A Cooperative Framework for Reliable Multicast Forwarding in Mobile Ad hoc NETworks

Transcription

1 IEEE International 30th International Conference Conference Distributed on Distributed Computing Computing Systems Systems Workshops Workshops A Cooperative Framework for eliable Multicast Forwarding in Mobile Ad hoc NETworks Giuseppe Araniti, Massimo Condoluci, Antonella Molinaro, Antonio Iera ATS Lab., DIMET Dep., University Mediterranea of eggio Calabria, Italy {araniti, antonella.molinaro, antonio.iera}@unirc.it Abstract In this paper an innovative framework based on the cooperation between MAC and routing protocols is proposed with the purpose of improving the reliability of multicast transmissions in a Mobile Ad hoc NETwork (MANET). The proposed framework adds new features to the On-Demand Multicast outing Protocol (ODMP) and to the IEEE MAC layer. This choice allows to exploit the mesh-based nature of the ODMP forwarding scheme, which offers higher reliability and connectivity among multicast members compared to other multicast routing protocols proposed for MANETs. Simulation results have shown the advantages achieved by the proposed solution compared to the straight forward application of legacy ODMP when coupled with either the standard IEEE MAC or the Multicast aware MAC Protocol (MMP). Several scenarios have been analyzed by varying the multicast group size, the node mobility, and by properly tuning the protocols parameters (e.g., the maximum number of MAC layer retransmissions). In the various examined scenarios, the proposed solution improves the performance in terms of successfully conveyed multicast packets with a negligible increase in the signaling overhead necessary to guarantee reliable data delivery. Keywords-Ad-hoc networking; multicast; mesh routing; MAC reliable; multihop. I. INTODUCTION Multicasting [1], by definition, is the transmission of Data packets to a group of terminals identified by a single IP destination address and is the natural solution to data delivery needs of group-oriented applications. This justifies the increasing attention of researchers to the multicast traffic handling issue, also in Mobile Ad hoc NETwork (MANET) [2], [3] scenarios. Difficulties in deploying effective solutions for multicast delivery in MANETs are due to the typical nature of such networks: neither a fixed infrastructure nor a central administration is present; nodes may move arbitrarily, thus causing frequent changes in the network topology; moreover, bandwidth and battery power are limited. Useless to say, a key role in guaranteeing a high-level performance to multicast traffic is played by the routing protocol [4]. Many multicasting protocols have been designed for MANET environments. The reference protocol addressed in this paper is On-Demand Multicast outing Protocol (ODMP) [5]; a mesh-based forwarding scheme, which offers higher reliability and connectivity with respect to other mesh- and tree-based multicast routing protocols [6], [7]. At the MAC layer, the typical approach to multicast traffic handling is to exploit the broadcast nature of the wireless channel and to map IP layer multicasting over broadcasting at the radio layer. Multicast packets are, thus, forwarded as one-hop broadcast packets. The basic Carrier-Sense Multiple Access with Collision Avoidance (CSMA/CA) protocol in IEEE [8] does not provide any reliable multicast transmission; in fact neither explicit ACK nor TS/CTS (equest To Send/Clear To Send) mechanism is allowed for either acknowledging the reception of multicast Data frames or reserving the radio channel for transmission, respectively. In order to differentiate broadcast and multicast traffic transmissions at the MAC layer and to guarantee a sort of protection against the hidden terminal problem, we propose a new multicast MAC policy named Dynamic Mesh based Multicast MAC Protocol (D3MP). As it will be discussed in section III-B, D3MP is an extension of the Multicast aware MAC Protocol (MMP) [9], taken as a reference in the present work. The joint use of ODMP and D3MP forces the Data packet forwarding scheme to assume a tree configuration instead of a mesh one. In order to re-establish a mesh configuration, a cooperation framework between routing and MAC layers is required. In particular, at the routing