Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks to Achieve Good Scalability and Reliability

Transcription

1 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks to Achieve Good Scalability and Reliability MohammadReza EffatParvar 1, Mehdi EffatParvar, and Mahmoud Fathy 1 Informatics Services Corporation,Tehran, Iran effatparvar@gmail.com Computer Eng Department, Islamic Azad University Ardebil Branch, Ardebil, Iran mehdi_effatparvar@yahoo.com Computer & IT Eng Department of Iran University of Science & Technology, Tehran, Iran mahfathy@iust.ac.ir Abstract. An ad hoc network is a dynamically reconfigurable wireless network with no fixed wired infrastructure. The primary concerns in ad hoc networks are bandwidth limitation and unpredictable dynamic topology. Reliable Multicast plays a significant role in many applications of Mobile Ad Hoc Networks (MANETs). The On-Demand Multicast Routing Protocol (ODMRP) was designed for multicast routing in ad hoc networks. In this paper, we propose a cluster-based on demand multicast routing protocol to the lack of extension of flat multicast routing protocols in ad hoc networks of large scale. Simulation results show that using clustering base ODMRP improves network performance in terms of end-to-end delay and control packets. Also we propose a local recovery approach to design a reliable multicast algorithm. By this approach nodes can join to multicast groups in short time and data delivery will be increase, so ODMRP transmits the data with good reliability. Keywords: Ad hoc networks, Multicast, Scalability, Reliability, ODMRP, Cluster. 1 Introduction Ad hoc networks consist of mobile nodes that autonomously establish connectivity via multi hop wireless communications. In many ad hoc applications, nodes need to collaborate to achieve common goals and are expected to communicate as a group rather than as pairs of individuals (point-to-point). Multicast is a very useful and efficient means of supporting group-oriented applications, especially in mobile/wireless environments where bandwidth is scarce and equipment has limited power [14]. Recently, wireless mesh networks have attracted much attention [,4]. These networks typically have low maintenance overhead, high data rates, and are not energy constrained. There are several problems which are specific to ad hoc networks. One of the problems is decrease of management capability as the size of ad hoc networks becomes O. Gervasi et al. (Eds.): ICCSA 008, Part II, LNCS 507, pp , 008. Springer-Verlag Berlin Heidelberg 008

2 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks 447 large. The other problem is frequent change of the topology of ad hoc networks due to node movement. In order to cope with these problems, the clustering scheme such as the autonomous clustering scheme [5] and Least Clusterhead Change (LCC) scheme has been proposed. The election of clusterheads has been a topic of many papers as described in [7]. In all of these papers the leader election guarantees that no node will be more than one hop away from a leader. Furthermore, their time complexity is O(n), where n is the number of nodes in the network. Mesh-based multicast protocols [8], provide alternative paths and a link failure need not trigger the re computation of a mesh. Previous studies showed that meshbased protocols are robust against topology change and are more suitable than tree-based protocols [9]. ODMRP is an ad hoc multicast routing protocol based on a multicast mesh [8]. Because importance of reliability there are some reliable protocols in MANET; that we can classify them in three categories: The categories are, Automatic Retransmission Request (ARQ) based, gossip-based and Forward Error Correction (FEC) based. The protocols are RMA [11], RALM [10] and ReAct [1] are belong to the ARQ-based category. In this paper, we propose a new method for clustering ODMRP; also we propose a new local recovery approach in ODMRP. Our clustering algorithm divides the network to some group, that these clusters are 1-hop clustering. By using clustering method protocol overhead will be decrease and we can establish scalable network. Local recovery method achieves fast recovery when route breakage happens, so destination can connect to source in new route or the same route. Simulation results show that our approaches improve the performance of ODMRP, so we can use ODMRP as a scalable and reliable protocol. The reminder of this paper is organized as follows. Sections describes issues of ODMRP reliability and scalability. Section describes our new clustering mechanism. In section 4 we show our local recovery approach. Section 5 follows with the simulation results and concluding remarks are made in section 6. Issues of Reliability and Scalability in ODMRP ODMRP [8] is a best-effort mesh-based multicast routing protocol. Hence, a major performance evaluation criterion is the average miss ratio, which is calculated as the number of packets that each receiver did not receive divided by the total number of packets sent by the source, and averaged over the total number of receivers in the multicast group. In this section, we study how the average miss ratio of ODMRP is affected by mobility, session size, and the transmission range. The simulation environment consists of 0 hosts in a 700 m * 700 m area. A single source multicasts byte DATA packet to a multicast group. Network bandwidth is Mbit/sec. MAC protocol is The random waypoint mobility pattern is used with equal minimum and maximum speeds of mobility (i.e., mobile hosts move with a constant speed). Several values of constant mobility speeds between 0 m/sec. (i.e., stationary hosts) and 5 m/sec. are considered. Figure 1 shows the miss ratio of ODMRP with different numbers of members.

