LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009
Performance Evaluation of Multicast Routing Protocols in MANET Guntupalli Lakshmikanth1, Sandeep Patel2, and Apurva Gaiwak3 1
Vignan University, Guntur, India. 2 Mars Telecom, India. 3MITM/Electronics Department, Indore, India Email:
[email protected],
[email protected]
Abstract— The advent of ubiquitous computing and proliferation of portable computing devices have raised the importance of mobile and wireless networking. A major challenge lies in adapting multicast communication to environments where mobility is unlimited and failures are frequent. This paper investigates the performance of multicast routing protocols aimed specifically at fully Mobile Ad Hoc Networks (MANET), the multicast protocol studied in this work are, On-Demand Multicast Routing Protocols (ODMRP), Adaptive Demand Driven Multicast Routing Protocol (ADMR) and Flooding in different mobility conditions e.g. Uniform, Manhattan, Exhibition, Random way point and Battlefield. This paper demonstrates that difference in protocol mechanics and different mobility patterns lead to performance differentials.
node stores the sequence number of the last packet it receives for each multicast group.[2]. 2. On Demand Multicast Routing Protocol (ODMRP): ODMRP[3] is a mesh-based demand-driven multicast protocol. A source periodically builds a multicast tree for a group by flooding a control packet throughout the network. The member nodes of the group respond to this control packet and help the source to establish multicast tree. The nodes which are on the tree use soft state, meaning their status as forwarders for a given group times out if not refreshed. Because the source rebuilds the tree periodically, the set of forwarders at any one time actually forms a mesh, providing robustness for the mobile receivers. 3. Adaptive Demand Driven Multicast Routing Protocol (ADMR): ADMR[4] creates source-specific multicast trees, using an on-demand mechanism that only creates a tree if there is at least one source and one receiver active for the group. Unlike ODMRP, receivers must explicitly join a multicast group. Sources periodically send a network wide flood, but only at a very low rate in order to recover from network partitions. In addition, forwarding nodes in the multicast tree may monitor the packet forwarding rate to determine when the tree has broken or the source has become silent. If a link has broken, a node can initiate a repair on its own, and if the source has stopped sending then any forwarding state is silently removed. Receivers likewise monitor the packet reception rate and can re-join the multicast tree if intermediate nodes have been unable to reconnect the tree.
Index Terms— MANET, ODMRP, ADMR, Flooding, Uniform, Manhattan, Exhibition, Random way point and Battlefield.
I. INTRODUCTION A mobile ad hoc network (MANET)[1] consists of a collection of dynamic nodes with sometimes rapidly changing multihop topologies with bandwidth constrained wireless links. Since each node has a limited transmission range, not all messages may reach all the intended hosts. To provide communications through the whole network the traffic from a source to destination path could be relayed through several intermediateneighboring nodes. Unlike typical wire line routing protocols, ad hoc routing protocol must address a diverse range of issues [1]. The goal of MANET is to extend mobility into the realm of autonomous, mobile, wireless domains, where a set of nodes form the network routing infrastructure in an ad hoc fashion. The multicast service is critical in applications characterized by the close collaboration of teams (e.g. rescue patrol, battalion, and scientists) with audio and video conferencing requirements and sharing of text and images.
III. SIMULATION METHODOLOGY A. Evaluation Metrics In the simulations presented in this paper the following parameter are analyzed to study the effects of mobility on each of the multicast routing protocols: Throughput: The ratio of the number of packets received to the number of packets sent. Delay: The difference between the time when the packet is sent by the source and when it is received by a receiver. Transmission Overhead: It is the ratio of the number of data messages transmitted (originated or forwarded) to the number of data messages received.
II. MULTICAST ROUTING PROTOCOLS This paper presents the following multicast routing pprotocols performance: Flooding, ODMRP and ADAMR. 1. Flooding: Flooding approach uses a simple protocol in which each node receiving a packet for a group first checks whether it is a duplicate and, if not, forwards the packet by retransmitting it. To check for duplicates, each 184 © 2009 ACADEMY PUBLISHER
LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009
A. For FLOODING Flooding attains very high throughput at the expense of high transmission overhead (Figure 1 and Figure 2). This is due to each node forwarding every non-duplicate packet it receives. Note that the transmission overhead of 7 is a worst-case scenario because there are 49 nodes forwarding every packet and only 7 receivers. From (Figure 1) Battlefield model results in a relatively low throughput for flooding, which is 85% as compared to other mobility models whose throughput is more than 95%. The reason is that all packets are flooded. Flooding also establishes a lower bound on delay for each of the mobility models (Figure 3). Delay is 60% higher for other mobility models than Battlefield and Exhibition because group members have a higher likelihood of being near the source (i.e. if the source is the group leader, following the same leader or visiting the same center). Delay is higher for Uniform and Manhattan models because nodes are likely to be both well connected and spread out in the entire field.