layer, a new version of ODMP, named Enhanced Xcasting- ODMP (EX-ODMP), is implemented. As discussed later, EX-ODMP takes advantage of the possibilities (i) to determinate the list of nexthops that will be delivered to D3MP protocol for multicast MAC transmission (Xcasting- ODMP) and (ii) to indicate, in the EPLY message, a list of reverse hops to the purposes of maintaining a mesh configuration (Enhanced X-ODMP). Hence, the cooperation between MAC (D3MP) and routing (EX-ODMP) protocols improves the end-to-end reliability of ODMP. Furthermore, this choice reduces the negative effects of MAC broadcast transmissions (as in legacy IEEE ) and, at the same time, assures reliable multicast packets delivery. The remaining part of the paper is organized as follows. Section II provides a brief overview of ODMP and MMP protocols and related works. In Section III our cooperative /10 $ IEEE DOI /ICDCSW /ICDCSW

2 Table I QUALITATIVE COMPAISON BETWEEN MULTICAST OUTING POTOCOLS Protocols ODMP CAMP SMP MAODV AMIS AMoute Standard MAC Protocol IEEE IEEE IEEE IEEE IEEE IEEE Configuration Mesh Mesh Mesh Tree Tree Tree outing Approach eactive Proactive eactive eactive Proactive Proactive Loop Free Yes Yes Yes Yes Yes No oute Setup Overhead Moderate Moderate Moderate High High High oute Maintenance Overhead Moderate High Moderate High Highest Highest eliability High Moderate Moderate Low Low Low Traffic Concentration Low High Low Highest Highest Highest Dependency of unicast protocol No Yes No No No Yes framework is introduced. The main results obtained by the simulation campaigns are the focus of Section IV. Conclusive remarks are presented in Section V. II. ESEACH BACKGOUND In the next sub-sections the research background related to our proposal is described. In particular, the main functionalities of ODMP and MMP are illustrated with the aim of allowing readers to better understand the novelties introduced in Section III. A. Multicast outing Protocols ODMP [5] is a mesh-based routing protocol that guarantees several paths among source and destination nodes. At the moment, it is widely accepted as the most promising multicast routing protocol [6], [7]. Indeed, ODMP shows a very attractive behaviour in terms of reliability compared to others, either mesh-based (i.e., Core-Assisted Mesh Protocol (CAMP) [10], Source outing-based Multicast Protocol (SMP) [11]) or tree-based multicast routing protocols (i.e., Ad hoc Multicast outing protocol utilizing Increasing idnumbers (AMIS) [12], Ad hoc Multicast outing protocol (AMoute) [13], Multicast Ad-hocOn-Demand Distance Vector (MAODV) [14]). CAMP and AMoute assumes the availability of routing information from a unicast routing protocol. This feature, in a mobility scenario, can require a period of network reconvergence to a subset of destinations. As a result, these nodes during the re-convergence interval will be marked as unreachable. Moreover, AMoute is not a loop-free protocol and this is critical for performance because it may cause serious congestion. In SMP stable paths and higher battery life are provided. SMP is shown to assure better quality of links and to minimize the possibility of link failures, nevertheless the protocol shows an interesting behaviour only when nodes tends to be stationary, while in other cases the reliability is moderate. AMIS builds a shared delivery tree to support a multicast session. Node movements and tree breaks are detected by a beaconing mechanism. Neighbors are considered to have moved away if three consecutive beacons are not received.the Data packets delivery follows a single path (tree configuration) and during delivery many multicast packets can be lost. MAODV is the most performing tree based routing protocols [6], and it presents the lowest routing overhead (with respect to the tree-based approach). Compared with ODMP, it well works in stationary scenarios, but it presents lower reliability in a mobility scenario. Table I provides a qualitative comparison between ODMP and other multicast routing protocols. As shown in Figure 1, ODMP establishes a mesh among its members for each multicast group through a Forwarding Group () [15] that is responsible for forwarding Data packets from Source (S) towards multicast eceivers (). The protocol uses two tables: Forwarding Group Table, for