the Surroundings Repair Method (SRM) for route maintenance. The ISM limits the ... by using the SRM, nodes attempt to repair the broken routes ..... computer science. ... Ranjan Bose (M '98) did his B.Tech. in Electrical Engineering from the.
Bandwidth-efficient routing protocol for mobile ad hoc networks R.S. Al-Qassas, A. Al-Ayyoub and M. Ould-Khaoua Abstract: Mobile ad hoc networks (MANETs) have dynamic irregular topologies by nature, and suffer from inherent limitations such as limited bandwidth and power. A number of routing protocols have been proposed in the past few years to deal with these issues efficiently. The paper proposes and evaluates a new routing protocol, referred to here as the vector routing protocol (VRP). One of the main features of the VRP is its lower communication overhead to establish a route from source to destination. Results obtained through simulation experiments reveal that the new VRP algorithm achieves a lower communication overhead than the well known DSDV and AODV protocols, especially in low mobility environments.
1
Introduction
The mobile ad hoc network (MANET) is a collection of wireless mobile nodes that forms a temporary network without the need for any infrastructure or centralised administration. In such an environment, it may be necessary for one mobile node to enlist the aid of others in forwarding a packet to its destination due to the limited propagation range of each mobile node’s wireless transmissions [1]. The communication in MANETs is peer-to-peer as the mobile nodes communicate directly with one another. In MANETs, the topology may change frequently as nodes may move or power themselves off to save energy. Consequently, we often need to collect connectivity information periodically in order to get a consistent view of the network, which in turn increase the bandwidth consumption overhead resulting from collecting this information. MANETs have limited bandwidth, and therefore need an efficient routing protocol that can establish and maintain routes for both stable and dynamic topologies with minimum bandwidth consumption. A number of routing protocols, including DSDV [2], CGSR [3], WRP [4], LAR [5], ABR [6], AODV [7], TORA [8], DSR [9] and ZRP [10], have been proposed to solve the routing problem. Ad hoc routing protocols can be classified into two main categories: proactive (or table driven) protocols and reactive (or source initiated on-demand driven) protocols. Proactive protocols, such as those described in [2 – 4] try to maintain consistent and up-to-date routing information (routes) from each node to every other node in the network. Topology updates are propagated throughout the network to
maintain a consistent view of the network. Keeping routes for all destinations has the advantage that communication with arbitrary destinations experiences minimal initial delay. Furthermore, a route can be immediately selected from the route table. However, these protocols have the disadvantage of generating additional control traffic that is needed to continually update stale route entries. Reactive protocols, such as those proposed in [6 – 9], establish routes only when they are requested by network nodes. When a source node requires a route to a destination, it initiates a route discovery process within the network. Once a route has been established, some form of route maintenance procedure is used to maintain it, until either the destination become inaccessible or the route is no longer desired. These protocols tend to use less bandwidth for maintaining the route tables at every node. However, the latency drastically increases, leading to long delays before a communication can start. This is because a route to the destination has to be acquired first. Moreover, these protocols have large control overhead when the number of source to destination connections is large. The overall goal of this present study is to develop an efficient routing protocol that can deal with the dynamic nature of MANETs and their bandwidth limitations. The protocol must be able to establish and maintain correct routes between any source and destination efficiently, by reducing the number and size of routing packets needed for establishing and maintaining routes in order to obtain better utilisation of the available bandwidth. The proposed protocol that will be presented below is still under development and extensive testing. In its current form it is able to deal efficiently with low mobility ad hoc networks. 2
q IEE, 2003 IEE Proceedings online no. 20030811 doi: 10.1049/ip-sen:20030811 Paper received 25th April 2003 R.S. Al-Qassas and M. Ould-Khaoua are with the Department of Computing Science, University of Glasgow Glasgow G12 8RZ, UK A. Al-Ayyoub is with the Faculty of Computer Studies, Arab Open University, Amman 11953, Jordan 230
Existing routing protocols
Before presenting the proposed routing protocol, this section describes two well-known routing protocols, notably the destination sequenced distance vector (DSDV) [2] and ad hoc on-demand distance vector (AODV) protocols [7]. This is to enable us to contrast the operations of our new protocol against those of the DSDV and AODV as they achieve good performance compared to their counterparts of proactive and reactive routing protocols. IEE Proc.-Softw., Vol. 150, No. 4, August 2003
2.1 Destination sequenced distance vector In this protocol each node has a routing table that contains for all reachable destinations the next-hop, number of hops for that destination (hop count) and the destination sequence number. DSDV requires that each node periodically broadcast routing updates. To take care of topology changes that may happen in the time between periodical updates, DSDV uses triggered updates. DSDV uses two types of update messages: full dump and incremental dump. The full dump carries all available routing information, while the incremental dump carries only the information that has changed since the last dump. DSDV utilises the sequence number to guarantee loop freedom. The sequence number shows the freshness of a route; routes with higher sequence numbers are preferable. If two or more routes have the same sequence number the route with a lower hop count is taken. When a node S detects that a route to a destination D has broken, the sequence number is increased and the hop count becomes infinite. So when node S advertises its routes after then, the route to D will have an infinite hop count and a larger sequence number so that the route is ignored by other nodes.
