A stability-based distributed routing mechanism to support unicast and multicast routing in ad hoc wireless network

Abstract

An ad hoc network can be envisioned as a collection of mobile routers, each equipped with a wireless transceiver, which are free to move about arbitrarily. In ad hoc wireless networks, even if two nodes are outside the wireless transmission range of each other, they may still be able to communicate in multiple hops using other intermediate nodes. However, the dynamics of these networks, as a consequence of mobility and disconnection of mobile hosts, pose a number of problems in designing routing schemes for effective communication between any pair of source and destination. In this paper, a stability-based unicast routing mechanism, that considers both link affinity and path stability in order to find out a stable route from source to destination, is proposed. It is then extended to support multicast routing as well where only local state information (at source) is utilized for constructing a multicast tree. The performance of the proposed scheme is evaluated on a simulated environment to show that the stability-based scheme provides a unified framework for both unicast and multicast routing and reduces the probability of route error drastically in both the cases.

Introduction

Ad hoc wireless networks are self-organizing network architectures of mobile nodes that are rapidly deployable and that adapt to the propagation conditions and to the traffic and mobility patterns of the network nodes [1]. An ad hoc network can be envisioned as a collection of mobile routers, each equipped with a wireless transceiver, which are free to move about arbitrarily. The mobility of the routers and the variability of other connecting factors result in a network with a potentially rapid and unpredictable changing topology. These networks may or may not be connected with the infrastructure such as internet, but still be available for use by a group of wireless mobile hosts that operate without any base-station or any centralized control. Applications of ad hoc networks include military tactical communication, emergency relief operations, commercial and educational use in remote areas, where the networking is mission-oriented and/or community-based.

There has been a growing interest in ad hoc networks in recent years [2], [3], [4], [5], [6], [7]. The basic assumption in an ad hoc network is that two nodes willing to communicate may be outside the wireless transmission range of each other but still be able to communicate if other nodes in the network are willing and capable of forwarding packets from them. However, the successful operation of an ad hoc network will be hampered, if an intermediate node, participating in a communication between two nodes, either moves out of range suddenly or switches itself off in between message transfer. The situation is worse, if there is no alternative path available between those two nodes.

In general, existing routing protocols can be classified either as proactive or as reactive [1]. In proactive protocols, the routing information within the network is always known beforehand through continuous route updates. The distance vector and link state protocols are examples of proactive scheme. Examples of proactive routing methods in ad hoc network environments are given in [8], [9]. However, these methods require knowing the topology of the entire network and this information needs to be propagated through the network. Consequently, in a highly dynamic environment, these schemes are less efficient.

Reactive protocols, on the other hand, invoke the route discovery procedure on demand only. The family of classical flooding algorithms belongs to this group. Examples of reactive protocols in the context of ad hoc networks can be found in [2], [3], [4], [5], [6]. It has been pointed out that proactive protocols are not suitable for highly mobile ad hoc network, since they consume a large portion of network capacity for continuously updating route information [10]. On the other hand, on-demand search procedure in reactive protocols generates large volume of control traffic and the actual data transmission is delayed until the route is determined. The features, problems and requirements associated with different routing schemes in ad-hoc network are illustrated in [11], [12], [13], [14] through qualitative and simulation studies.

Whatever may be the routing scheme, frequent interruption in a selected route would degrade the performance in terms of quality of service. Therefore, an important issue is to minimize route maintenance by selecting stable routes, rather than shortest routes. In an ad hoc network, relationship among nodes is based on providing some kind of service, and stability can be defined as the minimal interruption in that service. Hence, a notion of stability of a path and its evaluation mechanism in the context of dynamic topology changes in an ad-hoc network is introduced in this work and a distributed routing scheme among mobile hosts is proposed in order to find a path between them, which is stable in a specific context.

The idea of selecting stable routes within a dynamic network has been proposed in [5], [6]. In associativity based routing [5], an optimal route is selected based on the stability of the route. The notion of stability of an intermediate node is based on a Rule of Associativity that states that a node's association with its neighbor changes as it migrates from one wireless cell to another. This migration is such that, after this unstable period, there exists a period of stability, where the node will spend some dormant time within a wireless cell before its starts to move again. In signal-stability based adaptive routing [6], stable routes are selected based on signal strength. The signal strength criteria allow the protocol to differentiate between strong channels and weak channels. Each channel is characterized by the average signal strength at which packets are exchanged between the nodes at either end of the channel. The routes with strong channels are likely to be long-lived (i.e. stable). However, a major drawback of these methods is that the parameter stability is not explicitly evaluated. Moreover, the notion of stability of a path is dynamic and context-sensitive. Stability of a path is the span of life of that path at a given instant of time and it has to be seen in the context of providing a service. A path between a source and a destination would be stable if its span of life is sufficient to complete a required volume of data transfer from source to destination without possible interruption. Hence, a given path may be sufficiently stable to transfer a small volume of data between source and destination; but the same path may be unstable in a context where a large volume of data needs to be transferred.

