A Comparative Study of Power Dissipation in MANET Routing Protocols
Sylvia Grace, Rajeev Kumar, Anupama Potluri
Department of Computer and Information Sciences,
University of Hyderabad, Hyderabad 500 046.
, ,
Abstract
MANETs need to conserve power since most nodes are powered by a battery source. Here we study the relative power dissipation performance of two MANET routing protocols, AODV (Ad hoc On-demand Distance Vector) and DSR (Dynamic Source Route). The protocols chosen are significant due to their being candidates for standardization by IETF. The study was conducted through simulations using the ns-2 network simulator. We demonstrate that the performance of DSR and AODV varies because of the inherent difference in their Route Discovery and Route Maintenance mechanisms. Based on our observations, we make recommendations about the mobility scenarios in which each may be preferred.
1. Introduction
Mobility affects the power dissipation of the nodes in a MANET. This is because of the high overhead incurred in Route Discovery and Route Maintenance in mobile nodes as compared to static wired nodes. In static wired nodes, the network topology remains constant over a period of time, so the routes are stable. However in mobile nodes, because of the high mobility of the nodes the network topology is dynamic and the routes keep changing. So a mobile node in a MANET needs to be aware of the latest route to another node to communicate with it. Due to the high overhead associated with Route Discovery and Route Maintenance, reactive or on-demand protocols like AODV, TORA and DSR are preferred. AODV and DSR are significant as they are candidates for standardization by IETF. Although recent studies in this field have compared the performance of these two reactive protocols, the energy issue has not been adequately addressed. Most studies, [3], [4] and [5], have restricted the performance metrics to packet delivery fraction, average end-to-end delay of data packets, normalized routing load and normalized MAC load. However, Mobile Ad-Hoc Networks (MANETs) must contemplate energy-constrained operations due to the nodes’ strong dependency on batteries. This paper attempts to study the effect of mobility on power dissipation of nodes in a MANET. An earlier study [6] discusses the effect of different mobility models on the average power consumption of the network. Our work differs in the mobility model used as well as the attempt to understand the effect on individual nodes – especially the difference between nodes participating in both data plane and control plane operations. Through the simulations, we also demonstrate the mobility scenarios in which one protocol may be preferred over the other.
This paper is organized as follows. A brief description of AODV and DSR is provided in Section 2. A detailed description of our simulation environment follows in Section 3. It contains details of the network topology, the mobility models and the traffic model used in the simulation. The simulation results and their interpretations are provided in Section 4. A discussion of the results is done in Section 5. In Section 6, we draw conclusions and also make our recommendations for the mobility scenarios in which each protocol may be preferred over the other. The future research direction is described in Section 7.
2. A Description of the Protocols
2.1. AODV (Ad-hoc On-demand Distance Vector)
AODV is an on-demand protocol [1]. It discovers routes only when the source attempts to send packets. It uses traditional routing tables, one entry per destination. It relies on routing table entries to propagate an RREP (Route reply) back to the source and uses the table subsequently to route data packets to the destination. AODV uses sequence numbers maintained at each destination to determine freshness of routing information and to prevent routing loops. An important feature of AODV is the maintenance of timer-based states in each node, regarding utilization of individual routing table entries. A routing table entry is expired if not used recently. The recent specification of AODV uses an expanding ring search initially to control the RREQ (Route Request) flood.
2.2. DSR (Dynamic Source Route)
The key distinguishing feature of DSR is the use of source routing [2]. The data packets carry the source route in the packet header. Route discovery works by flooding the MANET with RREQs. Each node receiving a RREQ re- broadcasts it, unless it is the destination or it has a route to the destination in its route cache. RREQ and RREP are also source routed. DSR makes very aggressive use of route caching. Multiple routes to a destination, discovered initially during the Route Discovery process and later through promiscuous listening, are stored in the route cache. An important optimization is packet salvaging. An intermediate node can use an alternate route from its own cache when a data packet meets a failed link on its source route.
3. Simulation Environment
The primary approach for this study was computer simulations. The network simulator ns-2 developed by the VINT research group at University of California at Berkeley has been used for the simulations. The Monarch research group at Carnegie Mellon University extended the ns-2 simulator to include wireless scenarios with mobile nodes. The CMU wireless extension also includes energy models for the mobile nodes. This was used in our study.
3.1. MANET Model
A MANET of 30 nodes with a simulation time of 180 seconds was considered. The mobile nodes were placed on a 670 X 670 flat grid. Both DSR and AODV were used as the routing protocols. 10 pairs of unique source-destination were chosen out of 30 nodes. Thus, 20 different nodes were involved in the communication. The mobility of the nodes was in the range of 0-5m/s, 0-10m/s and 0-20m/s. For each case, the average power levels of the nodes in the MANET was plotted every 10 seconds. Then, the power levels of 3 source and destination pairs and 2 forwarding nodes (router) were plotted.
3.2. Mobility Model
For the MANET, the topology configurations used is: 670 m x 670 m field with 30 nodes. At a specified time, each mobile node starts its journey from current location to a specified destination with a chosen speed (0-5/ 0-10/ 0–20 m/s, depending on the mobility scenario). Once the destination is reached, a new destination is targeted. There is no fixed pause time. If the node reaches the destination ahead of its next move, it pauses; else, it continues to move. Three different mobility details were provided in three files, each corresponding to low, medium and high mobility. Identical mobility model was used in simulations across protocols to yield fair result.
