2008 International Conference on Computer Science and Software Engineering Performance comparison and analysis of routing strategies in Mobile ad hoc networks Fu Yongsheng, Wang Xinyu, Li Shanping Department of Computer Science Zhejiang University Hangzhou, Zhejiang, China, 310027 Abstract The performance of routing protocols in Mobile Ad hoc network (MANET) always attracts many attentions. As many previous works have shown, routing performance is greatly dependent to the availability and stability of wireless links. Although there are some studies reported to evaluate the performance of routing protocols in MANET, little work is done for the system overall performance, which is generally referred to as the network throughput and the endto-end delay. This paper evaluates the routing strategies of the three routing protocols (AODV, OLSR and SRMP) and compares their performance under two different mobility models. We study those protocols under varying metrics such as node mobility and network size. Our objective is to provide a qualitative assessment of overall performance of the routing strategies in different mobility model. The results illustrate that the routing strategies behind the three protocols affect their performance deeply. Meanwhile, different from previous results in the literature, the result shows that it is the combination of routing strategies behind the routing protocols to determine the performance in given scenarios. Keywords-performance, mobile ad-hoc, routing strategies, local broadcasting, mulitcasting, unicasting, hop-by-hop, source routing I. INTRODUCTION A mobile ad-hoc network (MANET) is a self-configuring network of mobile routers (and associated hosts) connected by wireless links the union of which form an arbitrary topology. The routers are free to move randomly and organize themselves arbitrarily; thus, the network's wireless topology may change rapidly and unpredictably. Connections are possible over multiple nodes (multi-hop ad hoc network). Fault Tolerance IP addressing Mobility Management TCP/UDP Security Clustering Qos/Multimedia Location services Multicasting/Broadcasting Multiple Access Bandwidth Management Power Mangement Routing Issues 2007 2006 2005 2004 0 100 200 300 400 500 600 700 800 900 Papers Figure 1. Trends of the MANET research from 2004 to 2007 Routing protocols then provide stable connections even if nodes are moving around. The trends of the papers published in IEEE Xplore [1] and ACM lib between 2004 and 2007 are illustrated in Fig. 1 And it shows that routing algorithms in this area are the hottest topics in the past years due to the fact that existing internet routing protocols were designed to support fixed infrastructure and they are unsuitable for MANETs. More than 4953 papers from 2004 to 2007 were searched using the keyword ad-hoc networks. The research trends before 2004 can be found in [2]. During these years, there have been many discussions about which type protocols perform better. However, the protocol performance is different based on scenarios and little work is done for the routing strategy overall performance, which is generally referred to as the network throughput and the end-toend delay. In this paper, we will present different routing strategies in both proactive and reactive protocols and discuss their performance factors in two different mobility models. This paper is organized as followings: Section 2 shows the related works in protocol performance of MANET. Protocols AODV, OLSR and SRMP are discussed in Section 3. We present the routing strategies in Section 4. Scenarios based on two mobility models and experiments are illustrated in Section 5. Final section ends into a summary on the performance factor of routing strategies and future work. II. RELATED WORKS Several simulation-based performance comparisons have been done for ad hoc routing protocols in the recent years. A detailed packet simulation comparative study of AODV, DSR and DSDV was reported in [5], highlighting that DSR and AODV achieve good performance at all mobility speeds whereas DSDV perform poorly under high speeds and high loads conditions respectively. Different from this study, proactive protocols have the best end-to-end delay and packets delivery fraction in [6] [7] but at the cost of higher routing load. Authors in [8] shown that AODV and DSR outperform OLSR at higher speeds and lower number of traffic streams and OLSR generates the lowest routing load. However, these studies are conducted under limited conditions as explained. And a study more limited to Qos was conducted in [8] and illustrated that DSR gets better performance of packet delivery fraction and routing overhead whereas OLSR shows the lowest end-to-end delay at lower network loads. On the other hand, authors in [9] and [10] discussed these protocols in more realistic environments and shown that AODV outperforms both DSR 978-0-7695-3336-0/08 $25.00 2008 IEEE DOI 10.1109/CSSE.2008.799 505
