Supporting multicasting in mobile ad-hoc wireless networks: issues, challenges, and current protocols

Transcription

1 WIRELESS COMMUNICATIONS AND MOBILE COMPUTING Wirel. Commun. Mob. Comput. 2002; 2: (DOI: /wcm.26) REVIEW Supporting multicasting in mobile ad-hoc wireless networks: issues, challenges, and current protocols Symeon Papavassiliou*, Department of Electrical and Computer Engineering New Jersey Institute of Technology New Jersey Center for Multimedia Research and New Jersey Center for Wireless Telecommunications Newark NJ , U.S.A. Beongku An Department of Electrical and Computer Engineering New Jersey Institute of Technology New Jersey Center for Multimedia Research and New Jersey Center for Wireless Telecommunications Newark NJ , U.S.A. Summary The basic philosophy of personal communication services is to provide user-to-user, location independent communication services. The emerging group communication wireless applications, such as multipoint data dissemination and multiparty conferencing tools have made the design and development of efficient multicast techniques in mobile ad-hoc networking environments a necessity and not just a desire. Multicast protocols in mobile ad-hoc networks have been an area of active research for the past couple of years. This paper summarizes the activities and recent advances in this work-in-progress area by identifying the main issues and challenges that multicast protocols are facing in mobile ad-hoc networking environments, and by surveying several existing multicasting protocols. This article presents a classification of the current multicast protocols, discusses the functionality of the individual existing protocols, and provides a qualitative comparison of their characteristics according to several distinct features and performance parameters. Furthermore, since many of the additional issues and constraints associated with the mobile ad-hoc networks are due, to a large extent, to the attribute of user mobility, we also present an overview of research and development efforts in the area of group mobility modeling in mobile ad-hoc networks. Copyright 2001 John Wiley & Sons, Ltd. KEY WORDS multicasting protocols mobile ad-hoc networks wireless networks mobility Published online: 28 vember 2001 Ł Correspondence to: S. Papavassiliou, Department of Electrical and Computer Engineering, New Jersey Institute of Technology, New Jersey Center for Multimedia Research and New Jersey Center for Wireless Telecommunications, Newark, NJ , U.S.A. papavassiliou@adm.njit.edu Contract/grant sponsor: The New Jersey Center for Multimedia Research, The New Jersey Center for Wireless Telecommunications and the New Jersey Institute of Technology; contract/grant number: Copyright 2001 John Wiley & Sons, Ltd.

2 116 S. PAPAVASSILIOU AND B. AN 1. Introduction As the technology and the popularity of the Internet have grown dramatically over the last few years, applications that require multicasting are becoming more widespread. Multicasting provides for efficient ways of transmitting data from one source to a group of receivers simultaneously. Since in multicasting the sender sends a single copy to all the receivers instead of broadcasting the data to all the receivers or sending a separate copy to each individual receiver, there are a lot of advantages associated with this, which vary from bandwidth efficiency to lower network and node overhead. For those reasons multicasting has emerged as one of the most focused areas in the field of networking and many multicast protocols have been proposed and developed in wired networks [1]. At the same time mobile wireless networking has enjoyed a dramatic increase in popularity over the last few years. The advances in hardware design, the rapid growth in the communication infrastructure, and the increased user requirement for mobility and geographic dispersion, continue to generate a tremendous need for dynamic mobile networking. The emerging wireless applications such as audio/video conferencing, distance learning, E-commerce, and distributed and multiparty games will benefit by multicast support in the underlying wireless and mobile networks. Many of the emerging and existing wireless and mobile networks deploy widely different technology, protocols, and wireless links, and therefore, the support for multicasting is an interesting challenge. Regardless of the network environment, multicast communication is a very useful and efficient means of supporting group-oriented applications. With limitations of bandwidth and other resources in wireless networks, the efficient support for multicast is even more important as resource conservation in multipoint communication, therefore making the design and development of efficient multicast protocols a necessity and not just a desire. Supporting multipoint communications for emerging applications in wireless and mobile networks is an interesting challenge especially due to the user mobility and limited resources. Multicasting techniques and approaches used in conventional wired networks [2,3,4] are not well suited for the mobile environment due to the considerable overhead produced by periodic route messages and their slow convergence to topological changes. There are currently two kinds of wireless networks. The first is known as the infrastructure-based network. The infrastructure needed to support such communication networks often includes the use of many fixed radio transceiver sites that serve as the gateways for communication with wireless units. The fixed transceivers are networked through fixed links. The second one is known as a mobile ad-hoc network. In such environments there are no dedicated base stations as in conventional cellular networks, and all nodes interact as peers for data forwarding. This distributed nature eliminates single points of failure and makes those mobile ad-hoc networks more robust and survivable than other wireless networks. In this paper we emphasize on the multicasting problem for mobile ad-hoc networks. The multicast issues in the infrastructure-based networks are somewhat less complex due to the availability of fixed infrastructure support for multicast tree/mesh building and maintenance. The main engineering obstacles in those networks have to do with the last hop delivery and membership tracking, and are not discussed in this paper. In the ad-hoc networks, where all nodes may be mobile/portable, the interconnection pattern changes between any two nodes over a period of time, therefore multicast routing involves frequent route discovery and managing tree/mesh for multicast using limited or scoped broadcast. This paper presents a first effort in the literature to identify, report, and summarize the work in progress and the most recent advances in the area of multicasting in ad-hoc networks. The main objective of this work is to reveal the main philosophy of each protocol in the literature, identify commonalities and differences among the various protocols, and categorize them accordingly. Our objectives are to understand and compare the main design characteristics of the different protocols, to identify their corresponding strengths and drawbacks, and finally present a comparative qualitative study. In general, the multicast routing protocols designed for infrastructure-based wireless networks (i.e. cellular networks) cannot be used in ad-hoc networks, mainly for the following three reasons: (a) the routes in ad-hoc networks change frequently, (b) the fixed infrastructure does not exist for route (re) computation, and (c) there are power, coverage and bandwidth restrictions. If such protocols were to be used in adhoc wireless networks the amount of processing and storage overhead may exceed the limited capabilities of mobile nodes. It has been argued [5] that just routing table updates can consume about half of the bandwidth even under medium mobility level under certain conditions.