3 448 M. EffatParvar, M. EffatParvar, and M. Fathy Fig. 1. Average miss ratio versus speed of mobility for different numbers of members, transmission range is 00m At zero speed of mobility, the hosts are stationary and thus, the topology of the ODMRP multicast mesh is fixed. Hence, when a source multicasts a data packet, only the receivers that are in the mesh may receive it but other receivers that are far away from the mesh will not. As mobility increases, distant receivers may, while moving, reach the multicast mesh and hence start receiving some data packets causing the average miss ratio to decrease (as shown in Figure 1). However, as mobility increases further, there are more frequent disconnections in the multicast mesh and hence packet loss increases. For larger session sizes, there are more forwarding nodes in the multicast mesh, and therefore more redundancy in the mesh, leading to falling back on alternate multicast links when a multicast link fails and thus, reducing the average miss ratio. ODMRP always tries to include the shortest path within the forwarding group, so almost it has lower latency than the other algorithm. In this section, Figures show the performance metrics as functions of group size. With fixed pause time as 60 seconds, we have one multicast group of size from 0 to 00. Figure () shows the result for packet delivery ratio. As group sizes increases, ODMRP delivers more fractions of packets. The reason is that the forwarding mesh becomes more reliable with more redundant paths as it increases its size. In this section, we study the performance behaviors with respect to the horizontal scalability. We consider the following 6 scenarios: 7 by, 48 by, 6 by 4, 4 by 6, 18 by 8 and 1 by 1. Here, 7 by means multicast groups, and 7 members per group. Thus, in all scenarios, the total number of receivers is fixed to 144. There is one source for each group. The traffic demand remains equal in all scenarios. Figure-(a) shows the packet delivery ratio and the variance among the groups in the network. As the number of groups increases, performance of ODMRP shows quick drop to less than 10% for 1 groups. The size of meshes does not decrease proportional to the group sizes.

4 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks Delivery Ratio (%) ODMRP-0 (m/sec) ODMRP-0 (m/sec) Group Size Fig.. Packet delivery ratio versus group size. (Pause time is 60s, 1 group, 1 source, speed 0 and 0 m/s). 100 Delivery Ratio (%) ODMRP-0 (m/sec) ODMRP-0 (m/sec) Control Traffic in per Received Data Bit Number of Groups (a) Packet delivery ratio. 4 1 ODMRP-0 (m/sec) 18 ODMRP-0 (m/sec) Number of Groups (b) Bit Overhead Fig.. Performance versus group size. (Pause time is 60s, 1 source per group, speed 0 and 0 m/s).

5 450 M. EffatParvar, M. EffatParvar, and M. Fathy Figure-(b) shows the results for relative control bit overhead. The control traffic incurred by ODMRP increases dramatically with the increase in the number of groups and node speed. Clustering Mechanism Many clustering algorithms have been proposed, one of them is The Max-Min D- cluster algorithm [1], it creates clusters with a given radius, but there is no way to limit the maximum cluster size. On the other hand, using the algorithm presented in [], it is not possible to limit the radius for the clusters. Newly many clustering algorithms worked on multicast routing in Ad hoc. In [15] we can see one of clustering approach that is implemented on ODMRP and adopted an enhanced weighted clustering algorithm (EWCA) to manage hierarchically its motion nodes, and forms the forwarding group mainly based on clusterheads. In this section we propose our clustering mechanism. Our clustering mechanism has stages, at first stage we divide the network in some 1-hop clusters and in second stage we define some condition and we determine the role of clusterheads and the cluster members. Generally clustering algorithms have some disadvantages like overhead, needing the nodes stop during the algorithm execution and depending on underlying protocols. In our algorithm clusterheads do as gateways, so only the clusterheads transmit the Join-Request packets. In our algorithm all assumption are like ODMRP but we define some new fields in Join-Request packet in order to cluster the network at the beginning of simulation, also we define a new packet to find the new clusterhead during the simulation. With above assumption, our new algorithm, that we call SC-ODMRP is described as follows: a) In the beginning of simulation each source sends the Join-Request packet (see figure 4). b) On receiving the first Join-Request, each node determines the sender as the clusterhead of itself. Then the node puts the clusterhead address in Join-Request packet and broadcasts it (see figure 5 ). c) When the Broadcasted Join-Request is received by the original sender, and it knows that has been selected as clusterhead by one node. C F G J 1 E H A B D I K Fig. 4. Sending the first Join-Request by the source