and DSDV, which proved that academic simulation does not always work well and reactive protocol is more realistic. Recently, Mbarushimana et al. [11] use OPNET simulator to contrast the two routing protocols (AODV and DSR) to the proactive approach OLSR and shown that OLSR get better performance, which is different from the conclusion of [5] and [8]. And the comparison in [11] focused on experiment. However, there is few comprehensive studies for ad hoc routing and we will focus on the routing strategies in different mobility models and evaluate the performance on NS [12] simulator. III. BACKGROUND In this paper, we picked up three ad hoc routing protocols: AODV (RFC-3561) [4], OLSR (RFC-3626) [3] and SRMP [16]. We shortly address the three protocols in the rest of this section. A. AODV (Ad-hoc On-demand Distance Vector) In AODV, when a source node has data traffic to send to a destination node, it first initiates a route discovery process. In this process, the source node broadcasts a Route Request (RREQ) packet. Neighbor nodes, which do not know an active route for the requested destination node, forward the packet to their neighbors until an active route is found or the maximum number of hops is reached. When an intermediate node knows an active route to the requested destination node, it sends a Route Reply (RREP) packet back to source node in unicast mode. Eventually, the source node receives the RREP packet and opens the route [15]. B. OLSR (Optimized Link State Routing) In OLSR, each node periodically constructs and maintains the set of neighbors that can be reached in 1-hop and 2-hops. Based on this, the dedicated MPR algorithm minimizes the number of active relays needed to cover all 2-hops neighbors. Such relays are called Multi-Point Relays (MPR). A node forwards a packet if and only if it has been elected as MPR by the sender node. In order to construct and maintain its routing tables, OLSR periodically transmit link state information over the MPR backbone. Upon convergence, an active route is created at each node to reach any destination node in the network [15]. C. SRMP (Source Routing-based Multicast Protocol) In SRMP, route selection takes place through establishing a multicast mesh, started at the multicast receivers, for each multicast session. SRMP is classified as a mesh-based protocol. A mesh (an arbitrary subnet) is built to connect multicast group members providing robustness against link failure due to topology changes and channel fading. To minimize the flooding scope, the Forwarding Group (FG) nodes concept is used in SRMP. In fact, a population of appropriate nodes is selected to cooperate to find the best route to deliver the packets to the destination. Four metrics are used in SRMP to select FG nodes: neighborhood association stability, link signal strength, battery life and link availability estimation. IV. ROUTING STRATEGIES There are different routing strategies behind routing protocols for mobile ad hoc networks. Using Biradar s classification [13], we classify them based on when and when the routing information is exchanged, how many nodes information need maintain and the scope of routing. We will focus on the routing strategies behind the three protocols (DSDV, OLSR and SRMP). A. Proactive vs. Reactive routing Proactive routing protocol attempts to maintain consistent, upto-date routing information from each node to every other node in the network. These protocols require each node to maintain one or more tables to store routing information, and they respond to changes in network topology by propagating updates routes through out the network in order to maintain a consistent network view. Different from proactive approach, Reactive routing protocol creates routes only when desired by source node. When a node requires a route to a destination, it initiates a route discovery process within the network. This process is completed once a route is found or all possible routes permutations have been examined. Once a route has been established, it is maintained by a route maintenance procedure until either the destination becomes inaccessible along every path from the source or until the route is no longer desired. B. Source Routing vs. Hop-by-hop routing Source Routing is a technique whereby the sender of a packet can specify the route that a packet should take through the network. With source routing, forwarding depends on the source of messages and the source node puts all the routing information on the header packet. The relaying nodes utilize this routing information and may decide to alter it in the packet to be forwarded. On the contrary, hop-by-hop routing is based on the fact that different nodes have different views of the network topology in ad hoc networks. This means, the route to the destination node is distributed in the next hop of the nodes along the path. The problem with hop-by-hop routing is that different nodes maintain different routing information and they may form a routing loop. C. Limited vs. Local broadcast Full broadcast, also named flood, intends to pass the source message to every node in ad hoc networks and the receiver nodes need to relay the message to other nodes without it. However, flooding storm may be formed with full broadcast, as even a small request may cost numberless messages interaction. So there is no routing protocols simply using full broadcast strategy. Limited broadcast, which relies primarily on replicating messages to enough nodes so the destination receives it, is more reasonable. As the TTL of the packet expires after n-hops, it only reaches all those nodes that are at most n-hops away from the original sender [3]. Compare with limited broadcast, local broadcast strategies only send messages within the senders reach and no receiver relays 506