3 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 117 This paper is organized as follows. In Section 2 we identify and describe the main issues and challenges that multicasting techniques have to face and address in mobile ad-hoc networking environments. Section 3 examines several multicast techniques designed for mobile ad-hoc networks, by first giving a classification of those multicast techniques, then describing the functionality of the individual existing protocols, and finally comparing their various characteristics. Section 4 presents an overview of research and development efforts in the area of mobility modeling, emphasizing on the impact that mobility models may have on the performance evaluation of multicast protocols in mobile ad-hoc networks and the need for the development of realistic mobility models suitable for the motion behavior of nodes supporting multicast applications. Finally Section 5 concludes the paper. 2. Issues and Challenges The development of efficient and applicable protocols to support the various networking operations in mobile ad-hoc networks presents many issues and challenges due to the fact that such networks have rapidly changing, random, multi-hop topologies. In this section the main issues and requirements that existing and future protocols should take into consideration are summarized. ž Dynamic multi-hop topology: The continuous and random movement of nodes in mobile ad-hoc networks results in rapidly changing multi-hop topologies. Since not every pair of nodes is within transmission range of each other, a packet must often be relayed over several hops before reaching its final destination. In addition, the ad-hoc wireless nodes may experience a change of topology during a multicast session. In general, strategies designed to support internetworking and multicasting in mobile ad-hoc networks should handle such topological changes with minimal overhead by limiting the scope of control packets that may have to be generated and propagated after a change in topology. ž Routing information (in)accuracy: In mobile adhoc networks, the available state information for routing inherently includes inaccuracies due mainly to user mobility, variable delays or even loss of control packets. Multicast protocols should attempt to reduce the impact of those inaccuracies on the efficiency and reliability of their operation. ž Scalability : Scalability in ad-hoc mobile wireless networks presents more complex problems than in wired or last-hop networks, due to the random movement of nodes and the bandwidth and power limitations. The protocol scalability is expressed in the efficient support of large numbers of users, links, nodes, simultaneous sessions, etc. ž Interoperability and adaptability: Current multicast solutions in ad-hoc networks differ substantially from those offered in fixed networks or last-hop networks. However, hosts should be able to migrate freely among ad-hoc, fixed infrastructure wireless, and traditional wired networks. In order to offer seamless and integrated multicast services, additional novel mechanisms should be developed and supported for interoperation of fixed and wireless multicast solutions. Additionally hosts should be able to switch on the fly among multiple mechanisms with the minimum of both effort and inconvenience (i.e. packet loss). ž Network resource usage efficiency: With limitations of bandwidth and other resources in wireless networks, the efficient construction and support of multicast trees and other mesh topologies is very critical to successful communication in wireless ad-hoc networks. Keeping accurate state multicast information at each router becomes impractical if the set of neighbors changes at very high rate. On the other hand, excessive propagation of control information (i.e. flooding) in highly mobile networks creates substantial overhead and consumes a considerable amount of resources. There is an inherent trade-off between efficient usage of network resources and robustness or reliability. ž Power consumption: Another important factor is the limited power supply in hand-held devices, which can seriously inhibit packet forwarding in an ad-hoc mobile environment. Hence techniques that take into account each node s power metrics and try to minimize power consumption are more desirable and provide a way to create more long-lived routes. There are two non-orthogonal approaches to optimizing the power consumption in wireless ad-hoc networks. The first is to minimize the overall power dissipated throughout the network on the average under the techniques adopted. The second is to minimize the average power dissipated by the individual nodes. ž Reliability and Security: Reliable services are very important in some applications such as military battlefields and emergency operations that are

4 118 S. PAPAVASSILIOU AND B. AN supported by mobile ad-hoc wireless networks. Offering reliable delivery of data to a group of fast moving nodes that change their position continuously, adds substantial complexity to the already complicated problem of efficient multicasting in ad-hoc networks. Moreover since multicast traffic may pass through unprotected routers and/or links and since the nature of such applications is many times of high security and importance (as is the case with military applications), additional efforts should be devoted to provide secure communication. ž Group membership (join/leave operation): Due to the user mobility the process of group management becomes of major concern in mobile ad-hoc networks. Since nodes are constantly moving, operations for leaving or joining a multicast group, or leaving at some point and joining at another, could place a significant amount of overhead in the multicast protocol. Thus group membership protocols should efficiently process such membership changes in order to minimize their impact on the overall protocol s performance. ž Quality of Service (QoS): Many, if not all, of the above considerations contribute to the user and/or network quality of service models. Many applications require support of a certain QoS for optimum performance. Offering such kinds of applications particularly in a wireless environment places additional limitations which need to be accounted for in developing protocols in ad-hoc networks. In general it is inadequate to consider QoS merely at the network level without considering the underlying media access control layer. Given the problems associated with the dynamics of the nodes (i.e. fluctuating link characteristics, node movement), supporting end-to-end QoS is a nontrivial issue that requires in-depth investigation. Currently, there is a trend toward an adaptive QoS approach instead of the plain resource reservation method with hard QoS guarantees. ž Mobility: In ad-hoc networks, all network components (nodes) may function as routers and may move at high speeds and in random directions. This node mobility could lead to inefficient multicast trees or mesh configurations, loss of packets or incorrect routing, therefore placing additional constraints on the design of multicast routing protocols that must overcome those mobility problems in an efficient way. Mobility patterns add another dimension to the problem of routing in mobile ad-hoc networks. Mobility contributes to a large extent to almost all the previous issues and constraints described in this section. In evaluating the performance of different techniques for ad-hoc networking, realistic mobility modeling is crucial. Mobility models for ad-hoc networks range from random individual models to more structured group-oriented models. Mobility of users will not only impact the networking techniques such as routing, multicasting and reconfiguration, but also the underlying physical layer constraints. 3. Existing Multicast Protocols The goal of this section is to provide an overview of the recent progress in the field of multicast protocol design and development in mobile ad-hoc networking environments. We first classify the existing protocols according to a taxonomy consisting of several distinct characteristics. Then we discuss the main characteristics of each protocol and highlight their respective strengths and drawbacks. Finally we provide a qualitative comparison of the various protocols based on a set of performance in mobile ad-hoc networks. It should be noted here that this paper does not intend to provide a full survey and quantitative performance comparison of all the various existing protocols. It rather presents a first effort in the literature to identify, report, and summarize the work in progress in the area of multicasting in ad-hoc networks. However, a qualitative comparison of the various protocols is provided. In Reference [6] a quantitative performance comparison study of a small subset of the current multicast protocols for ad-hoc networks is presented. In that study the authors simulated five protocols and presented results regarding some specific performance parameters of the protocols, such as: packet delivery ratio and control information as a function of traffic load, mobility speed and multicast group size. Our paper evaluates all the current protocols from a different angle, trying to identify the various design characteristics and strategies of the different protocols proposed in the literature. As we will see in detail in the following sections the scope of our effort is more general and broader, in the sense that we try to reveal the main philosophy of each protocol, identify commonalities and differences among the various protocols, and categorize them accordingly. Among our primary objectives is the understanding and comparison of the main design characteristics of the different protocols, as well as the evaluation of those protocols based on a set of

5 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 119 qualitative criteria and requirements, such as: routing philosophy, dependency on other protocols, reliability, quality of service support provision, etc Taxonomy The development of multicast routing protocols for ad-hoc mobile wireless networks has been recently a very active research topic. Several protocols have been developed for group-oriented communication in these networks. In general the multicast routing protocols can be classified as follows: ž Proactive multicast routing protocol tree based non-tree based ž Reactive multicast routing protocols tree based non-tree based Figure 1 shows a schematic representation of these categories along with the individual protocols that belong to each subgroup. In the following subsections we provide a discussion regarding the main characteristics, advantages and disadvantages of each category identified, which actually presents the basis used to classify the various multicasting protocols Proactive multicast vs. reactive multicast Proactive Multicasting Protocols maintain up-to-date multicasting information from each node to the other Tree ASTM AMRoute Ad-hoc Multicast Routing protocols Proactive n-tree CAMP MCEDAR Tree AODVM DMRP DSMR RBM AMRIS LAM Reactive n-tree ODMRP FGMP RMRP Geocast Fig. 1. Classification of ad-hoc multicast routing protocols. nodes that are members of the multicasting groups in the network. In these protocols each node maintains one or more routing information table. They maintain and update the necessary multicasting information constantly and with no regard to when and how frequently such multicasting routes are desired. Such proactive mechanisms involve the constant and often generation and propagation of the appropriate multicasting-related data to keep the associated tables and multicast topologies updated, in order to keep track of the wireless link status changes in the networks. Those protocols may consume substantial power and high bandwidth, especially in a highly mobile environment where the topological changes are quite fast and often, and have high storage capacity requirements. On the other hand their implementation is simple, and the source has always a route available to a multicast group to be used immediately when a node has data to send to the multicast group. Reactive Multicasting Protocols create multicast routes only on demand, that is only when such routes are desired by the source node. Group membership and multicast routes are established, maintained and updated on demand. Therefore the associated control overhead in those protocols heavily depends on the actual traffic characteristics of the network. In general and under normal network operation those protocols reduce the channel overhead while at the same time they may introduce route acquisition delay since the source has to wait until the appropriate multicast route is discovered Tree based multicast vs. non-tree based multicast Tree based Multicasting Protocols use the tree mechanism as the vehicle to forward multicast data to the appropriate destinations. Multicast trees provide an efficient and simple way of creating a single path between source and destination pairs. However, maintaining a routing tree for the purposes of multicasting packets when the underlying topology changes frequently can incur substantial control traffic. Additionally, during periods of routing-table instability mobile nodes may be forced to stop forwarding packets while they wait for the multicast routing tree to be reconstructed. There are two kinds of approaches that implement the tree based multicasting and have been adopted by the various multicast routing protocols. The first one per-source trees distributes the traffic evenly in the network (assuming that sources and