6 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks 451 C F G J 1 E H A B I K D Fig. 5. Determine the clusterhead by first Join-Request d) Stages b and c will be repeated by all nodes. e) For the other Join-Request packets except the first, only two groups broadcast it: clusterhead and the nodes that receive the packet for the first time. f) Each node can receive the packets from any clusterhead and can change its clusterhead. g) If a node does not receive the Join-Request after a specific period (in our simulation 6s) sends a Find-Cluster packet to its neighbors. The node puts the multicast group address in the packet and sets the Find-Cluster status field. The status field may have three states: 0: determine that this Find-Cluster packet is a question packet to discover the clusterhead. 1: means that the Find-Cluster packet sender suggests itself as the clusterhead for the receiver. : it means that the sender has accepted the receiver as its clusterhead. So finding a new clusterhead is done by sending three Find-Cluster packets with different status field. (see figures 6 and 7 ). C F G J 1 E H A B I D FC (1) FC (1) K Fig. 6. node K does not have a clusterhead so after receiving the Find-cluster it can choice the clusterhead

7 45 M. EffatParvar, M. EffatParvar, and M. Fathy C F G J 1 E H A B I D FC (,D) K Fig. 7. Node K accepted D as its clusterhead C F G J 1 E H A B I D K Fig. 8. Node D can send the packets to K After determining the clusterhead, node can join to multicast group by sending Join-Reply packet like ODMRP. In our algorithm clusterhead is determining as forwarding group nodes; so they can transmit the data to destinations. In section 5 we show our simulation results completely. Simulation results show that SC-ODMRP improves the routing overhead and end to end delay. 4 Local Recovery Mechanism and Reliable Multicasting After analyzing reliability and scalability on ODMRP, in this section we proposed a new recovery mechanism for ODMRP. Many routing algorithms use recovery method

8 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks 45 to increase the performance and decrease the delay between source and destination. Usually local recovery methods are used in tree based algorithm but sometimes because importance of data and our application we need to use this method in mesh based routing too. There is a number of works currently done on local route recovery, which include WAR (Witness-Aided Routing Protocol) [16], ABR (Associatively- Based Routing Protocol) and RDMAR (Relative Distance Micro discovery Ad-hoc Routing Protocol). In this section we introduce a new recovery approach while it keeping low complexity of the algorithm to a minimal. Complexity is an important attribute because nodes in MANET are generally small portable devices with little processing power, thus, the local recovery scheme must also cater to this property. Because of many reasons in ODMRP it is possible that the discovered routes between source and destination break. Generally movement of node in MANET cause the route breakage, also if this movement occurred by one FG node (forwarding group node) it is harmful than the other cases because by movement of FG node one of the important connecting route between this FG and all other node connected to that FG (e.g. destination or FG nodes) will break so they can not receive any packet until other route is discovered. So mostly we considered this case in our algorithm. C Destination Source A (FG) B (FG) M (FG) L (FG) (FG): forwarding group Fig. 9. Local Recover approach in our algorithm With reference to Figure 9, if the direct link (AB) breaks off, there should exist, in most cases, some indirect route from A to the original next node B through some neighbor node C. In these situations, if a request packet is sent out to find the original next hop or other node which is at the further part of the original route with limited timeto-live (e.g. hops), the possibility of repairing the current route should be high and the overhead should be much lower than using end-to-end global recovery. With the above assumptions, a new algorithm for the proposed this mechanism is described as follows: When an intermediate FG node discovers that the link to the next hop has broken it would: a) Save the data packet into local repair data buffer, set a timeout timer b) Send out a two-hop wide local recovery request, containing a sequence of nodes including all the further nodes on the original route from A till the destination node D.

