VOL. 2, NO. 3, SEPTEMBER 211 Performance Analysis of MANET Routing Protocols OLSR and AODV Jiri Hosek Faculty of Electrical Engineering and Communication, Brno University of Technology Email: hosek@feec.vutbr.cz Abstract This paper presents a performance analysis of two Mobile Ad Hoc Network (MANET) routing protocols - Ad Hoc On Demand Distance Vector (AODV) and Optimized Link State Routing (OLSR). For the behaviour simulation and evaluation of these protocols we used the OPNET Modeler simulation tool. Each routing protocol was configured into two network scenarios - with the default and modified parameters in order to achieve better transmission characteristics. The final evaluation is presented at the end of this paper. Due to many already discussed benefits wireless networks are modern and currently very popular way of communication. Since the basic principle was introduced in the 197s, significant expansion of wireless networks came up in last two decades. There are two main variations of the wireless networks. The first variation is called an infrastructure mode and typical representative of this network type is Wireless Local Area Network (WLAN). The infrastructure mode has a basic communication model composed of a central node (access point) and wireless clients that must only communicate over the access point [1]. The second wireless network type is an infrastructureless network, commonly known as an ad hoc network. Ad hoc networks are recently one of the most attractive research fields in telecommunication technology. The wireless ad hoc network is decentralized wireless network that does not rely on a preexisting communication infrastructure. Each wireless node is capable of movement and participates in routing by forwarding of data for the other neighbouring nodes [2]. The wireless ad hoc networks can be further classified into three groups based on applications [3]. The first group represents wireless mesh networks those are made up of wireless nodes organized in a mesh topology [2]. The wireless sensor networks (WSN) belong into the second group of the ad hoc wireless networks. Basic task of WSN is monitoring of environmental physical parameters and their transfer toward a control station. The last group of the wireless ad hoc networks is known as Mobile Ad hoc Networks (MANET). MANET represents a system of wireless mobile nodes which can freely and dynamically selforganize into arbitrary and temporary network topologies allowing people and devices to internetwork seamlessly [2]. The example of a mobile ad hoc network is shown in Fig. 1. This paper is focused on the MANET routing protocols therefore the following text will be devoted to this communication platform. 1 Introduction Figure 1: Mobile Ad Hoc Network 2 Mobile Ad Hoc Routing Protocols Since the topology of the wireless ad hoc networks is changed dynamically, routing is the main issue in the mobile ad hoc networks. The MANET routing protocols have to face high challenges due to frequently changing of topologies, low transmission power and asymmetric links. In contrast with the infrastructured networks, a routing process is integrated into the mobile nodes those have to act as both the client and the server, forwarding and receiving packets to or from the other nodes. The main goal of an ad hoc network routing algorithm is to correctly and efficiently establish a route between a pair of nodes in the network and then a transferred message can be delivered according to the expected parameters [1], [2]. The route establishment should be done with minimum overhead and bandwidth consumption. There have been developed many ad hoc routing protocols which can be divided into two main classes: proactive protocols and reactive on-demand protocols, as discussed in the following text. The proactive routing algorithms, also known as table-driven, aim to keep con- 22
VOL. 2, NO. 3, SEPTEMBER 211 sistent and up-todate routing information between every pair of nodes in the network by proactively propagating of route updates at fixed time intervals. Usually, each node maintains this information in the tables, thus protocols of this class are also called tabledriven algorithms. Examples of the proactive protocols are Destination-Sequenced Distance Vector (DSDV), Optimized Link-State Routing (OLSR), and Topology-Based Reverse Path Forwarding (TBRPF) Protocols [1], [3]. A different approach from proactive routing is sourceinitiated on-demand routing. Reactive routing algorithms establish the route to a given destination only when the node requests it by initiating of a route discovery process. This process is completed once the route is found or all the possible route permutations have been examined. Once the route has been established, the node keeps it until the destination is no longer accessible, or the route expires. Examples of the reactive protocols are Dynamic Source Routing (DSR) and Ad Hoc On-Demand Distance Vector (AODV) [1], [3]. The main goal of this paper is to compare the proactive and reactive ad hoc routing protocols. Therefore the performance comparison of OLSR and AODV, the most often used proactive (OLSR) and reactive (AODV) protocols in mobile ad hoc networks, are introduced. 