput compared to UDP. For routing protocols, reactive protocols achieved better performance than the proactive. We observed that DSDV did not establish UDP/TCP connection. The traffic generated (route table messages) did not reach the destination nodes (Figure 3) because of the high node mobility and size of the network.AOMDV and AODV performed better in PDR- the more the nodes the higher the PDR while in throughput, we observed that the more the number of nodes the lower the throughput for AOMDV/UDP, AODV/UDP and DSR/UDP.