maintaining information about the Forwarding Group; outing Table, created for the management of JOIN QUEIES/EPLIES. ODMP implements on-demand routing techniques; this means that a route to destination is found only when it is needed. A multicast source sends a Control packet, called JOIN QUEY, following the reception of the first Data packet to forward from the application, when no route and group membership is known. An intermediate node rebroadcasts the QUEY packet only if it is not duplicated. When a multicast group member receives a JOIN QUEY packet, it waits 25 ms for likely additional QUEIES, and then generates and delivers a JOIN EPLY. S Figure 1. JOIN EPLY JOIN QUEY Mesh establishment in ODMP. Forwarding Group Member

3 The EPLY is broadcasted, but the chosen reverse hop toward the source is included in the packet (in our proposal we will modify this procedure). Only stations on the reverse path accept the EPLY, other nodes will drop it. The node which receives a EPLY becomes a Forwarding Group member and it sets a flag in the Forwarding Table. This node rebroadcasts the EPLY indicating its reverse hop. The procedure continues until the EPLY reaches the source of traffic. No storing of the next hop address is required in this case, because each node utilizes multicast address at the network layer. In such a way, ODMP will generate the routing tree as shown in Figure 1 while the IEEE broadcast transmission protocol will guarantee a mesh forwarding of the Data packets. Indeed, a generic node could receive several copies of the same Data packet from different Forwarding nodes. All these packets will be processed and accepted by the MAC layer (broadcast transmission), while the ODMP protocol will discard the duplicated copies (multicast transmission). A mechanism called soft state is used to refresh the mesh. Until Data packets are available from the application, source node transmits a QUEY periodically (the period is 3 sec). A forwarder not receiving either Data or EPLY packets for more than 9 seconds leaves the forwarding group. For more details on ODMP the reader can refer to [16]. B. Multicast MAC Protocols Several protocols have been proposed in literature [9], [17], [18], in order to support reliable multicast transmissions at MAC layer. Among those, we consider in this subsection the most performing ones and only those protocols characterized to be an extension of IEEE [8]. MMP [9] supports multicast transmissions at the MAC layer through the introduction of a new feature: the Extended Multicast Header (EMH). Before sending a Data frame, with the help of the routing layer, each transmitting node creates the EMH and adds it to the Data frame header. EMH contains a list of MAC addresses of the wireless nodes which are supposed to receive the multicast frame. Nodes listed in the EMH have to reply with an ACK to the transmitter. To avoid collisions of ACK packets at the sender, ACKs are sent in the sequential order established by the EMH. In case the transmitter misses ACKs from any of its next hops, then it tries to retransmit the packet only to the node (or nodes) from which it did not receive any ACK. The retransmission is done through an TS/CTS-like mechanism, which combats the hidden terminal problem. In more details, a Multicast TS (MTS) is sent to nodes by which ACK is missing. These terminals will reply with a CTS and then the transmitter can retransmit the Data frame. In [9], authors set a maximum of 7 retransmissions; after this threshold is exceeded, the frame is dropped and the routing layer is informed of this event with a broken link message. Moreover, to reduce the collision of an ACK, the MMP protocol foresees that the transmitting node sets the Network Allocator Vector (NAV) field in the MAC frame. As a consequence, the radio channel will be assumed busy and the node does not try to transmit. For more details about MMP the reader can refer to [9]. Another interesting MAC protocol proposed in literature for multicast delivery is named A (ound-obin Acknowledge and etransmit, [17]). It implements at the MAC layer a lost frame retransmission procedure employing a simple acknowledge mechanism with the aim to avoid the problem of ACK implosion. The sender, by implementing a round-robin scheme, requests one of its neighbours