Energy Saving Dynamic Source Routing for Ad Hoc Wireless Networks Mohammed Tarique, Kemal E. Tepe, and Mohammad Naserian Electrical and Computer Eng. Dept. University of Windsor Windsor, Ontario N9B 3P4, CANADA Tel: +1-519-253-3000 ext 3426, Fax: +1-519-971-3695 tarique, ktepe,
[email protected]
Abstract In this paper Energy Saving Dynamic Source Routing (ESDSR) protocol is introduced to maximize the life-span of a mobile ad hoc network (MANET). Many theoritical studies show that energy consumption in MANET can be significantly reduced using energy-aware routing protocols compared to fixed-power minimum-hop routing protocols. Two approaches are broadly suggested for energy-aware routing protocols - transmission power control approach and load sharing approach. ESDSR integrates the advantages of those two approaches. In ESDSR, the routing decision is based on a load balancing approach. Once a routing decision is made, link by link transmit power adjustment per packet is done based on a transmit power control approach. We modified Dynamic Source Routing (DSR) protocol to make it energy aware by a network simulator (Network Simulator-2 of University of California). The simulation results show that the proposed ESDSR can save energy up to 40% per packet and it can send 20 % more packets to destinations by spending the same battery power in compare to DSR. keywords: Ad hoc networks, sensor networks, routing,energy saving, power adjustment,network life.
1. Introduction Mobile ad hoc network (MANET) is becoming increasingly popular as a means of providing instant networking to a group of people who may not be within the transmission range of one another. MANET is self-initializing, self-configuring and self-maintaining. It usually consists of battery-operated computing devices which cooperate with each other to transmit packet from a source node to a destination node. The availability of each node is equally important for the enforcement of that kind of cooperation. The
failure of a single node can greatly affect the network performance. Since mobile nodes are usually battery-operated, one of the major reasons of node failure is battery exhaustion. In order to maximize the life-time of a mobile node, it is important to reduce the energy consumption of a node while transmitting packet. In recent years, a number of studies have been done so far to achieve energy conservation in MANET. Those can be broadly classified as transmit power control approach protocols and load distribution approach protocols. The transmit power control approach protocols determine the optimal routing path that minimizes total transmission energy required to deliver data packets from a source to a destination. The load distribution approach protocols focus on balancing energy usage among the nodes by avoiding overutilized nodes while selecting a routing path. There is no clear consensus that any particular protocol or a class of protocols are suitable for all scenarios. We presented a brief review of some of those studies here. Routing protocol based on minimizing the amount of power (energy per bit) required to get a packet from a source to a destination has been proposed in [5]. The main disadvantage of that protocol is that it always select the leastpower route. As a result the nodes along the paths ’die’ very soon in compare to other nodes. GPS based transmit power adjustment has been proposed in [4]. GPS information does not contain information about the physical environment such as noise and interference. Many researchers try to adopt load balancing approach protocols to overcome those limitations of transmit power control approach protocols. Minimum battery cost routing protocol is introduced in [6]. It minimizes the summation of inverse of the remaining battery for all nodes on the routing path. In [7], a node determines whether to forward the route-request message or not, depending on its residual battery energy. If it has energy level higher than a threshold, node forwards the routerequest message, otherwise it drops the route request message. The main problem with the second approach protocol
is that it assumes that all nodes transmit at the same power irrespective of the location of the receiver which causes unnecessary over utilization of battery. A considerable energy can be saved if a node transmits at less power when the receiver is located closer to it. We introduce energy saving dynamic source routing (ESDSR) protocol in this paper, which integrates the advantages of those two approaches. In ESDSR, the nodes which has a ’tendency’ to ’die out’ very soon are avoided during the route discovery phase of this protocol. The ’tendency’ of the node to ’die out’ is expressed quantitatively as the ratio of the remaining battery energy and the current transmit power of the node. We call it ’expected life’ of the node. Once the routing decision is made, link by link transmit power adjustment is accomplished depending on the signal strength at which a node receives a packet. Since our protocol is an enhancement of Dynamic Source Routing (DSR) [2] protocol, we will describe the basic operation of DSR briefly in the following section.