6 120 S. PAPAVASSILIOU AND B. AN receivers are evenly distributed) and do not require any aggregation of the traffic to central points. However, in highly mobile networks such an approach may present many problems that mainly stem from the fact that the packets may never reach the destination since the network may find itself with obsolete routing tables which may point towards the wrong direction. The remedy to such a problem is to increase the routing update rate with mobility which in turn may result to increased control traffic that may present difficulties in the scaling of the protocol. The second tree-based approach is the shared-tree multicast technique which partially overcomes the problem mentioned earlier. However, the shared tree approach has the drawback that most of the traffic is generated on the backbone tree, rather than evenly distributed across the network and the paths are often not optimal. This may actually lead to low throughput efficiency. Additionally, as the entire network moves fast and the membership changes dynamically, the center points of the shared tree may be off center further aggravating the non optimality of the paths. n-tree based multicasting protocols use other kinds of approaches (i.e. mesh-based topologies) to deliver data to the members of the multicast groups. A multicast mesh is a subset of the network topology that provides at least one path (usually more than one) from each source to each receiver in the multicast group. The mesh provides richer connectivity among multicast members compared to tree-based protocols. Having redundant paths among different nodes of the mesh group helps overcome the problems of node displacement and channel fading. Therefore, unlike trees, frequent reconfigurations may not be required and communication disruption may rarely occur. On the other hand, to achieve that target, additional redundant traffic may be generated in the network (i.e. some kind of flooding redundancy). The advantages achieved by those approaches versus the drawback of the increase in the traffic that may be introduced depend on the degree of connectivity and redundancy among the nodes of the forwarding or mesh groups Multicast protocols: Overview and characteristics In this section we present an overview of the existing multicasting protocols for mobile ad-hoc networks, by presenting the key features and main characteristics of each protocol and highlighting their respective strengths and drawbacks. Most of those characteristics are complementary to the advantages and disadvantages that the various protocols naturally inherent depending on which category they belong to (i.e. proactive, reactive, tree based, non-tree based) Proactive multicast routing protocols Tree based protocols Adaptive shared tree multicast routing protocol: ASTM [7] is a hybrid adaptive per-source and shared tree algorithm that tries to optimize a delay link cost tradeoff. The main feature introduced by this protocol is the two-level mobility model that allows for the definition of a robust routing scheme and therefore the creation of a stable shared tree. The shared tree schemes exhibit lower throughput than the per-source shortest path solution at heavy load as expected due to concentration on the common tree. However, they present much less control overhead than the persource tree solution, since the latter must constantly refresh separate trees rooted at different sources. The shared tree solution offers better potential to scale to large networks. The shared tree approach is based on the notion of Rendezvous Point (RP) [8]. Sender members send multicast packets towards the RP, and receiver members send join requests to RP. Multicast packets are forwarded to receiver members along the multicast forwarding tree. The routing protocol is an extension to the wireless environment of the Destination-Sequenced Distance-Vector (DSDV) scheme originally developed for wireline networks [9]. Adaptive shared tree multicast allows to switch between RP-rooted shared tree and per-source tree dynamically on receiver/sender pair basis, using relative path length/links load tradeoffs as criteria. The most appropriate environment for those kind of algorithms is the one that presents at least two-level mobility nodes. In that case the protocol can take advantage of its hierarchical nature. des with low mobility can provide a stable backbone tree, onto which fast moving nodes can inject their traffic. However, because ASTM uses two-level mobility models based on clusters and clusterhead routing, frequent clusterhead changes can adversely affect the protocol s performance. Ad-hoc multicast routing protocol: AMRoute [10] allows for IP multicast in mobile ad-hoc networks by exploiting user-multicast trees and dynamic cores. It creates bidirectional shared trees for data distribution using only group senders and receivers as tree nodes. Unicast tunnels are used as the tree links to connect neighbors on the user-multicast tree. Thus AMRoute does not need to be supported by network

7 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 121 nodes that are not interested/capable of multicast, and thus the cost is incurred mainly by group senders and receivers. All member nodes in AMRoute need to support IP-in-IP encapsulation. AMRoute creates a per group multicast distribution tree using unicast tunnels connecting group members. The protocol has two key parts: mesh creation and tree creation. AMRoute continuously creates a mesh of bidirectional tunnels between a pair of group members. Using a subset of the available virtual mesh links, the protocol periodically creates a multicast distribution tree. AMRoute makes certain nodes as core nodes, in order to initiate the signaling component of AMRoute, such as detection of group members and tree setup. However, the core nodes differ significantly from those used in CBT-based protocols [11], since they are not a central point for data distribution. Implementing multicast protocol using only members has many advantages. For instance, provided that there remains a path among all nodes connected by mesh branches, the user-multicast tree need not change because of network changes. This independence improves robustness and reduces signaling overhead. Moreover, since non-members are not required to perform packet replication the processing and storage overhead needed to support the multicast tree is placed only to members. The penalty paid for using a user-multicast tree is reduced efficiency and increased delay, since non-member routers are not allowed to perform packet replication. As the network becomes more dynamic, maintaining a more optimal path becomes harder and requires increasingly higher signaling overhead. When mobility is increased the temporary loops and non-optimal trees are the drawbacks of AMRoute [6]. Finally, AMRoute assumes the existence of an underlying unicast protocol that can be utilized for unicast IP communication. Therefore it can be operated seamlessly over separate domains as long as they support IP-based unicast protocols. n-tree based protocols Core-assisted mesh protocol: The Core-Assisted Mesh Protocol (CAMP) [12] extends the notion of core-based tree (CBT) for Internet multicasting into multicast meshes that have much richer connectivity than trees. CAMP uses routing structures different from trees, without the need to flood an entire network with either data packets or control packets. CAMP differs from most multicast routing protocols in that it builds and maintains a multicast mesh for information distribution within each multicast group. A multicast mesh is a subset of the network topology that provides at least one path from each source to each receiver in the multicast group. Because a member router of a multicast mesh has redundant paths to any other router in the same mesh, topology changes are less likely to disrupt the flow of multicast data and to require the reconstruction of the routing structures that support packet forwarding. When a link fails breaking the reverse shortest path to a source, the router affected by the break may not have to do anything, because the new reverse shortest path may very well be part of the mesh already. Therefore, link failures are not very critical in CAMP. CAMP extends the basic receiver-initiated approach introduced in the CBT protocol [11] for the creation of multicast trees to enable the creation of multicast meshes. Cores are used only to limit the control traffic needed for receivers to join multicast groups. The failure of cores does not stop packet forwarding or the processes of maintaining the multicast meshes. The use of cores in CAMP eliminates the need for flooding. The basic packet forwarding scheme in CAMP consists of trying to forward multicast data packets along the paths within the mesh that first reach the member routers from the sources. CAMP tends to forward packets along the fastest routes from sources to receivers. Frequent mesh changes can adversely affect routing protocol performance in environments with very high mobility. Finally, for CAMP to work correctly it is necessary for the underlying routing protocol to work correctly in the presence of router failures and network partitions. This limits the number of underlying unicast routing protocols that can be used in conjunction with CAMP. Multicast core extraction distributed ad-hoc routing: MCEDAR [13] uses a mesh as the underlying multicasting infrastructure. MCEDAR is layered as a multicast extension on top of Core-Extraction Distributed Ad-hoc Routing(CEDAR) [14]. A set of hosts are distributedly and dynamically selected as the core of the network by approximating a minimum dominating set of the ad-hoc network using only local computational and local state. Each core host maintains the local topology of the hosts in its domain, and also performs route computation on behalf of these hosts. For each multicast group MCEDAR extracts a subgraph of the core-graph to function as the routing infrastructure. The subgraph is a mesh structure and is called the mgraph for the multicast group. Although the mgraph for a multicast group is a mesh infrastructure, the forwarding of data on the infrastructure