2.1 Optimized Link-State Routing (OLSR) Protocol OLSR is the proactive routing protocol for the mobile ad hoc networks. The protocol inherits stability of a link state algorithm and has the advantage of having routes immediately available when needed due to its proactive nature. OLSR minimizes overhead from flooding of control traffic by using only selected nodes, called Multi-Point Relays (MPRs), to retransmit control messages [4]. This technique significantly reduces the number of retransmissions required to flood a message to all nodes in the network. Upon receiving an update message, the node determines the routes (sequence of hops) to its known nodes. Each node selects its MPRs from the set of its neighbours. The set covers those nodes which are distant two hops away. The idea is that whenever the node broadcasts the message, only those nodes present in its MPR set are responsible for broadcasting the message [3]. Furthermore, as OLSR continuously maintains the routes to all destinations in the network, the protocol is beneficial for traffic patterns where a large subset of nodes are communicating with the another large subset of the nodes, and where the [source, destination] pairs change over time. The protocol is particularly suited for the large and dense networks, as the optimization is done by using MPRs those work well in this context. The higher optimization in the larger and more dense network can be achieved as compared to the classic link state algorithm [4]. 2.2 Ad Hoc On-Demand Distance Vector (AODV) Protocol The AODV protocol is one of the most used reactive ondemand routing protocols for the mobile ad hoc networks. AODV keeps the route table to store the next-hop routing information for destination nodes. Each routing table can be used for a period of time. If the route is not requested within the period, it expires and the new route needs to be found when needed. When a source node has data to be sent to a given destination, it looks for the route in its route table. In case there is the one route, it uses it to transmit the data packet. Otherwise, it initiates the route discovery process to find the route by broadcasting of a route request (RREQ) message to its neighbours [3]. The AODV routing protocol is designed for the mobile ad hoc networks with populations of tens to thousands of the mobile nodes. AODV can handle low, moderate and relatively high mobility rates, as well as a variety of data traffic levels. AODV has been designed to reduce dissemination of control traffic and eliminate overhead on data traffic, in order to improve scalability and performance [5]. 3 Simulation Model of Mobile Ad Hoc Routing Protocols in OPNET Modeler Modelling and comparison of OLSR and AODV routing protocols were realized in the simulation environment Optimized Network Engineering Tool (OPNET) Modeler which is well known software tool enabling design, simulation and analysis of different type of network technologies, architectures and protocols [6]. 3.1 Traffic Parameters In OPNET Modeler we created a model of the mobile ad hoc wireless network, shown in Fig. 2 composed of two simulation scenarios consisting of 72 mobile nodes in each one. The nodes were randomly placed in 5 x 5m campus environment. Network traffic was defined between several randomly selected pairs of the nodes through IP G711 Voice demand flow model component. Used traffic model generates interactive voice flow with the bit rate of 12kbps and average packet size of 12B. Traffic start time was set to 12 seconds from the simulation beginning. 3.2 Wireless and Mobility Parameters The mobility and wireless Local Area Network (LAN)parameters were identical for all nodes in every scenario. For the mobility definition was used the Random Waypoint profile in the Mobility Config object. All nodes were configured to randomly move within the defined wireless domain. The speed of each mobile node was defined by the uniform distribution between and 1m/s. The RX Group Config object was used for a limitation of the possible set of receivers based on distance 23