25
Randomway Point Uniform Manhattan Exhibition Battlefield
D e la y ( m illis e c )
20
15
10
5
0 5
10
15
20
25
30
35
40
45
50
Speed (m/sec)
Figure 3. Delay for Flooding
B. For ODMRP For ODMRP[5], throughput (Figure 4) depends on the model and the number of link changes roughly predicts the ordering from worst-to-best. Throughput for Exhibition model is 23% higher than from Battlefield model, despite a similar number of link changes, because of its much lower reachability. Throughput for the Uniform model is the only exception to this ordering which is 12% lower than Exhibition model and this can be explained by its lower node density. ODMRP is significantly better than flooding because of its ability to achieve good throughput with much lower transmission overhead (Figure 5). For ODMRP, approximately 2.2 packets are forwarded for every packet received. ODMRP could have very high throughput by increasing the join query rate, but then this becomes flooding at very high rates, with a corresponding increase in transmission overhead. For both transmission overhead and delay (Figure 6), the ordering among models is the same as for flooding. Delay for group-based mobility is 47% lower than the Uniform mobility model whose delay is the highest. The reason for this is it correlates well with node density, as Group based mobility results in group members having a higher likelihood of being near the source, which can be expected to reduce delay and transmission overhead. 100 90 80 70
T h ro u g h p u t(% )
IV. RESULTS
30
60
Randomway Point Uniform Manhattan Exhibiton Battlefield
50 40 30 20 10 0 5
10
15
20
25
30
35
40
45
50
Speed (m/sec)
Figure 4. Throughput for ODMRP 10 9
Randomway Point Uniform Manhattan Exhibition Battlefield
8
T r a n s m is s i o n O v e r h e a d
Control Overhead: The ratio of the number of control messages originated or forwarded over the combined total of data and control messages originated or forwarded. B. Methodology Simulations use GloMoSim-2.03, which is extended to include the Uniform, Manhattan, Exhibition, and Battlefield mobility models[5] along with multicast protocols ADMR and Flooding. The ODMRP implementation provided in the GloMoSim[6][7] distribution was also used. Each simulation was run with 50 nodes, randomly placed over a square field whose length and width is 1000 meters. The multicast traffic was generated through three multicast groups, each consisting of 7 receivers. Each multicast source uses a Constant Bit Rate (CBR) flow, transmitting a 64 byte packet every 250 milliseconds. IEEE 802.11 MAC protocol, with free-space radio signal propagation, with a 2Mbps channel were used. Each simulation was run for 600 seconds and averaging was done by running 10 simulations for each data point.
7 6 5 4 3 2 1 0 5
10
15
20
25
30
35
40
45
50
Speed (m/sec) 100 90
Figure 5. Transmission Overhead for ODMRP
80
T h r o u g h p u t( % )
70
The increased efficiency of ODMRP results in added control overhead, which was absent in case of flooding (Figure 7). The high values of overheads are due to the combination of low traffic rate (4 packets per second) and periodic flooding (once per 3 seconds). With higher traffic rates, the percentage of overhead becomes much lower. The ordering of models in this graph is again similar to that of node density, With ODMRP.
60
Randomway Point Uniform Manhattan Exhibition Battlefield
50 40 30 20 10 0 5
10
15
20
25
30
35
40
45
50
Speed (m/sec)
Figure 1.Throughput for Flooding 10 9
T r a n s m is s io n O v e r h e a d
8 7 6 5
Randomway Point Uniform Manhattan Exhibition Battlefield
4 3 2 1
Randomway Point Uniform Manhattan Exhibition Battlefield
14
0 5
10
15
20
25
30
35
40
45
50
12
D e la y ( m illis e c )
Speed (m/sec)
Figure 2. Transmission Overhead for Flooding.