to forward an acknowledgment, named broadcast acknowledgment (BrACK). It contains a sequence number, s for instance, and a bitmap specifying the reception (or not) of the previous frames with respect to s. Checking the BrACK bitmap, the sender identifies the lost frames and retransmits them. The A performances are interesting; nevertheless, with respect to IEEE and MMP, there is a strong dependence of the performance on the multicast group size. In particular, the packet delivery ratio decreases by increasing the multicast group size (please refer to [17]). For the reasons mentioned above, in this paper it has been decided to utilize, as reference protocols, ODMP and MMP for routing and MAC layers respectively, in order to obtain high performance regardless the multicast group size and the user mobility profile. C. Combining ODMP and MMP Protocols As previously discussed, MMP creates the EMH header containing the list of next hops. This list must be provided from the routing protocol (ODMP in this paper). As a consequence, the introduction of this functionality into ODMP has to be foreseen (we named it X-casting ODMP, X-ODMP). In more details, each terminal must store in the Forwarding Group Table the address of the nodes forwarding the EPLY towards the source. In such a way, each forwarder node will determine the list of next hops that will be delivered to MMP for multicast MAC transmission. In a multi-hop scenario, each node, over the path to the multicast receivers, locally updates the EMH field by including the MAC addresses of its next hops. The introduced feature does not modify the ODMP multicast route creation/refresh algorithm. It shall be underlined that the joint use of ODMP/MMP changes the mechanism of Data forwarding. In fact, differently from IEEE , the multicast MAC transmission that characterizes MMP does not assure to ODMP a mesh forwarding scheme for Data packets delivery. Indeed, when MMP is jointly used with ODMP, the packets delivery will follow a tree configuration. The explicit list of receivers,

4 indicated in the EMH, does not allow the generic node to receive several copies of the same Data packet from different Forwarder nodes. Although it may seem an advantage for multicast transmission, simulations presented in Section IV show that cons overcome pros. This demonstrates that it is not possible to simply couple ODMP and MMP and hope to achieve a good performance level in terms of reliability without the introduction of ad-hoc additional features. III. OU POPOSAL In the next sub-sections we describe the enhancements introduced both in routing and MAC layer with the purpose of solving issues presented in Section II. A. outing Layer: EX-ODMP As discussed in sub-section II-C, the joint use of ODMP and MMP changes the forwarding scheme of Data packets. In order to re-establish a mesh configuration, we define a new version of ODMP named Enhanced X-ODMP (EX- ODMP). It takes advantage of the possibility to specify a list of reverse hops in the EPLY. In more details, a generic receiver must wait 25 ms before to send a EPLY (according to what reported in ODMP standard documents [16]). During this period, we propose to collect several QUEIES. The addresses of nodes transmitting the QUEIES will be indicated as reverse hops in the EPLY. In so doing, the receiver will be a next hop for different Forwarding nodes and a mesh configuration is assured. According to our proposal, when a Data packet must be forwarded the routing agent creates the following packet: <Xcast Header, Data ID, ODMP Header + Payload>. It is worth noting that two new fields are added: Xcast header, contains IP addresses of the next hops (see section II-C) that will be translated into MAC addresses to create the EMH header for Data frame forwarding; Data ID, is a field introduced to uniquely identify Data packets. The source sets this field and Forwarders do not change it. The role of Data ID will be explained in Sub-section III-B. mesh configuration will be dynamically updated and adapted to the real network load, (ii) the number of channel accesses will be reduced and (iii) the channel will be busy for a shorter time. D3MP modifies the MMP frame structure by introducing a new field, named Frame ID, with the purpose of uniquely identifying a Data frame. As