VOL. 2, NO. 3, SEPTEMBER 211 Figure 3: Wireless LAN parameters expiry time of an entry in duplicate set table. Duplicate set table is used to avoid processing of duplicate messages received within this time interval [7]. Figure 2: Simulation model of the mobile ad hoc network between the nodes. We have set the Distance Threshold to 2 meters. The wireless LAN parameters, shown in Fig. 3, were configured with default values available in OPNET Modeler except data rate that was decreased to 1Mbps which better corresponds to transfer rate in the real environment. 3.3 Routing Protocols Parameters For the AODV protocol and OLSR as well, two scenarios were created. The first scenario with the OPNET Modeler default parameters and the second one with parameters obtained by an experimental research in order to achieve better transmission characteristics. Figures 4 and 5 show default and modified simulation parameters for the OLSR and AODV routing protocols respectively. The Willingness parameter was chosen to reduce the number of MPR nodes. This attribute specifies how willing the node is to be forwarding traffic on behalf of the other nodes. The Willingness set to Always indicates that the node always should be selected to carry traffic on behalf of the others [4]. The Hello Interval and the TC (Topology Control) Interval of OLSR were set to 3 and 7 seconds respectively. The Neighbor Hold Time was set to 9 seconds and entries in the topology table expired after 21 seconds. The Duplicate Message Hold Time defines the Figure 4: Default (left) and modified (right) OLSR simulation parameters The Gratuitous Route Reply Flag was enabled for AODV as it helps in reducing of time for the route discovery. The Active Route Timeout was increased to 3 seconds in order to extend lifetime of the routing table entry. The Hello Interval was also increased from the default value to decrease of congestion in the network [7]. The Allowed Hello Loss parameter defines the number of hello packet loss the node can sustain before the declaring link break. Due to the large number of the mobile nodes the Allowed Hello Loss and the Net Diameter were increased. 4 Simulation Results and Analysis The simulation of all four scenarios with duration of 3 minutes was performed. We focused especially on amount of routing traffic and packet end-to-end delay for each protocol. Further, the most important results of our research are described. Fig. 6 shows amount of routing traffic sent in the entire network. This statistcis includes the Hello messages 24
End-to-End delay (s) Traffic Sent (bits/s) VOL. 2, NO. 3, SEPTEMBER 211 2 18 AODV_default AODV_modified OLSR_default OLSR_modified 16 14 12 1 8 6 4 2 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 Figure 5: Default (left) and modified (right) AODV simulation parameters sent, TC messages sent and TC messages forwarded. Basic difference between the proactive (OLSR) and reactive (AODV) routing principle is evident. From the results it is clear that the OLSR builds up the routing tables immediately at the beginning of simulation and no data transfer is needed. On the other hand, AODV starts the routing process by time of the route request, at 12 seconds after simulation beginning in our scenario. OLSR generates more network load. The reasons are frequent updating of the routing table and MPR nodes selecting that is performed with every change of the network topology. Fig. 6 also shows that amount of routing traffic can be easily influenced by modification of routing protocols parameters. Fig. 7 shows the ETE (End-To-End) packet delay for IP G711 Voice demand flow between randomly selected pair of nodes. This statistics reflects the time between the packet creation at the source node and its reception at the destination node. It is evident that modified AODV routing parameters bring the significant reduction of ETE delay. With the default parameters the ETE delay was about 3 milliseconds which is non-acceptable value for interactive voice. The ETE delay for both OLSR scenarios was below 3 milliseconds. The reason of this low value is that OLSR knows the route to the destination node already before communication request arrival. Fig. 8 shows the total number of OLSR routing table calculations during the simulation. Routing table recalculation happens in OLSR whenever the neighborhood or topology change is detected on the node. The neighborhood change is checked only after the processing a Hello or TC message. High number of calculations is caused