8 122 S. PAPAVASSILIOU AND B. AN is done only on a source based tree, thus saving on redundant transmissions leading to efficient usage of the scarce bandwidth. Also, such a forwarding protocol implicitly creates a source based tree that represents the fastest delivery structure for each data packet of the multicast group. The inherent redundancy present in meshes increases the robustness of the mgraph. Thus, for instance, every link breakage in the underlying graph does not necessitate a reconfiguration of the infrastructure. The forwarding mechanism used by MCEDAR on the mesh creates an implicit source based forwarding tree which ensures that although MCEDAR maintains a mesh routing infrastructure the data forwarding occurs only on a source rooted minimum height tree. A core broadcast mechanism is also implemented using reliable unicast messages that scales linearly with the number of network nodes Reactive multicast routing protocols Tree based protocols Ad-hoc on-demand distance vector multicast routing protocol: AODVM [15] extends Adhoc On-Demand Distance Vector Routing (AODV) [16] to offer multicast capabilities which follow naturally from the way AODV establishes unicast routes. Unicast and multicast routes are discovered on demand and use a broadcast route discovery mechanism. Broadcast data delivery is provided by AODV by using the Source IP Address and Identification fields of the IP header as a unique identifier of the packet. Unlike other protocols, the AODV protocol is capable of unicast, broadcast, and multicast communication. Therefore AODV has an advantage over other routing protocols because it provides all three types of communication without being dependent on or requiring the use of any additional routing protocols. The multicast algorithm uses the Route Request (RREQ)/Route Reply(RREP) messages used in the AODV unicast protocol. As nodes join the multicast group, a multicast tree is composed of group members and nodes connecting the group members. Multicast group membership follows the model of the Mbone [3] in that it is dynamic. A node sends an RREQ message when it wishes to join a multicast group, or when it has data to send to a multicast group and it does not have a route to that group. As the RREQ is broadcast across the network, nodes set up pointers to establish the reverse route. If a node receives a join RREQ for a multicast group, it may reply if it is a router for the multicast group s tree and its recorded sequence number for the multicast group is at least as great as that contained in the RREQ. The first member of the multicast group becomes the leader for that group. A protocol that offers both forms of communication can be streamlined so that route information obtained when searching for a multicast route can also increase unicast routing knowledge, and vice versa. Thus, combining both types of communication into a single protocol simplifies coding and reduces control overhead. However, since routes are created on demand by the source, additional route acquisition delay may be observed. Distributed multicast routing protocol: DMRP [17] is a distributed, source-initiated protocol that combines multicast routing with resource reservation and maintains connections to desired destinations. Its operation is divided into three phases: route construction, route maintenance and route destruction. An initial set of routes are created during the route construction phase. When mobility causes link failures, the maintenance phase takes over and re-establishes routes to each of the destinations. When a session terminates, the source sends a Transceiver/Frequency Release tification(tfrl) along established routes toward the destinations. When a source desires a route to a set of destinations, it checks its connectivity map to see if any destination is reachable in one or two hops. If this is true and a transmission frequency is available, the source checks to see if it has an available transceiver. If a frequency and a transceiver is available, the source determines the optimum frequency at which to transmit. The second version of this protocol enhances the first version and strives to accomplish multicast routing by giving each node two degrees of freedom: each node can choose the frequency at which to transmit and also the transmission power level. Power control is applied to tradeoff between routing delays and frequency reuse factor. DMRP has two main strengths. First, its distributed nature alleviates the need for storing global topological connectivity information, resulting in low complexity. Second, the resource reservation feature allows routes only if sufficient bandwidth exists along every link of the route and gives the protocol the capability of QoS support. Finally for this protocol to work correctly, an underlying protocol is assumed which gives each node knowledge of its neighboring nodes, up to two hops away, and also informs the node of the frequencies in use by those neighbors.