them out of the sends reach. Nodes do not need to know all network topology and only focus on the local scope. OLSR belongs to this type. Multipoint Relaying (MPR) used in OLSR protocol is a classical local broadcast strategy. D. Multicasting vs. Unicasting In a unicast network, several copies of the data are sent from the source to each destination. Over a multicast network, data is sent from the source only once and the network must transmit the data to multiple destinations. OSLR and AODV use unicast to broadcast their packets. It will consume more bandwidth and energy for data copies transport. On the contrary, SRMP based on source routing multicasts its message to the destinations, for saving storage and power. TABLE I. SUMMARY OF ANALYSIS ON ROUTING STRATEGIES Protocol Philosophy Source Routing Broadcast Routes AODV Reactive No Ring Flood Unicast OLSR Proactive No local broadcast Unicast SRMP Reactive Yes Ring Flood Multicast V. PERFORMANCE EVALUATION & ANALYSIS The simulation mainly focuses on the performance of the routing strategies to react on the different scenarios in MANET. Because the three protocols (AODV, OLSR and SRMP) cover different routing strategies mentioned above, we will discuss these routing strategies based on the simulation results of the three protocols. In this paper, we evaluate the performance in terms of network throughput and the end-toend delay. A. Simulation Model and Scenarios The routing protocols were simulated on the platform NS-2 [12] with the CMU extensions by Johnson et al. [14], which is a discrete event simulator developed by the University of California at Berkeley and the VINT project. It provides substantial support for simulating multi-hop wireless networks complete with physical, data link, and medium access control (MAC) layer models on NS-2. In this study, we focus on two mobility models: Random Waypoint (RWP) model and Reference Point Group Mobility (RPGM) model. The mobility scenarios for Random Waypoint model are generated under NS-2, while the RPGM mobility scenarios are generated using Bonn-Motion tool [19]. The network parameters are shown in TABLE II. 1) RWP model In this model [17], all nodes are uniformly distributed around the simulation space. Movement of nodes has pause and motion periods. A mobile node selects a random destination and moves to that destination with a uniform speed. Upon reaching that destination it pauses for a certain pause time, and then selects another random destination performing the previous process. This is repeated along the simulation time. 2) RPGM Model This model [18] defines a random motion of a group of mobile nodes as well as a random motion of each mobile node within a given group. Each group has its own mobility behavior. Every node, within a specified group, follows a logical center reference point. The motion of the groups is explicitly defined by giving a motion path for the group. The path is viewed as a sequence of checkpoints along the path. The nodes belonging to a group are usually randomly distributed around the reference point. Both group s motion and motion of mobile nodes within a group follow the RWP model. TABLE II. NETWORK PARAMETERS Parameter Name Value Simulator NS-2 (2.31) Network Traffic Model CBR Network Area 500 x 500 meter Node Tx Distance 250 meter Data Packet Size 512 bytes payload RPGM Group Size 4 Sending Buffer 64 packets Simulation Time 100 seconds B. Simulation Results and Analysis We have two scenarios for simulation study: mobility (pause time) and network size. In our first simulation for mobility analysis, each packet starts its journey from a random location to a random destination with a fixed speed 20 m/s. Once the destination is reached, another random destination is targeted after a pause. The pause time, which affects the relative speeds of the packets, is varied from 0s to 100s and the simulations are run for 100 simulated seconds. Fig. 2 and Fig. 3 show the throughputs and end-to-end delay respectively in RWP model. In case of fixed nodes and pause time equals run time in the network, all three protocols perform well with a high throughput. But as the pause time decrease from 100s to 0s, the mobility increases and more packets are dropped due to unavailable routes. Low pause time means high pressure for packet forwarding and high mobility. So OLSR with proactive routing strategy faces serious challenge as more predefined data invalid and it is closer to AODV. Although SRMP is a source routing protocol, it benefits from its multicasting strategy and combination of single and multiple paths. Fig. 2 (b) shows that OSLR outperforms AODV and SRMP in RPGM model. It is also noticed that impact of RWP model on the throughputs outperforms that of RPGM model. RPGM promotes nodes movements in groups and the local broadcast strategy optimizes group communication. In fact, there is a higher probability for nodes in RPGM groups to be within the same multicast group, allowing more data packets reception within each group of nodes. When the pause time is 0 (highest mobility), SRMP outperforms OLSR and AODV remarkably. 507