2. DSR Protocol The DSR protocol is based on source routing. Network nodes cooperate to forward packets for each other to allow communication over multiple hops between nodes which are not directly within the transmission range of one another. The originator of each packet determines an ordered list of nodes through which the packet will travel to reach the destination. DSR protocol composed of two main mechanisms- route discovery and route maintenance. Route discovery is the mechanism by which a sending node finds a route to the destination. When a source node originates a new packet, it places in the packet header a ”source route” through which the packet will travel to reach the destination. When the source node has some packets to send to a destination, it will search its cache to find a route. If it cannot find a route in the cache, it will initiate a route discovery to find a route to that destination. To initiate the route discovery, source node transmits a route request as a local broadcast packet, which is received by all nodes currently within wireless transmission range of this node. Each route request contains the source and the destination addresses and it also contains a unique request id. Each request also records a list of the nodes through which it has been forwarded. When a node receives this route request packet, it checks whether it is the destination of this route request or not. If it is the destination, it sends a route reply to the source node after copying the accumulated route from the route request packet in the route reply packet. When the source node receives the route reply packet, it records the new source route in its cache and send packet using this new route that it has just discovered. If the node receiving the route request packet determines that it is not the destina-
1
2
3
5 4
Figure 1. Routing in DSR Pt1,E1 1
Pt2,E2
Pt3,E3
2
3
5 4
Pt5,E5
Pt4,E4
Figure 2. Routing in ESDSR tion, it appends its own address in the route request packet and propagates it as a broadcast packet with the same request id. Route maintenance is the mechanism by which a node is able to detect any changes in the network topology such that it cannot send a packet using a route because a link along the route is broken. In DSR, each node transmitting the packet is responsible for confirming that data can flow over the link from that node to the next hop. An acknowledgment provides confirmation that the link is capable of carrying data to the next hop. This type of acknowledgment is provided by the existing MAC layer protocol, (IEEE 802.11). After sending the packet to the next hop, the transmitting node waits for an acknowledgment. If it does not receive an acknowledgment, the transmitting node treats the link to the next hop is ’broken’. It marks all the routes in the route cache that uses that link as ’invalid’. It will return a route error to each node that has sent a packet over that broken link so that all those nodes can update their own route cache as well. After receiving the broken route information from the route error message, the source node tries to find another route from its route cache. If it cannot find any other alternative route available in the cache, it will initiate another route discovery to find another route for that destination. The routing decision is made in DSR based on minimum hop. Routing decision based on minimum hop is depicted in Figure 1. Node 1 is the source and Node 5 is the destination. The source discovered two paths 1 − 2 − 3 − 5 and 1 − 4 − 5 by route discovery mechanism. The source node 1 will choose path 1 − 4 − 5 because it has less number of hops. Once a routing decision is made, the source node starts transmitting packet at full transmit power irrespective of the position of the next hop. When node 4 for-
wards the packet to the destination node 5, it also transmits at full transmit power. The DSR protocol is not energy efficient. In ESDSR we will solve those limitations.