9 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 123 Dynamic source multicast routing protocol: DSMR [4] extends the dynamic source routing protocol (DSR) [18] to support multicasting. Mobile nodes are required to maintain route caches that contain the source routes of which the mobile is aware. Entries in the route cache are continually updated as new routes are learned. Stations that want to communicate with the group have to know the address of all members. The protocol consists of two major phases: route discovery and route maintenance. A mobile node with no topology knowledge that wants to communicate with a group starts the DSR Route Discovery process for each group member. Afterwards it waits until all necessary routes are discovered. It then calculates a minimal spanning (multicast) tree and puts the tree into the header of its packets. Each packet is broadcasted. A mobile node receiving this packet does a simple look-up into the multicast tree. If it is mentioned in the tree it sends an acknowledgment to its father in the tree. Each mobile node waits a certain amount of time for all direct sons in order to acknowledge the packet. If there is a mobile node that did not acknowledge the packet, it assumes a topology change. The subtree starting at this son does not exist anymore. To enable the initial mobile node to start a new Route Discovery process for this network part, it sends back a Route Error packet, enclosing this subtree. As with AODVM, since routes are created on demand by the source, additional delay may be observed. Reservation-based multicast (RBM) routing protocol: RBM [19] is a combined multicast routing, resource reservation and admission control. In RBM, the concept of receiver-initiated operation is borrowed from existing and proposed Internet protocols and the use of Rendezvous Point (RP) to link sources with destinations in a mobile network is borrowed from the protocol independent multicast (PIM) [18] protocol or the CBT protocol [11]. These techniques are integrated and coupled to an underlying unicast routing protocol for mobile networks. RBM emphasizes combined routing and reservation in the face of bandwidth constraint and searches multiple routes while building source-specific trees. RBM routes hierarchically encoded streams with varying fidelities based on prioritized user requests, real-time delivery requirements and network bandwidth constraints. The routing process can be broken into two stages: sourceto-rp and RP-to-destination. Each multicast group is assigned a set of possibly mobile RPs. Each source attempts to send its information to all RPs. Destinations join the group by subscribing to one or more RPs. In source-to-rp routing, each source attempts to send its information to all RPs and attempts to deliver the highest QoS level. The objective of the route selection process is to find a route for each signal from the selected RPs to the destination meeting the destination s prioritized QoS and, optionally, realtime delivery requirements. RBM utilizes a greedy heuristic algorithm for combined admission control and resource reservation to resolve user self-conflicts during route construction. RBM is overlaid on top of multiple, independently executing versions of the underlying, distributed unicast routing protocol (CE) [20] for mobile networks. Because RBM requires heuristic RP selection satisfying some requirements and constraints, its complexity adversely affects routing protocol performance. Since RBM relies on the core nodes, it may not provide for efficient implementation in environments that support high mobility. Ad-hoc multicast routing protocol utilizing increasing ID-numbers: AMRIS [21] is an on-demand multicast protocol that constructs a shared delivery tree to support multiple senders and receivers within a multicast session. AMRIS does not require a separate unicast routing protocol. The basic idea of this protocol is to assign every node in a multicast session an ID number (known as msm-id). A delivery tree rooted at a particular node called Sid (the root of the delivery tree) joins up the nodes participating in the multicast session. The relationship between the msm-id and Sid is that msm-ids increase in numerical value as they radiate away from Sid. The msm-ids allow nodes that have broken off from the delivery tree to rejoin the delivery tree in a localized fashion without causing permanent routing loops. Each node sends a periodic beacon to signal their presence to neighboring nodes. The beacon contains the msm-ids that each node presently has. The Sid has the smallest msm-id within that multicast session. Initially, the special node Sid broadcasts a NEW- SESSION packet. The NEW-SESSION includes the Sid s msm-id (multicast session member ID). Neighbor nodes, upon receiving the packet, calculate their own msm-ids which are larger than the one specified in the packet. The msm-ids thus increase as they radiate from the Sid. The nodes rebroadcast the NEW- SESSION message with the msm-id replaced by their own msm-ids. des participating in AMRIS do not require any global consistent routing state. AMRIS allows for loop-free operation and allows nodes to recover from broken links within one multicast beacon period. Repairs to damaged links are performed Copyright 2001 John Wiley & Sons, Ltd. Wirel. Commun. Mob. Comput. 2002; 2:

10 124 S. PAPAVASSILIOU AND B. AN locally without need for any central controlling node thus increasing survivability. Lightweight adaptive multicast protocol: LAM [22] is a group-shared tree based protocol which is coupled with a specific underlying unicast routing protocol the Temporal Ordered Routing Algorithm (TORA) [23]. LAM can be viewed as an integration of CBT [11] protocol and TORA. The coupling of LAM and TORA on one hand limits the portability of LAM to various underlying technologies, while on the other hand it increases reaction efficiency (lowering the protocol s control overhead) as LAM can benefit from TORA s mechanisms. TORA, which is a distributed and highly adaptive mechanism, provides a unicast routing infrastructure, upon which LAM builds a group-shared multicast routing tree for each multicast group. This tree is centered at a pre-selected node called a CORE. TORA is a source-initiated routing protocol, i.e. routes are built reactively as needed to reflect current traffic patterns. TORA builds a Directed Acyclic Graph (DAG) over the physical network topology. Typically TORA builds multiple routes towards the destinations. Since TORA already builds a DAG destined at the CORE, LAM does not need to worry about loop formation as long as child-to-parent tree links follow the downstream links in the TORA DAG, therefore simplifying the corresponding operation of CBT. As expected, LAM shares many common behavior with other protocols that are based on CBT. One feature that LAM has adopted from TORA and is different from CBT is the method to repair a link failure between the parent and the child. In CBT when an on-tree node discovers the link between itself and its parent is broken it must flush the sub-tree below itself. After all nodes on the sub-tree remove themselves from the multicasting tree they then start rejoining the tree individually. This operation is quite complicated and results in high time and communication complexity. However, in LAM there is no flush operation. The portion of the tree that is affected following any failure is typically highly localized. The packet data forwarding is similar to CBT s: if the source is ontree, it sends data packets to all of its tree neighbors; if the source is not part of the multicasting group, it sends the data packets towards the CORE. The use of shared tree and CORE concepts give LAM the advantages of reduced network control overhead and scalability. However, in LAM the CORE is the critical point for the whole group, and having only one CORE for a group makes LAM vulnerable and not sufficiently robust. In addition, the single CORE configuration may create data traffic concentration at the CORE. To alleviate this problem, a mechanism that maintains multiple cores per group must be developed as an integrated part of LAM. n-tree based protocols On-demand multicast routing protocol: ODMRP [5,24] applies on-demand routing techniques to avoid channel overhead and improve scalability. In ODMRP, group membership and multicast routes are established and updated by the source on demand. While a multicast source has packets to send, it periodically broadcasts to the entire network a member advertising packet. This periodic transmission refreshes the membership information and updates the route. This process of constructing and updating the routes from sources to receivers builds a mesh of nodes: the forwarding group. The forwarding group consists of a subset of nodes that is responsible of forwarding the multicast packets via scoped flooding rather than multicast tree scheme. It supports shortest paths between any member pairs. All nodes inside the forwarding group nodes forward multicast data packets. The created mesh subnetwork provides richer connectivity among multicast members compared to trees. Flooding redundancy among forwarding groups helps overcome node displacements and channel fading. Hence, unlike trees, frequent reconfigurations are not required. As an on-demand routing mechanism, ODMRP reduces unnecessary channel overhead. As long as the number of members is high (dense networks) the forwarding group creates richer connectivity among members and therefore this mesh makes the protocol quite robust to speed. ODMRP cannot only work with any unicast routing protocol, but it can function as both multicast and unicast. Thus, ODMRP can run without any underlying unicast protocol. On the other hand, since routes are setup on demand, delay is increased. Also the richer the connectivity, the higher the effect of flooding on network congestion. Forwarding group multicast protocol: FGMP [25] is a hybrid technique between flooding and shortest tree multicast. The underlying principle behind this protocol is to exploit the inherent broadcast property of wireless media and minimize the need for building some sort of logical infrastructure in mobile ad-hoc networks. This protocol keeps track not of links like the traditional multicast protocols (i.e. PIM