0 10 80 90 100 Throughputs in RWP model Throughputs in RPGM model Throughputs ratio Throughput Ratio (a) (b) Figure 2. Performance on Throughput with pause time (20 nodes) Delay in RWP model 0.18 Delay in RPGM model 0.18 0.16 0.16 (c) (d) Figure 3. Performance on Aver End-to-End delay with pause time (20 nodes On the other hand, because proactive routing prepares node connections before requirement and pause time affects little for topology changes, so the average end-to-end delay (shown in Fig. 3) is stable and not high for pause time changes. So OLSR gets low delay with the pause time changes. On the contrary, reactive routing is affected much by pause time, as AODV. SRMP applies multicasting strategy and has a short response time. OLSR and SRMP both perform well in high mobility scenarios. When the pause time is 0 (highest mobility), the delay of AODV is remarkable. OLSR with local broadcast routing requires more control packets all the time, so it has pool throughput in high mobility scenarios. Fig. 3 shows that RWP has less impact on the delay compared to RPGM for all mobility cases. This returns to the behavior of RPGM model, allowing nodes movement in groups, each group following a random pattern. At this case, there is a high probability for receivers to follow different movement patterns creating different groups in their movements. The second simulation focuses on the effects of network size. It varies from 20 nodes to 70 nodes, with null pause time and max node movement speed 20 m/s. Fig. 4 and Fig. 5 show the performance of throughputs and end-to-end delay respectively on network size changes. SRMP gets high throughput in large networks as it is more possible to find the best sing-path with more neighbors and multicasting strategy can bring more efforts. Meanwhile, the SRMP delay is low and close to OLSR. Reactive and Source routing strategy in SRMP maintain effective data after each query and performs well. The deep reason is multicasting which involves more neighbors in large network. AODV outperforms OLSR when the nodes are 20, in small network. However, its throughput decreases in a linear fashion (shown in Fig. 4 (e)). As a reactive protocol, AODV requires more control packet for on-demand packet delivery in a large network. The delay is higher than SRMP and OLSR, as in AODV a destination replies to the least RREQ and the time cost for RREQ is high in ring-expiration flooding mechanism of AODV. We observe that OLSR has better performance in Fig. 4. In large networks, nodes may communicate mostly with physically nearby nodes. If local communication predominates, path lengths could remain nearly constant as the network grows, and so OLSR with local broadcasting can get better throughput. 508
RWP has less impact than RPGM model in large networks (as shown in Fig. 4). The denser the network, the more effective the group communication is. It is a tricky change that the throughput increase when network expand from 20 to 50 and drop down in a larger network (70 nodes), for both models. This observation is quite normal since throughput decrease in large networks for all protocols, but nodes less than 50 do not mean large network in NS-2 simulator. Throughput Ratio Throughputs in RWP model Delay in RWP model (g) Delay in RPGM model Thourghput Ratio 0.58 0.56 0.54 0.52 0.48 0.46 0.44 0.42 0.38 0.36 (e) Throughputs in RPGM model 0.62 0.62 (f) Figure 4. Performance on Throughput with (0s pause time) In Fig. 5, OLSR and SRMP both keep low delay in large networks. Less routing overhead helps SRMP save time. OSLR uses local broadcast strategy to optimize broadcast overhead. Indeed, SRMP causes fewer overheads thanks to its source routing approach saving the overhead needed to find the next hop. In addition, it exerts no periodic messages, uses the FG concept, which minimizes the flooding scope, and applies simple maintenance mechanisms making use of data transmission with no extra control overhead. It is observed that RPGM always exhibits fewer overheads compared to RWP as it uses no pause times during its motion pattern and the nodes movements are not completely random. This feature in RPGM model allows the source to allocate the receivers faster and allows the receivers to construct their mesh towards the source faster, implying less control overhead needed from both sides [16]. 0.58 0.56 0.54 0.52 0.48 0.46 0.44 0.42 0.38 0.36 (h) Figure 5. Performance on Aver End-to-End delay with (0s pause time) C. Summary Simulations in this paper are based on NS-2[12] different from OPNET in [11]. We have the following observations from the simulations above: It is an indiscreet conclusion in [11] that proactive protocols are better suited to CBR traffic, as the results in this paper showed that the proactive protocol SRMP gets high throughput and low delay in spite of the mobility model and mobility scenarios. Source routing strategy combines multicasting can get good performance. Proactive and reactive strategy can not determine the protocol performance alone. Although the experiments illustrated that OLSR gets high throughput and maintain low end-to-end delay in different scenarios, but scarifies its routing load. OLSR is better suited to group communication model RPGM. The results also showed that proactive protocol OLSR is inferior to reactive SRMP in terms of throughput in both mobility scenarios and network size, different from [11] showed. 509