This work first proposes a stability-based unicast routing mechanism and then extends it to a multicast routing mechanism that depends only on local state information (at source) for constructing a multicast tree. Multicast communication in the context of ad hoc wireless network is a very useful and efficient means of supporting group-oriented applications; where the need for one-to-many data dissemination is quite frequent in critical situations such as disaster recovery or battlefield scenarios. Instead of sending data via multiple unicasts, multicast routing reduces the communication costs by minimizing the link bandwidth consumption and delivery delay.

Research in the area of routing in ad-hoc wireless network has mostly concentrated on designing effective routing schemes for unicast communication. Those routing algorithms were not designed with multicast extensions in mind. Therefore, they do not naturally support multicast routing solutions [15]. Since fixed network multicasting is based on state in routers (either hard or soft), it is fundamentally unsuitable for ad hoc network where topology is changing frequently due to unconstrained mobility. It has been shown that the performance of both hard- and soft-state multicast tree maintenance mechanisms degrade rapidly with increased mobility [16]. Traditional multicast approaches that rely on maintaining and exchanging multicast-related state information are not suitable in highly dynamic ad-hoc network with frequent and unpredictable changing topology [15].

Currently proposed ad hoc multicast routing schemes [16] lie on a spectrum that spans from pure Internet multicast routing based schemes to a pure flooding scheme. Internet multicast routing schemes, as it is currently, generally require the routing nodes to maintain fairly large amount of state information for routing and to use processing power of hosts rather liberally. Feasibility of supporting continuous unlimited mobility is also a question with Internet routing schemes. Only flooding control packets may support unlimited continuous mobility. Flooding will also reduce the amount of state information kept at mobile hosts, and will provide reliable and timely delivery.

FGMP [17], the Forwarding Group Multicast Protocol, proposes a scheme that is hybrid between flooding and source based tree multicast. The proposed multicast protocol scheme keeps track not of links but of groups of nodes, which participate, in multicast packets forwarding. To each multicast group G is associated a forwarding group, FG. Any node in FG is in charge of forwarding (broadcast) multicast packets of G. The nodes to be included in FG are elected according to members' requests. Instead of data packets, small membership advertisement packets are used to reduce overhead caused by broadcasting. However, in order to advertise the membership, each receiver periodically and globally floods its member information.

AMRoute [18], the Ad hoc Multicast Routing Protocol, creates a peer group multicast distribution tree using unicast tunnels connecting group members. The protocol has two main components: mesh creation and tree creation. Certain nodes are designated as logical core nodes that initiate mesh and tree creation; however, the core can migrate dynamically according to group membership and network connectivity. Logical cores are responsible for initiating and managing the signaling component of AMRoute, such as detection of group members and tree set up. Bi-directional tunnels are created between pairs of group members that are close together, thus forming a mesh. Using a subset of available mesh links, the protocol periodically creates a multicast distribution tree. AMRoute assumes the existence of an underlying unicast routing protocol and its performance is influenced by the characteristics of the unicast routing protocol being used. The AMRoute simulation runs on top of TORA [5] as underlying unicast protocol. The network dynamicity was emulated by keeping node location fixed and breaking/connecting links between neighboring nodes. Thus, the effect of actual node mobility on the performance is difficult to interpret. Moreover, the signaling generated by underlying unicast protocol (TORA in this case) is not considered in the measurements.

The Core-Assisted Mesh Protocol (CAMP) [19] generalizes the notion of core-based trees introduced for internet multicasting into multicast meshes that have much richer connectivity than trees. A shared multicast mesh is defined for each multicast group. The advantage of using such meshes is to maintain the connectivity even while the network routers move frequently. CAMP consists of the maintenance of multicast meshes and loop-free packet forwarding over such meshes. Multicast packets for a group are forwarded along the shortest path from sources to receivers defined within the group's mesh. CAMP rebuilds meshes at least as fast as CBT and PIM can rebuild trees. However, the effect of mobility on the performance has not been clearly evaluated. The topology under experimentation has 30 routers with high connectivity (average of six neighbors each) and at the most 15 routers out of 30 are assumed to be mobile.

AMRIS [20], the Ad hoc Multicast Routing Protocol utilizing increasing id-number S, assigns an identifier to each node in a multicast session. A pre-multicast session delivery tree rooted at a special node (by necessity a sender) in the session joins all the group members. The tree structure is maintained by assigning identifiers in increasing order from the tree root outward to the other group members. All nodes are required to process the tree set up and maintenance messages that are transmitted by the root periodically. It has been assumed that most multicast applications are long-lived; therefore rapid route reconstruction is of greater importance compared to rapid route discovery. The performance of the proposed scheme has yet to be evaluated.

ODMRP [21], the On-Demand Multicast Routing Protocol, also uses a mesh-based approach for data delivery and uses a forwarding group concept. It requires sources rather than destinations to initiate the mesh building by periodic flooding of control packets. It applies on-demand procedures to dynamically build routes and maintain multicast group membership. A soft-state approach is taken to maintain multicast group member.

This paper proposes a multicast routing mechanism that depends only on local state information (at source) for constructing a multicast tree. It is demand- driven in the sense that whenever a source needs to communicate with a set of destinations, it discovers the routes and creates a multicast tree dynamically. It has been shown that the proposed multicast routing scheme reduces both the control traffic and the data traffic and decreases the delivery delay considerably when compared to multiple unicasts.