11 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 125 [8], CBT [11], DVMRP [26]) but of groups of nodes which participate in multicast packet forwarding. Multicast forwarding is based on nodes (routes) which are going to accept multicast packets rather than on links on which multicast packets are forwarded. A forwarding group FG is associated with each multicast group G. Any node in FG is in charge of forwarding (broadcast) multicast packets of G. That is, when a forwarding node receives a multicast packet, it will broadcast that packet if it is not a duplicate. All neighbors can hear it, but only neighbor nodes that belong to the FG will broadcast it in turn, if needed. FGMP reduces channel and storage overhead by using scoped flooding over the nodes that belong to the forwarding group. However, frequent changes of forwarding groups in a fast-moving environment can adversely affect the protocol s performance. This approach provides a feasible solution only in small networks. Reliable multicast routing protocol: The basic motivation for the development of RMRP [27] is based on the argument that keeping accurate state about multicast group membership of all nodes neighbors is impractical if the set of neighbors changes at a very high rate. This method proposes the use of plain flooding as a natural solution to the reliable delivery of multicast packet in a highly dynamic environment. Flooding requires each node to keep track of its current neighbors only. Keeping track of neighbors is certainly much simpler than keeping track of group memberships for both neighbors and non-neighbors. Because flooding presents a stateless and topology-independent mechanism for reliable multicasting it could be quite effective in fast highly mobile ad-hoc networks. However, when mobility intervals are small and node speed is sufficiently high, even flooding becomes unreliable. Moreover there is a tradeoff between robustness and efficiency. Excessive control flooding in such highly mobile networks may provide substantial overhead, making the applicability of such techniques inappropriate for large networks. Location-based multicast routing protocol: Geocasting [28,29] is a variant of the conventional multicasting. In conventional multicasting protocols define a multicast group as a collection of nodes that register to a multicast group. For geocasting, the multicast group (or geocast group) consists of all the nodes within a specified geographical region. This scheme uses physical location information for mobile nodes, which can be obtained from a global positioning system (GPS). Using location information of the source and the specified multicast region, this scheme attempts to reduce the overhead of route discovery by reducing the number of nodes outside the multicast region, to whom a multicast packet is propagated. This location-based scheme is essentially identical to multicast flooding, with the modification that a node which is not in the forwarding zone does not forward a multicast packet to its neighbors. The forwarding zone is defined similarly to that of location-aided unicast routing in Reference [28]. Thus, the geocasting mechanism attempts to limit the forwarding space for a multicast packet and results in lower message delivery overhead, as compared to conventional multicast flooding. For this protocol to work correctly all mobile nodes should contain a GPS card and should be able to determine if they are part of the forwarding zone Comparison In the following we present a set of parameters that we are going to use for the protocols qualitative comparison. For each parameter we also provide the set of possible values. We find that most of the papers surveyed do not explicitly address each of these considerations. The following is an attempted comprehensive list of parameters that provides for a fair and meaningful basis for comparison between the protocols presented. This comparison provides a first effort to characterize and identify the qualitative behavior of the various multicasting protocols, and is by no means exhaustive or quantitative. Therefore no parameters such as time and communications complexity of the various protocols, that may require a detailed modeling, simulation and/or analysis of each technique, are evaluated here. 1. Routing philosophy: (flat, hierarchical) 2. Multicast control overhead: (high, low) 3. Route acquisition delay: (high, low) 4. Quality of service support: (yes, no) 5. Reliability: (high, medium, low) 6. Reaction to high mobility: (good, medium, poor) 7. Scalability property: (good, medium, poor) 8. Need for underlying unicast routing protocol: (yes*, yes, no) 9. Flooding: (yes, limited, no) 10. Support of multiple routes: (yes, no) Regarding the Flooding parameter and its associated values, within the content of this paper, the value Limited refers to the flooding of controlonly information within a forwarding group, while

12 126 S. PAPAVASSILIOU AND B. AN the value Yes refers to cases where data flooding is required. Similarly, the values of the parameter Need for underlying unicast routing protocol have the following meanings: Yes* refers to those multicast protocols where their operation depends on specific underlying unicast routing protocol(s); Yes refers to cases where the corresponding multicast routing protocols require the use of any underlying unicast routing protocol and do not depend on specific implementations or routing algorithms; means that no separate underlying unicast routing protocol is required to support the multicasting operation. Regarding the Routing philosophy parameter we identified two categories (values), flat or hierarchical. Within the context of this paper and comparison, the attribute flat is assigned to protocols where all nodes act (physically and logically) as peers in the network and there is not any kind of traffic aggregation to central points, while the value hierarchical is assigned to those protocols that either explicitly or implicitly build some kind of hierarchy. The hierarchy could be based on the use of multiple levels (i.e. multiple mobility levels, creation of backbone, etc.), or on the use of special nodes that may act as information concentrators such as core nodes, RPs, clusterheads, etc. However, in the special case where the concept of core nodes is used for limiting only control information, and those nodes are not necessarily used as data forwarding core nodes, then the corresponding protocol is identified as flat. Finally it should be noted that most of the values used in this qualitative evaluation have relative meaning (i.e. this is a comparative evaluation), and no absolute values are used. Figure 2 presents the comparisons of multicast routing protocols. 4. Impact of Mobility Modeling on Multicast Protocols Performance In order to evaluate in depth the performance of the various multicasting protocols a large number of parameters should be evaluated under different scenarios and conditions. Metrics that can been used include [6], but are not limited to: packet delivery ratio, number of data packets transmitted per data packet delivered, number of control bytes transmitted per data bytes delivered, number of control and data packets transmitted per data packet delivered. All those parameters should be evaluated under different scenarios such as: mobility patterns, number of senders, multicast group size, network traffic load, network size and density, etc. Many of those scenarios or procedures have been used in the past for the evaluation of many different protocols in wired and wireless networks. However, the attribute of mobility as described before, adds an additional dimension to the evaluation of multicasting in networks with mobile nodes. The lack of infrastructure and the capability that nodes have to move freely within the mobile ad-hoc networking environment are the main characteristics that make mobile ad-hoc networks attractive communication solutions. However, as we can see from the discussion in the previous sections many of the issues and challenges associated Multicast Routing Proactive Reactive Protocols Comparison Tree n-tree Tree n-tree Parameters ASTM AMROUTE CAMP MCEDAR AODVM DMRP DSMR RBM AMRIS LAM ODMRP FGMP RMRP Geocast Routing Philosophy (Flat, Hierarchical) Hierarch. Flat Flat Hierarch. Flat Flat Flat Hierarch. Flat Hierarch. Flat Flat Flat Flat Multicast Control Overhead (, Low) Low Low Low Low Low Low Low Low Low Low Route Acquition Delay (, Low) Quality of Service Support (Yes, ) Low Low Low Low Yes Yes Reliability (, Medium, Low) Low Low Medium Medium Low Low Low Medium Low Low Medium Reaction to Mobility (Good, Medium, Poor) Medium Poor Medium Medium Poor Poor Poor Medium Poor Medium Good Medium Good Good Scalability Property (Good, Medium, Poor) Good Poor Good Medium Poor Medium Poor Good Poor Good Medium Poor Poor Poor Need for Underlying Unicast Protocol(Yes*,Yes, ) Yes* Yes Yes Yes* Yes* Yes* Yes* Yes* Yes* Flooding (Yes, Limited, ) Limited Limited LimitedLimited Yes Yes Support of Multiple Routes (Yes, ) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes*: Special Protocol Required Limited Flooding : Only Control Information Flooding Fig. 2. Comparisons of multicast routing protocols in mobile ad-hoc wireless networks.