VI. CONCLUSION In this paper, four routing strategies were discussed in the three routing protocols (AODV, OLSR and SRMP), based on NS-2 [12] simulations. To get a comprehensive evaluation report, the study was simulated in two mobility models: RWP and RPGM. The results showed that hop-by-hop routing is not absolutely superior to source routing. Different strategy combinations enforce source routing functions (e.g. in SRMP). Meanwhile, large network will benefit from local broadcast and multicasting strategies, as space locality is more important in this type networks. Proactive protocols perform better than reactive ones in high mobility scenarios, but not in large networks. Furthermore, the simulation showed it is the routing strategies behind protocols to determine the performance, such as local broadcast, hop-by-hop routing and multicasting strategies. So reactive protocol SRMP combines source routing and multicasting and outperforms OLSR in terms of throughput and get the same low delay as OLSR. In future, we will focus on how to get stable and acceptable performance in dynamic ad hoc networks by constructing virtual bone networks using local broadcasting strategy in OLSR. REFERENCES [1] IEEE, http://www.computer.org/. [2] Dow C R, Dow C R, Lin P J, et al, A study of recent research trends and experimental guidelines in mobile ad-hoc network, Advanced Information Networking and Applications, P72-771, 2005. [3] T. Clausen, P. Jacquet, "Optimized Link State Routing Protocol (OLSR)", IETF RFC 3626, October 2003. [4] C. Perkins, E. Belding-Royer, S. Das, quet, Ad hoc On-Demand Distance Vector (AODV) Routing, RFC 3561, July 2003. [5] J Broch, D. Maltz, D. Johnson, Y.C. Hu, and J. Jetcheva, A performance comparison of multihop wireless ad hoc network routing protocols, MOBICOM, P85-97, October 1998. [6] S. R. Das, R. Casta?eda, J. Yan, R. Sengupta, Comparative Performance Evaluation of Routing Protocols for Mobile Ad hoc Networks, Proceedings of the International Conference On Computer Communications and Networks (ICCCN), P153-161, 1998. [7] T. Clausen, P Jacket, L Viennot, Comparative study of Routing Protocols for Mobile Ad Hoc Networks, The First Annual Mediterranean Ad Hoc Networking Workshop. September 2002. [8] J. Novatnack, L. Greenwald, H. Arora. Evaluating Ad hoc Routing Protocols With Respect to Quality of Service, Wireless and Mobile Computing, Networking and Communications (WiMob), P205-212, Vol 3, 2005. [9] S. Jaap, M. Bechler and L. Wolf, Evaluation of Routing Protocols for Vehicular Ad Hoc Networks in City Traffic Scenarios, International Conference on ITS Telecommunications, France, 2005. [10] Ahmed A M, Mohamed O K, Performance Analysis of MANET Routing Protocols in the Presence of Self-Similar Traffic, Mohamed O K. Local Computer Networks, Proceedings 31st IEEE Conference, P801-807, 2006. [11] Mbarushimana C, Shahrabi A, Comparative Study of Reactive and Proactive Routing Protocols Performance in Mobile Ad Hoc Networks,. Advanced Information Networking and Applications Workshops (AINAW), P679-684, 2007. [12] Kevin Fall and Kannan Varadhan, editors. ns notes and documentation.the VINT Project, UC Berkeley, LBL, USC/ISI, and Xerox PARC, November 1997. Available from http://wwwmash.cs.berkeley.edu/ns. [13] Biradar R V, Patil V C, Classification and Comparison of Routing Techniques in Wireless Ad Hoc Networks, Ad Hoc and Ubiquitous Computing (ISAUHC), P7-12, 2006. [14] CMU Monarch Group, CMU Monarch extensions to the NS-2 simulator, Available from http://monarch.cs.cmu.edu/cmu-ns.html, 1998. [15] Jerome Haerri, Fethi Filali, Christian Bonnet, Performance Comparison of AODV and OLSR in VANETs Urban Environments under Realistic Mobility Patterns Institute Eurecomz Department of Mobile Communications, 2006. [16] H. Moustafa and H. Labiod, A performance Analysis of Source Routing-based Multicast Protocol (SRMP) Using Different Mobility Models, IEEE ICC2004, 20-24 June 2004 - Paris, France. [17] C. Bettstetter, H. Hartenstein, and X. Pérez-Costa, Stochastic Properties of the Random Waypoint Mobility Model: Epoch Length, Direction Distribution, and Cell Change Rate, ACM MSWiM 02, 2002. [18] X. Hong, M. Gerla, G. Pei, and C. Chiang, A group mobility model for ad hoc wireless networks, In proceedings of the ACM International workshop on Modeling and Simulation of Wireless and Mobile Systems (MSWiM), August 1999. [19] Institute of Computer Science IV, University of Bonn, BonnMotion a mobility scenario generation and analysis tool, URL: http://www.cs.unibonn.de/iv/bonnmotion/ 510