9 454 M. EffatParvar, M. EffatParvar, and M. Fathy Any node, on receiving the local recovery request, will check whether it is on the node list, and if it is, would send back a local recovery reply c) On receiving local recovery request from an intermediate node, A salvages the data packet at the local repair buffer and transmits it using the repaired route. d) If no reply received after the local repair timer expires, the data packet is dropped, and an error message is send back to the source from A, the source may then initiate a new round of end-to-end route discovery. For doing the local recovering method we changed the Join-Reply packet in ODMRP, so each receiver generates a Join-Table packet that is broadcast to its neighbors. The Join-Table packet contains the multicast groups address sequence of (source address, next hop address) pairs and a count of the number of pairs. In one pair the source address and the next hop address is the same and it is equal with receiver address. When a node receives a Join-Table, it checks if the next node address of one of the entries matches its own address. If it does the node realize that it is on the path to the source and thus becomes a part of the forwarding group for that source by setting its forwarding group flag. It then adds its own Join-Table to last received Join-Table and broadcast it. By this method each FG node knows the other node on the path between itself and destination. So FG node can recover the breakage route in suitable time with minimal complexity and without need the source. We call our algorithm LR-ODMRP and in next section we will compare it with ODMRP. 5 Simulation Results and Performance Evaluation The simulation code was implemented within the Global Mobile Simulation (Glo- MoSim) library [1]. In the local recovery approach simulation, we modeled a network of 50 mobile hosts placed randomly within a m area. Radio propagation range for each node was 50 meters and channel capacity was Mbit/sec. Each simulation runs for 00 seconds of simulation time. The IEEE Distributed Coordination Function was used as the medium access control protocol. We used Constant Bit Rate as our traffic. The size of data payload was 56 bytes. the source sends data at the rate of 0 packets/second, also We used Random-Way point as mobility model. For clustering mechanism we modeled the network by increasing the number of nodes and the speed. Nodes placed randomly within m area. Radio propagation range for each node was 76 meters and channel capacity was Mbit/sec.The size of data payload was 51 bytes and MAC protocol is IEEE and simulation time is 00s. Figure 10 show the protocol reliability varies with mobility (node speed) in our local recovery approach. This comparing shows that at lower speed the difference in packet delivery ratio is not outstanding. In both case increased mobility requires that forwarding group member be updated more frequently also by increasing mobility causes frequent link change so by recovery method we can increase the packet delivery ratio. Figure 11 plots control overhead per data byte transferred as a function of mobility. Because of recovery packet in LR-ODMRP, this method slight increase overhead at higher speed.

10 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks Packet Delivery Ratio Vs Mobility: 5 Senders, 0 Receiver: 50 kbps Packet Delivery Ratio (%) ODMRP LR-ODMRP Velocity (km/hr) Fig. 10. Packet delivery ratio as a function of node mobility Control Bytes / Data Byte Routing Overhead Vs Mobility: 5 Senders, 0 Receiver: 50 kbps Velocity (km/hr) ODMRP LR-ODMRP Fig. 11. Routing overhead as a function of node mobility Control Bytes / Data Bytes ODMRP SC-ODMRP Number of Node Fig. 1. Routing overhead as versus of number of nodes ( groups, pause time 10s, speed 10m/s)

11 456 M. EffatParvar, M. EffatParvar, and M. Fathy ODMRP has poor performance even at low mobility. LR-ODMRP act like ODMRP but it can recover the link breakage and data packet drops. In our clustering mechanism as you see in figure 1 we decreased the routing overhead. By increasing the number of nodes SC-ODMRP has better act. Also in our simulation as you see the figure 1 SC-ODMRP has less end to end delay, but in some cases because of existing direct path between sources and destinations ODMRP end to end delay is less than our algorithm. 8 end-to-end Delay (ms) 6 4 ODMRP SC-ODMRP ` Number of Node Fig. 1. End to end delay as versus number of nodes ( groups, pause time 10s, speed 10m/s) Our simulation showed that in most cases SC-ODMRP data delivery ratio is like ODMRP (figure 14 ). 100 packet delivery ratio (%) Number of Node ODMRP SC-ODMRP Fig. 14. Data delivery ratio as versus number of nodes ( groups, pause time 10s, speed 10m/s) But sometimes because of delay on sending the Find-cluster packet destination loss the data packets.