13 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 127 with the mobile ad-hoc networks stem from that fact: user mobility. For instance, issues associated with the dynamic change of the network topology, the additional overhead involved in the group management process, the difficulties in the construction and maintenance of tree/mesh multicast topologies, the inaccuracy in the routing information, and the increased difficulty in supporting QoS, are due to a large extent to the attribute of user mobility. Among the main objectives of the protocol s performance evaluation process is to define an effective operating region and estimate the overhead of the protocol s components within this region. To achieve this, one of the most important factors is to determine the sensitivity of each of the system s/protocol s components to the transmission range and mobility of nodes Mobility modeling and necessity in mobile ad-hoc networks In wireless networks, mobile nodes can move in many different ways. Mobility modeling attempts to describe the mobility behavior of an individual node or a set of nodes. Mobility models are commonly used to analyze newly designed systems or protocols in both cellular and ad-hoc wireless networks. There is no doubt that the node mobility behavior influences most of the performance-related parameters. Ad-hoc networks are more sensitive to mobility than cellular wireless networks since in the latter the base stations are stationary providing a core fixed backbone network, and the state related to routing and multicasting information changes only when a mobile user leaves the cell, irrespective of relative connectivity with other mobile users. While in cellular networks mobility models are mainly focused on individual movements since communications are primarily point to point rather than among groups, in ad-hoc networks communications are often among teams which tend to coordinate their movements. Hence the need arises for developing efficient and realistic group mobility models in addition to the individual mobility models for ad-hoc networks, especially when evaluating multicast protocols that involve communication among teams. Those models may be used to optimize the design of the multicast strategies and are used as realistic models for the motion patterns in order to evaluate the various protocols performance. Most of the earlier research on mobility patterns was based on cellular networks [30]. Recently, mobility models have been explored also in ad-hoc networks. In general we may classify the existing mobility models according to the strategy of the dependent behavior among mobile nodes (i.e. individual mobility and group mobility), as well as according to the mechanisms used to get the main elements that govern the next moving time interval (i.e. random mobility and pre-information aided mobility). In ad-hoc wireless networks the mobility models focus on the motion behavior between mobility epochs, which are the smallest time periods in which we can assume that a mobile node moves in a constant direction at a constant speed. In the random mobility model the speed and direction of motion in a new time interval have no relation to their past values in the previous epoch. In pre-information aided mobility, the speed and direction of individual motion in a new time interval is related to their past values (i.e. previous values, predefined trajectory). A group mobility model for dependent behavior in ad-hoc networks must capture both motion dependencies over time epochs and the relationship among members of the same group. This team relationship makes it possible to partition the network into several groups, each one with its own mobility behavior. Most of the mobility models that have been developed and used in the literature deal with the motion behavior of the individual nodes ( [31 37]). We strongly believe that in order to evaluate the various multicast protocols under realistic environments and obtain meaningful results we need to use group mobility models in addition to the modeling of each individual node in order to capture the motion dependencies over time and the relationship among the nodes. Therefore the development of realistic group mobility models to be used for the evaluation of the various multicasting and routing protocols in ad-hoc networks is more a need rather than a desire. However, the group mobility modeling development has been an under-appreciated area and very few efforts are reported in the literature regarding the modeling of the group behavior. In the following we present a brief overview of representative existing group mobility models, by noting that the purpose of this discussion is to emphasize the need of group mobility modeling and to register the first few efforts towards this direction Group mobility In the Reference Point Group Mobility (RPGM) model [38],each group (set of nodes) has a logical center. The center s motion defines the entire group s motion behavior (group trajectory), including location, speed, direction, acceleration, etc. Each node is assigned a reference point which follows the group movement.

14 128 S. PAPAVASSILIOU AND B. AN Figure 3 describes the RPGM with a two-group model [38].! V gi presents the motion vectors of each group. The reference point of a node moves from RP to RP C 1 with the group motion vector! GM(! V gi ). The new node position is generated by adding a random motion vector! RM to the new reference point RP C 1. The length of the random motion vector! RM is uniformly distributed within a certain radius centered at the reference point and its direction is uniformly distributed between 0 to 360 degrees. The random vector! GM is independent from the node s previous location. The RPGM defines a sequence of check points (path) to model the motion behaviors of group. When the group center reaches a new check point, it computes the new motion vector! V gi from current check point and next check point locations. This model can be used to model a battlefield situation where different battalions are carrying out same operations in different areas; or to model the rescue operation in a disaster recovery area where multiple rescue teams are randomly spread out over a given area (yet each group may have a unique motion pattern); or to model the interaction between exhibitors and attendees in a convention scenario where attendees may roam from room to room to attend demos and/or project presentations. The Exponential Correlated Random Mobility- (ECRM) [39] model reproduces all possible movements, including individual and group, by adjusting the parameters of a motion function. The new position of a network element (group of nodes or individual nodes)! b t C 1 is a function of the previous position! b t, to which a random deviation!. GM RP(τ) RM RP(τ + 1) Vg 1 Fig. 3. The RPGM model. is added. The Vg 2 model is described by the following equation: b t C 1 D b t e 1/ C s 1 e 2/ 1 Where b t is the position (, ) of a group or a node at time t, is a time constant that regulates the rate of change, is the variance that regulates the variance of change, s is the speed of the node, and is a Gaussian random variable. Using this model, the movement of each group as a whole is controlled independently of the movements of other groups and nodes within the group. At each time step, a group moves a random distance in a randomly selected direction. The and variables, specified separately for the distance and direction, control the nature of the movement. In general, smaller values of result in more random movement, and larger values of result in more variation from a given direction. Within a group, all nodes have the same set of and variables; nodes in different groups may have different sets of variables. This mobility model permits representation of typical node movements in a tactical network, such as the maneuvers of a battalion consisting of several squads of soldiers. The squads may be represented as groups that move as a whole; the soldiers within a squad, while exhibiting some randomness of movement, are aligned with the overall movement of their squad. ECRM requires a complete set of (, ) per group to define the motion of entire network. The drawback is that it is not easy to force a given motion pattern by selecting the parameters. 5. Conclusion In this paper, we present the activities in the workin-progress area of the development of multicasting protocols in mobile ad-hoc networks, by identifying the main issues and challenges that multicast protocols are facing in these environments, and by surveying several existing multicasting protocols. This article presents a first effort in the literature to classify all the current multicast protocols, discusses the functionality of the individual existing protocols, and provides a qualitative comparison of their characteristics according to several distinct features and performance parameters, such as: routing philosophy, dependency on other protocols, reliability, scalability property, mobility support, quality of service support provision, etc. The main objectives of this paper are to identify the various design characteristics and strategies of the different protocols proposed in the literature,