2.2 Ad hoc on-demand distance vector AODV has been suggested as an improvement over DSDV; it typically minimises the number of required broadcasts by creating routes on an on-demand basis. When a node wants to send a message to another node and it does not already have a valid route to that destination, it initiates a path discovery process to locate the other node. The node broadcasts a route request packet to its neighbours, which in turn forward the request to their neighbours, and so on, until either the destination or an intermediate node with a fresh route to the destination is located. Once the route request reaches the destination or an intermediate node with a fresh route, the destination/intermediate node responds by unicasting a route reply packet back to the neighbour from which it first received the route request. When a node detects that a route to a node is no longer valid, it removes the routing entry and sends a link failure message to the neighbours that are actively using the route, informing them that this route is no longer valid. For this purpose, AODV uses an active neighbour list to keep track of the neighbours that are using a particular route. The nodes that receive this message will repeat this procedure. The message will eventually be received by the affected sources, which can choose to either stop sending data or request a new route by sending out a new rout request. AODV uses destination sequence numbers to ensure that all routes are loop-free and contain the most recent route information. Each node maintains its own sequence number. The source node includes in the route request the most recent sequence number it has for the destination. Intermediate nodes can reply to the route request only if they have a route to the destination whose corresponding destination sequence number is greater than or equal to that contained in the route request. 3
Proposed vector routing protocol
This section describes the proposed vector routing protocol (VRP). Let the network contains n nodes, each having unique identity 0; 1; 2 . . . ðn 1Þ: We say that there is a link between any two nodes if each of the two nodes is within the transmission range of the other, assuming that the link is bidirectional. As in most previous studies [2 – 6, 7], we IEE Proc.-Softw., Vol. 150, No. 4, August 2003
assume that all nodes are equal and that all transmission errors are recovered. In VRP, each node has three vectors: neighbourhood, routing and access vectors. The neighbourhood vector is used to store neighbourhood information. The routing vector contains the needed information to send packets to their destinations. The access vector contains accessibility information which is used for computing the routing vector. Before describing the routing protocol let us define these vectors. Let vu be the neighbourhood vector for node u, where u v ¼ ðvu0 ; vu1 ; . . . ; vuðn1Þ Þ. The vector vu consists of n bits, where vui ¼ 1 for every neighbouring node i. All bits in this vector are initialised with the value 0, except vuu ¼ 1. Let ru be the routing vector for node u, where u u Þ. The vector ru consists of n numbers, r ¼ ðr0u ; r1u ; . . . ; rðn1Þ u where ri presents the id of the next node along the path to the destination node i. We say that node u has a path to the destination node i, if and only if riu 6¼ u. Initially, riu ¼ u for all 0 i n 1 and 0 u n 1. Let au be the access vector for node u, where u a ¼ ðau0 ; au1 ; . . . ; auðn1Þ Þ. The vector au consists of n bits, where the ith bit indicates weather the node u has a path to the node i or not. We say that node u has a path to node i if aui ¼ 1; otherwise we say there is no path connecting u and i at this time. All bits in this vector are initialised with the value 0, except auu ¼ 1. Every node u uses the access vectors of its neighbouring nodes for computing the routing vector ru , as will be seen in the sequel. The proposed routing protocol computes the routing information in a distributed fashion. The computation consists of two phases. The first phase is the start-up phase. This is executed when the network is first activated. In this phase, the protocol establishes routing paths to all nodes in the network. The second phase is the maintenance phase, where the routing protocol deals with topology changes in order to maintain correct routing information. The start-up phase is described in Fig. 1. As a first stage, each node u broadcasts ‘hello’ messages to detect its neighbouring nodes. When node u receives a “hello” message from a neighbouring node i it sets vui to 1. After detecting all neighbouring nodes, the two vectors au and ru are updated by using (1) and (2), respectively. At this stage the information in ru is up-to-date for the nodes that are within one hop from node u. In other words, au ¼ v u riu ¼