3.3. Traffic Model
Continuous bit rate (CBR) traffic source were used. The CBR traffic, once started, continued throughout the simulation. 10 source –destination pairs are chosen in such a manner that they are spread across the network and the path between them changes often. The hop distance of some paths is less and some is more. The forwarding nodes that participate in data plane operation (i.e. routing) were chosen from nodes that handle multiple traffic at some point in time. The same traffic scenario was used in simulations with different mobility models and across protocols to yield fair result.
3.4. Energy Level of a Node
The initial energy of every node was set to 120 Joules. A high initial energy level was chosen to prevent ‘node deaths’. In case of low initial energy levels, it was observed that few nodes, which actively participated in control and data plane operations, exhausted their battery power after some time. They were then unable to participate in further operations. Power consumption for transmission of a data packet was taken as 0.6 Joules. This is chosen irrespective of packet sizes. For receiving a data packet it was taken to be 0.3 Joules. The residual energy recorded in the trace files every 10 seconds was extracted and plotted for all scenarios.
4. Performance results
4.1.Effect of mobility
4.1.1 Performance of DSR in different mobility scenarios
DSR performs well in low mobility scenario. This is because of availability of multiple paths in route caches and packet salvaging. In low mobility scenarios, the link failures are few in number. If a path fails, the node uses another path from its route cache. The possibility of all routes becoming stale is less. So Route Discovery is initiated infrequently. Less routing overhead translates to lesser power dissipation of nodes. Also successful route caching reduces the propagation of route request. If an intermediate node is unable to forward the data packets then it resorts to packet salvaging. In case of a failed link on the source route, the intermediate node uses an alternate route from its own cache. In the worst case that its route cache is stale, it may also initiate Route Discovery to the destination. These optimizations added to the protocol further reduce the propagation of Route Requests. Fewer nodes then participate in Route Discovery. So the average energy level of MANET remains high.
As the mobility of the nodes increases, DSR’s performance decreases greatly as time increases (see Figure 1). Initially, however, the performance is slightly better than low mobility. This can be observed in the time interval of approximately 20-60s. For a discussion of these results, please refer to Section 5. In a medium mobility scenario, the intermediate node may still be able to find an alternate path from its cache. However, this cannot be guaranteed at all times. In the unlikely event that its route cache is stale, it may initiate Route Discovery and successfully discover an alternate path. Packet salvaging may still be successful as the chance that the destination has become unreachable is remote. But decrease in the average energy levels is observed.
This deterioration is significant in high mobility scenario. Due to high mobility of nodes, there are frequent link failures. The possibility of all
the routes in the cache being stale is very high. So route caching and packet salvaging offer no added advantage. However, they add to the end-to-end delay thereby worsening the overall performance. The frequent link failures encountered in high mobility scenario necessitates frequent Route Discovery. So more routing overhead translates to lower energy levels of nodes in the MANET. As the mobility increases the degradation in performance becomes more evident.
4.1.2. Performance of AODV in different mobility scenarios
In low mobility scenarios, the link failures are few in number. As a node has only one next hop address corresponding to a destination, Route Discovery needs to be initiated only if a path fails. Since there are few link failures, Route Discovery
is not initiated frequently. So less routing overhead translates to lesser power dissipation of nodes. The average energy level of MANET is not affected greatly. Moreover, AODV uses expanded ring search to contain the propagation of RREQs (Route Requests). This reduces the number of nodes participating in Route Discovery. This further ensures that the average power consumption is less.
It can be seen from Figure 2 that the power consumption is not as drastically different for AODV as in DSR. This shows that up to a point mobility does not have a big impact on power consumption in AODV.
Performance, however, degrades significantly in high mobility scenario as can readily be observed from Figure 2. Due to high mobility of nodes, there are frequent link failures. This necessitates frequent Route Discovery. So more routing overhead translates to lower energy levels of nodes in the MANET. As more nodes are affected by the control plane operations, a marked decrease in average power consumption is seen.
4.2. Effect of varying the sources
The performance of AODV and DSR is compared by varying the number of sources as well as the mobility.
At low mobility, DSR outperforms AODV in all cases (Figure 3). However, the difference in their performance is not much when the number of sources is 5. As the number of sources increases, i.e. network load increases, significant difference can be observed.
In medium mobility scenario, there is not much difference in the performance of the two protocols. This is evident from Figure 4.
At high mobility, initially the performance of both the protocols is comparable. However, when the number of sources is 5 and 15, DSR gains over AODV at the later stages. This is not observed when the number of sources is 10.
4. 3. Performance of forwarding node versus non-forwarding node
A forwarding node is a node that actively participates in control plane as well as data plane operations. A non-forwarding node, in contrast, participates only in control plane operations. The forwarding node that was considered is Mobile Node 9. It was chosen, as it takes part in data plane operation. It also handles more than one flow at different times. Node 15 was chosen as a forwarding node. It lies on the edge of the MANET and hence does not participate in forwarding. At medium mobility, the difference in the energy consumed by Node 9 and Node 15 is not significant. However, DSR outperforms AODV at both low and high mobility scenarios.