introduced in Section III-A, the multicast traffic source identifies uniquely a Data packet through the Data ID field provided by the routing agent. The MAC agent copies the content of Data ID into the Frame ID. In so doing, each Data Frame is uniquely identified and D3MP can check whether neighboring nodes are transmitting the same data towards the same destinations. In Figure 2 a possible scenario is reported to explain the procedure applied by D3MP in more details. Two different nodes, A and B, that are forwarding a multicast packet are shown (obviously, this procedure is applicable to a multi-hop scenario). In particular, we assume that the Extended Multicast Headers of nodes A and B contain the following addresses, respectively: EMH: {B,C,D,E}, Node A; EMH: {A,C,E}, Node B. We suppose that the channel is busy when node A tries to transmit. In this case, node A has to wait before sending its frame and during this time we suppose that it receives the frame by B. Node A checks the received Frame ID and compares it with the Frame ID of the packet that should be forwarded. If these are the same, then node A deletes B from its EMH, thus avoiding to send the same packet to B. Furthermore, for the same reasons, node A deletes from its EMH all the nodes listed within B s EMH. After this procedure, the A s EMH will contain only the D node address and unnecessary transmissions are not performed. However, simulation results demonstrate that D3MP assures a mesh configuration so that each node can receive the same packet from several routes; this improving multicast reliability. B. MAC Layer: D3MP Dynamic Mesh based Multicast MAC Protocol (D3MP) is a new policy which offers reliable multicast transmissions at the MAC layer. Data forwarding mechanism and retransmission scheme are both based on MMP. Nevertheless, D3MP introduces a new procedure that allows to increase the performance of Data forwarding when, at the routing layer, a mesh-based algorithm is exploited (EX-ODMP in this paper). In more details, this procedure dynamically tries to reduce unnecessary transmissions at the MAC layer by checking whether neighbouring nodes are transmitting the same Data towards the same destinations. As a consequence, (i) the Figure 2. D3MP mechanism of EMH modification

5 IV. SIMULATION MODEL AND ESULTS In this section a simulative comparison study among ODMP/IEEE , X-ODMP/MMP, EX- ODMP/MMP and EX-ODMP/D3MP is illustrated. Simulations are carried out by means of ns2.33 [19]; obtained results are reported within 95% confidence interval. The following metrics are used to compare the performance of different envisaged solutions: Packet Delivery atio (PD): defining the ratio of the number of Data packets received by a multicast group member over the number of packets which should have been received. The value of PD is the average of the PDs of each receiver; Signalling Overhead: defined as the ratio of the Control bytes sent by each node of the network over the Data bytes transmitted by the multicast source. If a node forwards a Control packet, then the bytes of the packet are included in the metric; Forwarding Signalling: representing the ratio of the total signalling bytes sent by each node of the network to the Data bytes transmitted by the multicast source. In this metric ACK and MTS/CTS frames are taken into account; Delay: defining the average delay in delivering a Data packet. The main assumptions in the test campaigns are available from Table II. In the analyzed scenario, node position, speed, and movement direction are generated by utilizing a random waypoint based application provided by ns2.33 and called setdest. In more details, nodes are uniformly distributed and their speeds uniformly vary in a given range (in our case [0, 20] m/s). Moreover, we take into account only one source while the multicast group size is varied during the simulation campaigns to testify protocol scalability and in particular to demonstrate that our proposal (with respect to previous studies [5], [6]) allows increasing the performance in terms of PD almost independently of the group multicast size. Table II SIMULATION PAAMETES Parameters Value Number of nodes 50 Number of sources 1 Traffic duration 600 sec X-Dimension 1000 m Y-Dimension 1000 m Propagation Model TwoayGround Transmission