by the total number of nodes and their random mobility. The total number of OLSR Hello messages sent by all nodes is shown in Fig. 9. It is evident that the number of Hello messages is lower for the modified OLSR routing parameters. This decrease is about 33%.,45,4,35,3,25,2,15,1,5 Figure 6: Amount of routing traffic sent AODV_default AODV_modified OLSR_default OLSR_modified 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 Figure 7: End-to-end packet delay for IP G711 Voice demand flow between selected pair of nodes The average route discovery time for the AODV protocol is given in Fig. 1. The time to discover a route to a specific destination is the time when the route request was sent out to discover the route to the destination node until the route reply is received back [6]. Fig. 8 shows that the average route discover time for scenario with modified AODV attributes is higher than with default parameters. The reason of it is longer Hello Interval and Active Route Timeout, see Fig. 5. When no route is found to the destination, the node drops the packets queued. This statistics, shown in Fig. 11, represents the total number of application packets discarded by all nodes in the network. The number of packets dropped is higher for the modified AODV routing parameters. It happens due to less number of Hello messages sent by mobile nodes which causes slower reaction in case of a topology change. The average number of dropped packets for modified AODV routing parameters is 9.43 which is 1.25% of total application packets sent during the simula- 25
Total number of Hello messages sent (-) Total number of packets drop (-) Number of routing table calculations Route discovery time (s) VOL. 2, NO. 3, SEPTEMBER 211 5 45 OLSR_default OLSR_modified 5 4,5 AODV_default AODV_modified 4 4 35 3,5 3 25 2 15 1 5 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 3 2,5 2 1,5 1,5 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 Figure 8: Number of OLSR routing table calculations 25 OLSR_default OLSR_modified 225 2 Figure 1: Average route discovery time for AODV protocol 18 16 AODV_default AODV_modified 175 14 15 125 1 75 5 25 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 12 1 8 6 4 2 12 24 36 48 6 72 84 96 18 12 132 144 156 168 18 Figure 9: Number of OLSR Hello messages sent tion. This packet loss is acceptable for the wireless mobile ad hoc networks. 5 Conclusion The aim of our work was to compare the performance of reactive AODV and proactive OLSR ad hoc routing protocols. For evaluation of these protocols we used the OP- NET Modeler simulation environment where we created the model of wireless ad hoc network composed of four scenarios; two scenarios with default OPNET Modeler routing parameters and the other two with modified OLSR and AODV parameters. The purpose of modification of the routing attributes was reduction of network load and end-to-end delay generated by the ad hoc routing protocols. During the analysis we confirmed that the modification of selected routing parameters is beneficial, however it can bring about deterioration of transfer characteristics for some network topologies with specific mobility level. Our analysis also shows that OLSR offers better per- Figure 11: Total number of packets dropped for AODV protocol formance in terms of packet end-to-end delay. On the other hand, as we expected, AODV generates lower network load. Finally, taking into account our research we can say that the selection of the ad hoc routing protocol should be done based on the network topology and the type of network traffic operated. Acknowledgment This paper has been supported by the Grant Agency of the Czech Republic (Grant No. GA12/9/113) and the Ministry of Education, Youth and Sports of the Czech Republic (Project No. MSM2163513). References [1] Royer, M. E., Toh, C., A Review of Current Routing Protocols for Ad Hoc Mobile Wireless Networks. IEEE 26
VOL. 2, NO. 3, SEPTEMBER 211 Personal Communications, 1999, pp. 46-55. [2] Illyas, M., The Handbook of Ad Hoc Wireless Networks, CRC Press, 23. [3] Boukerche, Z., Algorithms and Protocols for Wireless and Mobile Ad Hoc Networks, Wiley, 29. [4] Clausen, T., Jacquet, P., Optimized Link State Routing Protocol (OLSR) - RFC 3626, 23. [5] Perkins, C., Belding, E., Das, S., Ad hoc On-Demand Distance Vector (AODV) Routing - RFC 3561, 23. [6] Opnet Technologies, OPNET Modeler Product Documentation Release 16., 21. [7] Qasim, N., Said F., Aghvami, H., Mobile Ad Hoc Networking Protocols Evaluation through Simulation for Quality of Service. IAENG International Journal of Computer Science, 29, pp. 1-9. 27