15 MULTICASTING IN MOBILE AD-HOC WIRELESS NETWORKS 129 reveal the main philosophy of each protocol, identify commonalities and differences among the various protocols, and categorize them accordingly. Furthermore, since many of the additional issues and constraints associated with the mobile ad-hoc networks are due to a large extent, to the attribute of user mobility, we present an overview of research and development efforts in the area of mobility modeling in mobile ad-hoc networks, to be used as means to provide a realistic environment to carry out the performance evaluation process of multicasting protocols. Since in ad-hoc networks communications are often among teams which tend to coordinate their movements, and multicasting protocols deal by default with the dissemination of information to groups of nodes, our discussion mainly emphasizes the need to design, develop and use advanced realistic mobility models that provide for the modeling and emulation of the motion of individual nodes as well as of motion patterns of the corresponding groups and their interrelations. References 1. Obraczka K. Multicast transport protocols: A survey and taxonomy. IEEE Communications Magazine 1998; 36(1): Deering S. Host extension for IP multicasting. Internet RFC 1112 August Eriksson H. MBONE: The multicast backbone. Communications of the ACM 1994; 37(8): Merrers J, Filios G. Multicast communication in ad-hoc networks. In Proc. of Vehicular Technology Conference(VTC 98), Chiang C-C, Gerla M. On-demand multicast in mobile wireless networks. In Proc. of ICNP 98, Austin, Texas, October 14 16, Lee SJ, Su W, Hsu J, Gerla M, Bagrodia R. A performance comparison study of ad hoc wireless multicast protocols. In Proc. of IEEE INFOCOM 2000, Chiang C-C, Gerla M, Zhang L. Adaptive shared tree multicast in mobile wireless networks. In Proc. of IEEE Globecom 98, vember Deering S, Estrin DL, Van Jacobson DF, Liu C-G, Wei L. The PIM architecture for wide-area multicast routing. IEEE/ACM Transactions on Networking 1996; 4(2). 9. Perkins CE, Bhagwat P. ly dynamic destination-sequenced distance-vector routing (DSDV) for mobile computers. In ACM SIGCOMM, 1994; Bommaiah E, Liu M, McAuley A, Talpade R. AMRoute: adhoc multicast routing protocol. Internet-Draft, draft-talpademanet-amroute-00.txt, August 1998; Work in progress. 11. Ballardie A, Francis P, Crowcroft J. Core based trees (CBT): An architecture for scalable inter-domain multicast routing. In Proc. of the 1993 ACM SIGCOMM Conference, San Francisco, CA, September 1993; Garcia-Luna-Aceves JJ, Madruga E. A multicast routing protocol for ad-hoc networks. In Proc. of IEEE INFOCOM 99, New York, NY, June Sinha P, Sivakumar R, Bharghavan V. MCEDAR: multicast core-extraction distributed ad-hoc routing. In Proc. of the Wireless Communications and Networking Conference, Sinha P, Sivakumar R, Bharghavan V. CEDAR: Coreextraction distributed ad-hoc routing. In Proc. of IEEE INFOCOM 99, Royer EM, Perkins CE. Multicast operation of the ad-hoc ondemand distance vector routing protocol. In Proc. of MOBI- COM 99, Seattle, Washington, USA. 16. Perkins CE, Royer EM. Ad-hoc on-demand distance vector routing. In Proc. of 2nd IEEE Wksp. Mobile Computer system and Applications, February 1999; Bhattacharya R, Ephremides A. A distributed multicast routing protocol for ad-hoc (flat) mobile wireless networks. ACM- /Baltzer Journal of Cluster Computing: Special Issue on Mobile Computing 1998; 1(2). 18. Johnson DB, Maltz DA. Dynamic source routing in ad-hoc wireless networks. Mobile Computing 1996; Scott Corson M, Batsell SG. A reservation-based multicast(rbm) routing protocol for mobile networks: Initial route construction phase. ACM/Baltzer Wireless Networks 1995; 1(4): Ephremides A, Corson MS. A distributed routing protocol for mobile wireless networks. ACM/Baltzer Wireless Networks 1995; 1: Wu CW, Tay YC. AMRIS: A multicast protocol for ad-hoc wireless networks. In Proc. of MILCOM 99, 1999; 1: Ji L, Corson MS. A lightweight adaptive multicast algorithm. In Proc. of Globecom 98, Park V, Corson MS. A highly adaptive distributed routing algorithm for mobile wireless networks. In Proc. of IEEE INFOCOM 97, Kobe, Japan, Lee SJ, Chiang C-C. On-demand multicast routing protocol. In Proc. of IEEE WCNC 99, New Orleans, LA, 1999; Chiang C-C, Gerla M, Zhang L. Forwarding group multicast protocol (FGMP) for multihop, mobile wireless networks. ACM/Baltzer Journal of Mobile Computing 1998; 1(2). 26. Deering SE, Cheriton DR. Multicast routing in datagram internetworks and extended LANS. ACM Transaction on Computer Systems 1990; 37(2): Ho C, Obraczka K, Tsudik G, Viswanath K. Flooding for reliable multicast in multi-hop ad hoc networks. In Proc. of MOBI- COM 99, Seattle, WA, USA. 28. Ko Y-B, Vaidya NH, Location-aided routing (LAR) in mobile ad hoc networks. In Proc. of MOBICOM 98 October Navas JG, Imielinski T. GeoCast Geographic addressing and routing. In Proc. of MOBICOM 97 Conference, Budapest, Hungary, 1997; Markoulidakis JG, Lyberopoulos GL, Tsirkas DF, Sykas ED. Mobility modeling in third-generation mobile telecommunications systems. IEEE Personal Communications, August Basagni S, Chlamtac I, Syrotiuk VR, Woodward BA. A distance routing effect algorithm for mobility(dream). In Proc. of Mobicom 98, Dallas, Texas, USA, January 1998; Johnson DB, Maltz DA. Dynamic source routing in adhoc wireless networks. Kluwer Academic Publishes, 1996: McDonald AB, Znati TF A mobility-based framework for adaptive clustering in wireless ad hoc networks. IEEE Journal on Selected Areas in Communications 1996; 17(8). 34. Chiang C-C. Wireless Network Multicasting. Ph.D dissertation, University of California, Los Angeles, Department of Computer Science, May Hass ZJ. A new routing protocol for the reconfigurable wireless networks. In Proc. of ICUPC 97, 1997; Ko YB, Vaidya NH. Geocasting in mobile ad hoc networks: location-based multicast algorithms. In Proc. of IWMCSA 99, New Orleans, USA. 37. Das S, Castaneda R, Yan J, Sengupta R. Comparative performance evaluation of routing protocols for mobile, ad-hoc networks. In 7th Int. Conf. on Computer Communications and Networks (IC3N), October 1998;

16 130 S. PAPAVASSILIOU AND B. AN 38. Hong X, Gerla M, Pei G, Chiang C-C, A group mobility model for ad-hoc wireless networks. In Proc. of ACM/IEEE MSWiM 99, Seattle, WA, August, Ramanathan R, Steenstrup M. Hierarchically-organized, multihop mobile wireless networks. Mobile Networks and Applications 1998; 3: Authors Biographies Symeon Papavassiliou was born in Athens, Greece, in December He received the Diploma in Electrical Engineering from the National Technical University of Athens, Greece, in 1990 and the M.Sc. and Ph.D. degrees in Electrical Engineering from Polytechnic University, Brooklyn, New York in 1992 and 1995 respectively. During his stay at Polytechnic University he had been a Research Fellow in the Center for Advanced Technology in Telecommunications (CATT). From 1995 to 1996 Dr. Papavassiliou was a Technical Staff Member at AT& T Bell Laboratories in Holmdel, New Jersey, and from 1996 to August 1999 he was a Senior Technical Staff Member at AT&T Laboratories in Middletown, New Jersey. From June 1996 till August 1999 he was also an Adjunct Professor at the Electrical Engineering Department of Polytechnic University, Brooklyn, NY. Since August 1999he has been an AssistantProfessor at the Electrical and Computer Engineering Department of New Jersey Institute of Technology, Newark, New Jersey. Dr. Papavassiliou was awarded the Best Paper Award in INFOCOM 94 and the AT&T Division Recognition and Achievement Award in Dr. Papavassiliou has an established record of publications, he has been a reviewer for many journals, conferences, and for the National Science Foundation (NSF) Dr. Papavassiliou is acting as External Technical Auditor/Evaluator for the European Commission s Advanced Communications Technology and Services Research Programs. His main research interests lie in the areas of computer and communication networks with emphasis on wireless communications and high-speed networks, network design and management, TCP/IP and internetworking, computer network modeling and performance evaluation and optimization of stochastic systems. Dr. Papavassiliou is a member of IEEE and the Technical Chamber of Greece. Beongku An received his B.S. degree in electronic engineering from Kyungpook National University, South Korea, in 1988 and his M.S. degree in electrical engineering from Polytechnic University, Brooklyn, New York, in He is currently working towards a Ph.D. degree in electrical engineering at New Jersey Institute of Technology, New Jersey. His research interests include multicast protocols for ad-hoc networks and Internet, mobility modeling and management, and geocasting technologies.