ange 250 m Traffic CB Transport Protocol UDP Bandwidth 2 Mbps Packet Size 512 bytes ate of source 2 pkt/sec Table III EX-ODMP/D3MP: PD AND DELAY DEPENDING ON NUMBE OF ETANSMISSIONS # etransmissions PD Delay [sec] Table IV EX-ODMP/MMP: PD AND DELAY VS. NUMBE OF ETANSMISSIONS # etransmissions PD Delay [sec] New features introduced by both EX-ODMP and D3MP impose us to tune opportunely the number of retransmissions at the MAC layer, that in standard MMP is equal to 7 [9]. Indeed, as shown in the following tables, the Data/ACK mechanism increases the network congestion and, as a consequence, causes a PD decrease. Obviously, the same mechanism increases the delay due to a greater channel occupation. We demonstrated (see Table III) that, in a mesh configuration characterized by a cooperative framework EX- ODMP/D3MP, the retransmissions number must be decreased with the purpose of increasing the PD and of decreasing the delay. In particular, the most relevant results in term of packet delivery ratio are obtained when the number of retransmissions is equal to 4. This value also guarantees a decrease in terms of delay. Similar studies have been performed by considering EX- ODMP/MMP (see Table IV). Also in such a situation 4 retransmissions guarantee the highest PD allowing also a reduction in terms of delay with respect to 7 retransmissions. Finally, the refresh period for routing protocols is assumed equal to 3 sec, as reported in [16]. Figure 3 shows the PD; as expected ODMP/ has a higher PD than X-ODMP/MMP. This result is a consequence of the tree configuration created by the joint use of X-ODMP/MMP. Indeed, in ODMP/802.11, if a node does not join the multicast group (case of failed EPLY transmission), then it can still receive Data packets by neighbouring nodes thanks to the broadcast transmission at the MAC layer. While, in X-ODMP/MMP, if a EPLY signalling is lost, then: (i) the Forwarder cannot memorize the receiver address in its Forwarding Table; (ii) the EMH will not contain the receiver address; (iii) the receiver will not be able to receive from its neighbours

6 Packet Delivery atio 0,9 0,8 0,7 Forwarding Signalling ,6 0 Figure 3. PD as a function of the number of receivers. Figure 5. Forwarding Signalling as a function of the number of receivers. 2, ODMP/ X-ODMP/MMP EX-ODMP/MMP EX-ODMP/D3MP Signalling Overhead 2 1,5 1 Delay [sec] ,5 Figure 4. Signalling Overhead as a function of the number of receivers. Figure 6. Delay as a function of the number of receivers. any Data packet (MAC multicast transmission). It clearly emerges that, to improve the PD of X-ODMP/MMP, it is required to increase the number of receivers joined to the multicast group. This result is obtained when we introduce the Enhanced X-ODMP (see Figure 3). Indeed, the EX-ODMP/MMP solution improves the performance with respect to X-ODMP/MMP. Nevertheless, when compared with standard ODMP/802.11, this latter is more efficient. The most performing results in terms of PD are obtained when the proposed protocol, D3MP, is coupled with EX- ODMP. The dynamic mechanism implemented at the MAC layer allows a better exploitation of the radio channel with a consequent improvement of PD, also in the presence of multicast groups of small dimensions. The enhancement introduced into ODMP does not significantly change the control signalling procedure of routing protocol. As shown in Figure 4, the Signalling Overhead is almost the same for all the considered solutions. On the contrary, as both MMP and D3MP are based on a Data/ACK mechanism for packet forwarding and on MTS/CTS scheme for retransmission procedure, the Forwarding Signalling (Figure 5) increases with respect to ODMP/ Nevertheless, this signalling increase is the price to pay to guarantee a more reliable data delivery. However, it is worth noting that D3MP is more performing than MMP in terms of Forwarding Signalling. This is a consequence of the dynamic reduction of unnecessary transmissions. Finally, the end-to-end Delay trend is reported in Figure 6 when varying the number of receivers. Also in this case the dynamic mechanism introduced in D3MP allows a manifest reduction in terms of Delay with respect to MMP