i u
if vui ¼ 1 if vui ¼ 0
ð1Þ
ð2Þ
As a next stage, each node u sends au to every neighbouring node w, and waits to receive aw from them. When u receives the vector aw , it updates the two vectors au and ru using (3) and (4). After updating the vectors, a new stage starts: node u sends the new au to every neighbouring node w and waits to receive aw . When node u receives the vector aw , it updates the two vectors au and ru : 1 if aui ¼ 0 and awi ¼ 1 aui ¼ ð3Þ 0 if awi ¼ 0 and riu ¼ w riu ¼
w u
if aui ¼ 0 and awi ¼ 1 if aui ¼ 0
ð4Þ
The process of sending, receiving and updating the vectors continues until one of the following conditions is met: (i) until each bit in the vector au is 1, implying that the node 231
Fig. 1 Outline of the start-up phase in the proposed VRP
u has routes to all nodes in the network; (ii) until the value of au remains unchanged for two successive stages, implying that the node u has routes to all reachable nodes in its subnetwork, and there is partition in the network. The process of computing routing information needs at most h stages, where h is the diameter of the network. The maintenance phase described in Fig. 2 is triggered when node u detects a change in the set of neighbouring nodes. When a node u detects that the node y, which is one of its neighbouring nodes, is out of its transmission range, it immediately updates the three elements auy ¼ 0; vuy ¼ 0 and ryu ¼ u, and then it sends au to its neighbouring nodes. On the other hand, when the node u detects the existence of a new neighbouring node y, it updates the vectors auy ¼ 1; vuy ¼ 1 and r uy ¼ y, then it sends au to the neighbouring nodes if auy was zero before update. Each time node u receives aw from a neighbouring node w it updates the two vectors au and ru using (3) and (5). Then it sends au to the neighbouring nodes only if the received vector aw is different from au before update.
Neighbourhood changes could be detected by using ‘hello’ messages at the network layer or by beaconing messages at the MAC layer. In our present study, the VRP protocol uses the MAC layer functionalities [11 – 13] for neighbourhood detection: w if aui ¼ 0 and awi ¼ 1 riu ¼ u if awi ¼ 0 and riu ¼ w ð5Þ
4
Simulation and performance evaluation
The viability of the proposed protocol has been assessed through simulation techniques and using a common network simulation software, namely ns-2 simulator [14]. In our simulations we have tested VRP on widely accepted simulation parameters [15 – 17]. The evaluation is based on the simulation of 50 wireless nodes forming a MANET over a flat space of size ð1500 m 300 mÞ for 900 s of simulated time. We have used constant bit rate (CBR) data traffic
Fig. 2 Outline of the maintenance phase in the proposed VRP 232
IEE Proc.-Softw., Vol. 150, No. 4, August 2003
sources with packet sizes of 64 bytes. The sending rate is 4 packets per second. The number of CBR sources has varied at 10, 20, 30, 50 and 100 sources. Nodes in the simulation move according to the random waypoint model [9] with a maximum speed of 1 m/s. Each node remains stationary for a pause time. When the pause time expires, the node selects a random destination in the simulation space and moves towards it. When a node reaches its destination, it pauses again for the same pause time. The previously described node behaviour is repeated until the end of simulation. The used pause times in this simulation are 300, 600 and 900 s. We have used scenario files that contain the description of the simulation environment that includes node distribution over the simulation space, node movements and communication patterns. The scenario files differ in the number of traffic sources and pause time. In this study we have used the two standard performance