10
8
6
4
5
10
15
20
25
30
35
Speed (m/sec)
Figure 6. Delay for ODMRP
185 © 2009 ACADEMY PUBLISHER
40
45
50
LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 2, November 2009 50
100
45
90
Randomway Point Uniform Manhattan Exhibition Battlefield
T r a f f icO v e r h e a d
70 60
40 35
T r a f f i c O v e r h e a d
80
50 40 30
30 25 20
Randomway Point Uniform Manhattan Exhibition Battlefield
15 10
20
5
10
0 5
0 5
10
15
20
25
30
35
40
45
10
15
20
25
30
35
40
45
50
Speed (m/sec)
50
Speed (m/sec)
Figure 13. Control Overhead for ADMR
Figure 7. Control Overhead for ODMRP
C. For ADMR The most sophisticated of all the three protocols studied, ADMR is able to maintain high throughput for nearly all of the mobility models (80 - 90%) even as speed increases (Figure 8). This is due to two mechanisms in ADMR. First, forwarding nodes are able to initiate local repair of the multicast tree when they determine that packet loss is occurring. Second, receivers experiencing high packet loss can ask ADMR to switch to flooding. For some models, both Flooding and ODMRP are able to achieve higher throughput than ADMR at low speeds. This could indicate that local repair can be used more efficiently to recover from loss at low speed. The consequence of performing adaptive flooding is that this increases transmission overhead for ADMR when speed is increased (Figure 9). As with ODMRP and Flooding, the relative performance of the mobility models correlates to node density for both transmission overhead and delay (Figure 10). ADMR have slightly higher delay of 4% than ODMRP; this is due to the increased number of control messages in ADMR, which may lead to collisions and retransmissions at the MAC layer. Control overhead for ADMR may decrease as the speed increases, depending on the mobility model (Figure 11). This is because ADMR switches to flooding, which decreases the amount of control traffic due to local repair and member adaptation to loss. This trend is not as evident with the group based mobility models because flooding in areas of high node density can lead to more collisions and hence more control traffic (when nodes try to recover from the resulting loss).
VI CONCLUSION In this work performance of multicast routing protocols namely Flooding, ODMRP and ADMR in different mobility models, which are named as Uniform, Manhattan, Exhibition, Random way point and Battlefield was studied. Regardless of the mobility model, ODMRP performance degrades as speed increases, whereas ADMR is able to maintain throughput greater than 80%. ADMR is able to maintain high throughput because (a) forwarding nodes are able to initiate local repair of the multicast tree and (b) receivers experiencing high packet loss can ask ADMR to switch to flooding. For both ODMRP and ADMR, the transmission overhead, control overhead, and delay vary according to the mobility model. Group-based mobility models, which lead to higher node density, result in a greater chance that multicast group members will be located near the source. This leads to a savings in transmission overhead and delay. High density also decreases control overhead for ODMRP, since JOIN REPLY messages travel fewer hops. For ADMR, however, control overhead increases with density. This happens because ADMR switches to flooding more frequently whenever there is congestion of packets. The results in this work indicate that characterizing link variations and density fluctuations for any user movement is crucial towards understanding routing performance. REFERENCES [1] S. Corson and J. Macker. “Mobile Ad hoc Networking (MANET): Routing Protocols Performance Issues and Evaluation Considerations”, RFC 2501, IETF, January 1999. Available at http:// www.ietf.org/rfc/rfc2501.txt [2] Tony Ballardie, Paul Francis, and Jon Crowcroft. “Core Based Trees (CBT): An architecture for Scalable InterDomain Multicast Routing”, In Proceeding of ACM SIGCOMM ’93, San Francisco, pp. 85-95, October 1993. [3] Demand Multicast Routing Protocol (ODMRP) for Ad Hoc Networks, INTERNET-DRAFT, IETF MANET Working Group, Sung-Ju Lee, William Su, Mario Gerla, University of California, Los Angeles, January 2000 . [4] The Adaptive Demand-Driven Multicast Routing Protocol for Mobile Ad Hoc Networks (ADMR), INTERNETDRAFT, IETF MANET Working Group, J. Jetcheva and D.B. Johnson [5] X. Hong, M. Gerla, G. Pei, and C. Chiang. A Group Mobility Model for Ad Hoc Wireless Networks. In ACM MSWiM, August 1999. X. Hong, M. Gerla, G. Pei, and C. Chiang. A Group Mobility Model for Ad Hoc Wireless Networks. In ACM MSWiM, August 1999. [6] Lokesh Bajaj, Mineo Takai, Rajat Ahuja, Ken Tang, Rajive Bagrodia, Mario Gerla,“GloMoSim: A Scalable Network Simulation Environment” [7] http://pcl.cs.ucla.edu/projects/glomosim/
100 90 80
T h ro u g h p u t(% )
70 60 50
Randomway Point Uniform Manhattan Exhibiton Battlefield
40 30 20 10 0 5
10
15
20
25
30
35
40
45
50
45
50
Speed (m/sec)
Figure 10. Throughput for ADMR 10 9
T ra n s m is s io nO v e rh e a d
8 7
Randomway Point Uniform Manhattan Exhibition Battlefield
6 5 4 3 2 1 0 5
10
15
20
25
30
35
40
Speed (m/sec)
Figure 11. Transmission Overhead for ADMR 30
D e la y ( m il lis e c )
25
Randomway Point Uniform Manhattan Exhibition Battlefield
20
15
10
5
0 5
10
15
20
25
30
35
40
45
50
Speed (m/sec)
Figure 12. Delay for ADMR
186 © 2009 ACADEMY PUBLISHER