7 In both X-ODMP/MMP and EX-ODMP/MMP solutions the number of retransmissions, DATA/ACK and MTS/CTS mechanisms introduce a high latency in packet delivery; while the Delay introduced by D3MP is comparable with IEEE V. CONCLUSION In this paper we presented a study on the performance of ODMP implemented together with an extension of IEEE , called D3MP. This protocol introduces multicast reliability into a MANET, through a policy of retransmission of lost frames implemented at the MAC layer. Through a simulation study we compared D3MP with standard IEEE and with MMP, which is one of the more efficient reliable multicast MAC protocol. We demonstrated that when ODMP and MMP protocol are jointly used the PD decreases with respect to ODMP/ This is mainly due to the fact that, differently from IEEE , multicast MAC transmission that characterizes MMP does not assure to ODMP a mesh forwarding scheme for Data packets delivery. Indeed, when MMP is jointly used with ODMP the packets delivery will follow a tree configuration. ODMP features have been augmented by introducing EX-ODMP, with the capability of mesh creation when coupled with a multicast MAC protocol. We analyzed the performance of EX-ODMP/MMP and also proposed a policy for dynamic EMH creation into MMP (D3MP) to improve the use of radio channel. We demonstrated that the cooperative use of EX-ODMP and D3MP increases the reliability with respect to other analyzed solutions. ACKNOWLEDGEMENT Special thanks go to Mr. Giuseppe Zerbo for its help in simulator implementation and test activity. EFEENCES [1] D. P. Agrawal, C. M. Cordeiro, H. Gossain, Multicast over wireless mobile ad hoc networks: Present and future directions,ieee Network, Special Issue on Multicasting: An Enabling Technology, Issue 1, Jan./Feb [2] P. Kermani, Special issue on advances in mobile ad hoc networking, IEEE Personal Communications Magazine, vol. 8(1),Feb [3] C. E. Perkins, Ad Hoc Networking, Addison Wesley, [4] S. Deering, Multicast outing in a Datagram Network, Ph. D. dissertation, Stanford Univ., [7] S. J. Lee, W. Su, J. Hsu, M. Gerla,. Bagrodia, A Performance Comparison Study of Ad Hoc Wireless Multicast Protocols, INFOCOM 2000, 19th Annual Joint Conference of the IEEE Computer and Communications Societies, Proceedings, IEEE, Vol. 2 (2000), pp vol.2. [8] IEEE Std , IEEE standard for wireless lan medium access control (MAC) and physical layer (PHY) specification, June [9] K. Anand, D. P. Agrawal, H. Gossain, N. Nandiraju, Supporting MAC layer multicast in IEEE based MANETs: Issues and Solutions, Proceed. of the IEEE International Conf. on Local Computer Networks, [10] J. J. Garcia-Luna-Aceves, and E. L. Madruga, The Core- Assisted Mesh Protocol, IEEE Journal on selected areas in communications, Vol. 17(8), August [11] H. Moustafa and H. Labiod, A Performance Comparison of Multicast outing Protocols In Ad hoc Networks, PIMC 2003, 14th IEEE Proceedings on Personal, Indoor and Mobile adio Communications, 2003, pp , Vol.1. [12] C. W. Wu, Y. C. Tay, AMIS: A Multicast Protocol for Ad hoc Wireless Networks, Military Communications Conference Proceedings, MILCOM, Vol. 1, pp , [13] J. Xie,.. Talpade, A. MCauley, M. Liu, AMoute: Ad Hoc Multicast outing Protocol, Mobile Networks and Applications, Vol. 7, pp , [14] C. Perkins, E. oyer, Multicast operation of the ad hoc ondemand distance vector routing protocol, Proc. ACM Mobicom 99 Conf., (1), pp , Aug [15] C. C. Chiang, M. Gerla and L. Zhang, Forwarding Group Multicast Protocol (MP) for multihop, mobile wireless networks, Cluster Computing, Vol. 1(2), pp , [16] S. J. Lee, W. Su and M. Gerla, On-Demand Multicast outing Protocol (ODMP) for ad hoc networks, Internet Draft, Work in progress draft-ietf-manet-odmrp-02.txt, January [17] J. Xie, A. Das, S. Nandi and A. K. Gupta, Improving the reliability of IEEE broadcast scheme for multicasting in mobile ad hoc networks, IEEE Proc. Commun., Vol. 153(2), April [18] W.SiandC.Li,MAC: A eliable Multicast MAC Protocol for Wireless Ad Hoc Networks, In Proc. of the 2004 International Conference on Parallel Processing (ICPP 2004), Aug [19] NS-2 Network Simulator, [5] M. Gerla, S.J. Lee, W. Su, On-demand multicast routing protocol in multihop wireless mobile networks, Mobile Networks and Applications, vol. 7: , Feb [6] G. Tsudik, K. Viswanath, K. Obraczka, Exploring mesh and tree-based multicast routing protocols for MANETs, IEEE Transactions on Mobile Computing, vol. 5(1), Jan