metrics: routing overhead and packet delivery ratio. The routing overhead metric is further refined into two sub-metrics: routing overhead (in kbytes) and number of routing packets. The routing overhead (in kbytes) is defined as the sum of routing packets sizes sent during the simulation time. For packets sent over multi-hop, each hop is considered as packet sent. Therefore, for each hop along the multi-hop path we add the packet size to the overall routing overhead. Of course, low routing overhead means better bandwidth utilisation. The number of routing packets is the total number of routing packets sent during the simulation time. Again, for packets sent over multi-hop paths, each hop is considered as a packet sent. When the number of routing packets is high this may increase the probability of packet collision and may delay the delivery of data packets to their final destination. The packet delivery ratio is defined as the total number of received packets at the final destinations to the number of packets sent from traffic sources. This metric shows the completeness and correctness of the routing protocol because it shows how much the routing protocol is able to deliver data packets to their final destinations. The above performance metrics have been applied to the three protocols: VRP, DSDV and AODV. The results we have obtained are for two kinds of simulation environments. The first one is a stable environment with different numbers of traffic sources while the second is a dynamic environment with a low number of traffic sources. The proposed protocol (VRP) deals with dynamic environments using the maintenance phase, which is triggered when the MANET topology changes. These changes are detected using the MAC layer functionalities that include neighbourhood change detection using beaconing messages and link failure detection. In this paper, we have used only link failure detection. As a consequence, the results for dynamic simulation environments do not reflect the true level of performance that the VRP could potentially achieve. In this paper what we have measured is the effects of using the link failure detection alone without using the neighbourhood changes detection. Figure 3 depicts results for the routing overhead for the three routing protocols VRP, AODV and DSDV in stable simulation environments with different numbers of traffic sources, varied from 10 sources to 100 sources. The Figure reveals that VRP achieves good performance especially when the traffic load is high. It also shows that the VRP provides a better performance than DSDV in all cases. Figure 4 shows the routing overhead for the three routing protocols in dynamic simulation environments with pause IEE Proc.-Softw., Vol. 150, No. 4, August 2003
Fig. 3 Routing overhead for stable simulation environment with different numbers of traffic sources
Fig. 4 Routing overhead for dynamic simulation environment with fixed number of traffic sources equals 10
times 300, 600 and 900 s but with a low number of traffic sources (10 sources). The results show that VRP has a lower routing overhead than its DSDV counterpart. The VRP and AODV exhibit very close performance behaviour. Bearing in mind that VRP maintains a complete set of routes to all nodes in the network, contrary to AODV, the quality of VRP routing overhead is very acceptable compared to its counterparts. Figure 5 shows the number of routing packets used by the three routing protocols in stable simulation environments with different numbers of traffic sources, varied from 10 sources to 100 sources. We can see in this Figure that VRP achieves better performance than DSDV under all traffic conditions. Moreover, the performance of VRP is better than AODV when the number of traffic sources is more than 30 sources and the induced routing overhead of VRP becomes remarkably lower than both DSDV and AODV when the number of traffic sources is over 50. Figure 6 shows the number of routing packets used by the three routing protocols in dynamic simulation environments with pause times 300, 600 and 900 s but with low numbers of traffic sources. VRP performs as well as DSDV; however, AODV outperforms the other two protocols under these circumstances. This is expected for an on-demand protocol with quite low traffic loads [15].