12 Improvement of on Demand Multicast Routing Protocol in Ad Hoc Networks Conclusion We introduced new techniques for clustering and local recovery. These techniques were applied to ODMRP and evaluated via several simulation scenarios. Our simulation showed that LR-ODMRP is more reliable than the ODMRP; also it has good performance when we increase the speeds of mobility and multicast group member, we can use LR-ODMRP in large scale topology and we can increase the packet delivery ratio and reliability. SC-ODMRP improves the network overhead in comparison of ODMRP by creating clusters and restricts the packet broadcasting. For the future works we intend to compact our local recovery method with clustering algorithm and we will extend it on other multicast routing protocols. References 1. Amis, D., Prakash, R., Vuong, T.H.P., Huynh, D.T.: Max-Min D-Cluster Formation in Wireless Ad Hoc Networks. Proceedings of IEEE, 41 (March 000). Banerjee, S., Khuller, S.: A Clustering Scheme for Hierarchical Control in Multi-hop Wireless Networks, Technical Report CS-TR-410, University of Maryland, College Park (February 000). MIT Roofnet, 4. Mesh Networks Inc., mesh networks technology overview, 5. Ohta, T., Inoue, S., Kakuda, Y.: An adaptive multihop clustering scheme for highly mobile ad hoc networks. In: Proc. 6th IEEE International Symposium on Autonomous Decentralized Systems (ISADS 00), pp (00) 6. Darehshoorzadeh, A., Dehghan, M., Jahed Motlagh, M.R.: Quality of Service Support for ODMRP Unicast Routing in Ad hoc Networks. In: International Conference on Digital Telecommunications (ICDT 006), Côte d Azur, France, August 9-1 (006) 7. Baker, D.J., Ephremides, A.: The Architectural Organization of a Mobile Radio Network via a Distributed Algorithm. IEEE Transactions on Communications, (1981) 8. Gerla, M., Lee, S.-J., Su., W.: On-demand multicast routing protocol (ODMRP) for ad hoc networks, IETF Internet Draft, draft-ietf-manet-odmrp-0.txt (000) 9. Lee, S., Su, W., Hsu, J., Gerla, M., Bagrodia, R.: A performance comparison study of ad hoc wireless multicast protocols. In: INFOCOM 000 (March 000) 10. Tang, K., Obraczka, K., Lee, S.-J., Gerla, M.: A reliable, congestion-control led multicast transport protocol in multimedia multi-hop networks. In: The 5th International Symposium on Wireless Personal Multimedia Communications, pp (00) 11. Gopalsamy, T., Singhal, M., Panda, D., Sadayappan, P.: A reliable multicast algorithm for mobile ad hoc networks. In: Proceedings of ICDCS (00) 1. Rajendran, V., Yi, Y., Obraczka, K., Lee, S.J., Tang, K., Gerla, M.: Reliable, Adaptive, Congestion-Controlled Adhoc Multicast Transport Protocol: Combining Source-based and Local Recovery, UCSC Technical Report (00) 1. UCLA Parallel Computing Laboratory, GloMoSim: A scalable simulation environment for wireless and wired network system 14. Iiyas, M.: The Handbook of Ad Hoc Wireless Networks. CRC Press, Boca Raton (00) 15. Sun, X., Liu, W., Zhang, Z., Zhao, Y.: CODMRP: Cluster-Based On Demand Multicast Routing Protocol, WiCOM (September 006) 16. Aron, I.D., Gupta, S.K.S.: A Witness-Aided Routing Protocol for Mobile Ad Hoc Networks with Unidirectional Links. In: Leong, H.V., Li, B., Lee, W.-C., Yin, L. (eds.) MDA LNCS, vol. 1748, pp. 4. Springer, Heidelberg (1999)