Fig. 5 Number of routing packets for stable simulation environment with different numbers of traffic sources 233
should the neighbourhood changes be captured and accounted for. 5
Fig. 6 Number of routing packets for dynamic simulation environment with fixed number of traffic sources equals 10
Fig. 7 Packet delivery ratio for stable simulation environment with different numbers of traffic sources
Fig. 8 Packet delivery ratio for dynamic simulation environment with fixed number of traffic sources equals 10
Figure 7 shows the packet delivery ratio for the three routing protocols in a stable simulation environment with different numbers of traffic sources. The results achieved by the three protocols are close to each other. However, VRP achieves better results as the number of traffic sources approaches 100. Figure 8 shows the packet delivery ratio for a dynamic simulation environment with pause times 300, 600 and 900 s but with a fixed number of traffic sources (10 sources). The Figure suggests a poor VRP delivery ratio for frequent topology changes. However, this result does not reflect the true level of performance that the VRP could potentially achieve as we have explained previously in this Section. What we have done here is an investigation of the possibility of using link failure detection alone without using neighbourhood change detection, and how much this would affect the protocol performance. A more accurate measure for the VRP delivery ratio would be obtained,
234
Conclusions
This paper has proposed the vector routing protocol (VRP) as a new routing algorithm for MANETs. Although the VRP is proactive as it tries to maintain routes to all nodes in the network, it does not flood the network with routing packets. The VRP relies only on local communication when establishing routes between nodes. The new protocol yields optimal routes in terms of the number of hops when the network topology is stable, even though it considers path optimality as a second priority. A comparative analysis against the well known AODV and DSDV protocols has revealed that the VRP manages to achieve a higher packet delivery ratio in stable environments while maintaining minimal communication overhead. The authors plan to exploit the use of the MAC layer functionalities for detecting topology changes, which includes neighbourhood changes and link failure detection. This would result in a great improvement in the performance of VRP, enabling it to deal efficiently with highly mobile networks. As a next step of this research work, they plan to conduct extensive experiments to evaluate the performance of the VRP in high mobility environments. 6
References
1 Johnson, D.B.: ‘Routing in ad hoc networks of mobile hosts’. Proc. IEEE Computer Society, Workshop on Mobile computing systems and applications, Santa Cruz, CA, December 1994, pp. 158–163 2 Perkins, C.E., and Bhagwat, P.: ‘Highly dynamic destination-sequenced distance vector routing (DSDV) for mobile computers’. Proc. ACM SIGCOMM’94, London, September 1994, pp. 234 –244 3 Chiang, C.-C.: ‘Routing in clustered multihop, mobile wireless networks with fading channel’. Proc. IEEE SICON’97, April 1997, pp. 197 –211 4 Murthy, S., and Garcia-Luna-Aceves, J.J.: ‘An efficient routing protocol for wireless networks’, ACM Mob. Netw. Appl. J., 1996, pp. 183–197, Special Issue on Routing in mobile communication networks 5 Ko, Y.-B., and Vaidya, N.H.: ‘Location-aided routing (LAR) in mobile ad hoc networks’, Wirel. Netw., 2000, 6, (4), pp. 307–321 6 Toh, C.-K.: ‘Associativity based routing for ad hoc mobile networks’, Wirel. Pers. Commun. J., 1997, 4, (2), pp. 103–139, Special Issue on Mobile networking and computing systems 7 Perkins, C.E., and Royer, E.M.: ‘Ad-hoc on-demand distance vector routing’. Proc. 2nd IEEE Workshop on Mobile computing systems and applications, New Orleans, LA, February 1999, pp. 90 –100 8 Park, V.D., and Corson, M.S.: ‘A highly adaptive distributed routing algorithm for mobile wireless networks’. Proc. IEEE INFOCOM ’97, Kobe, Japan, April 1997, pp. 1405– 1413 9 Johnson, D.B., and Maltz, D.A.: ‘Dynamic source routing in ad hoc wireless networks’, Chapter 5 in Imielinski, T. and Korth, H. (Eds.): ‘Mobile computing’ (Kluwer Publishing Company, 1996), pp. 153–181 10 Haas, Z., and Pearlman, M.: ‘The performance of query control schemes for the zone routing protocol’, IEEE/ACM Trans. Netw. (TON), 2001, 9, (4), pp. 427–438 11 Geier, J.: ‘Wireless LANs’ (Sams, Indianapolis, IN, 2001, 2nd edition) 12 Perkins, C.E.: ‘Ad hoc networking’ (Addison-Wesley, Boston, 2000) 13 Toh, C.-K.: ‘Ad hoc mobile wireless networks; protocols and systems’ (Prentice Hall, Upper Saddle River, NJ, 2001) 14 ‘The network simulator’ ns-2, http://www.isi.edu/nsnam/ns 15 Broch, J., Maltz, D.A., Johnson, D.B., Hu, Y.C., and Jetcheva, J.: ‘A performance comparison of multi-hop wireless ad hoc network routing protocols’. Proc. of ACM/IEEE MobiCom, October 1998, pp. 85–97 16 Das, S.R., Perkins, C.E., and Royer, E.M.: ‘Performance comparison of two on-demand routing protocols for ad hoc networks’. Proc. INFOCOM’, March 2000, pp. 3–12 17 Perkins, C.E., Royer, E.M., and Das, S.R.: ‘Performance comparison of two on-demand routing protocols for ad hoc networks’, IEEE Pers. Commun., 2001, 8, pp. 16– 28
IEE Proc.-Softw., Vol. 150, No. 4, August 2003