3. ESDSR In ESDSR, the mobile nodes which are very likely to drain out of batteries are avoided in the route discovery phase of protocol. The energy level and the transmit power level of a node are taken into account while making routing decision. The ratios of current energy levels and the transmit power levels of nodes indicate how likely those nodes will deplete battery. In order to do that a source node finds a route R(t ) at time t such that the following cost function is maximized: C(R, t) = max (Rj (t)) Ei Rj (t) = min Pti
(1) (2)
where, Rj (t) is the minimum energy to transmit power ratio for the path j, Ei is remaining energy of node i on the discovered path and Pti is the transmit power of node i on the discovered path. Route discovery mechanism in ESDSR is illustrated in Figure 2. Node 1 is the source node and Node 5 is the destination node. Assume that all nodes have empty caches. The energy levels of the nodes and their current transmit power levels are shown in the Figure 2 at time t when the source initiated the route discovery. The source initiates a route discovery by broadcasting the route request packet. Node 2 and Node 4 are within the transmission range of Node 1. Since intermediate nodes 2 and 4 are not the destination, those two nodes add their own node ids in the request packet and rebroadcasts that route request packet. Once the destination node 5 receives the route request packet, it sends a reply to the source node 1 by reversing the path through which it receives the request packet. Let us assume that the destination Node 5 replies back to the source source node using the route 5 − 3 − 2 − 1. When the intermediate Node 3 receives the reply packet it estimates its ’expected life’ using PEtii and let us assume this value is 0.2. Node 3 records this value in that reply packet and forwards the reply packet to next hop which is Node 2. Node 2 estimates its ’expected life’ using the same formula. Let us assume that value is 0.1. It also reads the value recorded in the reply packet (which is 0.2). Node 2 replaces the ’expected life’ recorded in the reply packet because its ’expected life’ is less than that recorded in the reply packet. Thus the reply packet carry the value of the ’minimum expected life’ of the path 1 − 2 − 3 − 5 which is equal to 0.1. The source Node 1 records this path in the cache. Let us assume the source Node 1 discovers another
path 1−4−5 which has minimum ’expected value’ of 0.05. The source then selects the path 1 − 2 − 3 − 5 because that path has higher ’minimum expected life’ instead of choosing the shortest path 1 − 4 − 5. But DSR protocol always chooses path 1 − 4 − 5 because it is the shortest path. Once a source discovers paths and selects a path as mentioned above, it starts sending data packet using that route. Once a routing decision has been made, link by link power adjustment need to be accomplished in the following way. Every node records its transmit power in the data packet and sends that data packet to the next hop. When the next hop receives that data packet at power Precv , it reads the transmit power Ptx from the packet and recalculates the transmit power for the previous hop usingPmin = Ptx − Precv + Pthreshold
(3)
where all the values are in dbW. Pthreshold is the required threshold power of the receiving node for successful reception of the packet. The typical value of Pthreshold in LAN 802.11 is 3.652−10watt. To overcome the problem of unstable links due to channel fluctuations, a margin Pmargin in dBW is included in Equation 3 . Hence Equation 3 becomes Pmin = Ptx − Precv + Pthreshold + Pmargin
(4)
The recalculated transmit power is recorded in a power table. Each node maintains a power table which records the target node’s ID and the transmit power for that target node. The recalculated transmit power is recorded in the MAC packet (ACK packet). When the ACK packet is received by the transmitting node, it records the modified transmit power in the power table and transmits packet at that power. In our experiment we maintain a margin of 1dB. Usually a margin of 3dB has been maintained in [8]. Since in our case the transmit power is monitored packet by packet, we maintain a margin of 1dB. The advantages of our packet by packet monitoring is that if channel conditions changes during the packet transmission, the transmit power also changes accordingly. Since our protocol maintain a small margin, it can save more energy than the protocol mentioned in [8]. The power table is updated each time a node receives a packet from its neighbor. When a node has some packet to send to a node, it searches its power table to find the required transmit power for that node and transmit packet at that power. When a node can not find a record in the power table for a particular node (which will be the case when two nodes never exchanged packet before), it transmits at default power level which is 280mW for 250 meter range. In order to maintain the full functionality of route discovery and route maintenance of DSR, all routing packets(route re-
Typical values of K1 and K2 in 802.11 MAC environment at 2M bps bit rate are 4µs per bytes and 42µJ respectively. The variation of transmit power with respect to distance is depicted in Figure 3. The energy saving due to the controlled transmit power is shown in Figure 4. It is clear from that figure that almost 70 % of energy can be saved if we use controlled transmit power instead of fixed transmit power.
4. Simulation Model For simulations, we used ns-2 [3] with the CMU wireless extension. We used DSR version with flow state disabled. We also disabled the tap function. Since the receiving power is constant and a fixed amount of energy is dissipated when a node receives the packet, we set receiving power to zero. The medium access control (MAC) protocol was based on IEEE 802.11 with 2 Megabits per second raw capacity. The 802.11 distributed coordination function used Request-To-Send (RTS) and Clear-To-Send (CTS) control packets for unicast data transmission, and implemented a form of virtual carrier sensing and channel reservation to reduce the impact of hidden terminal problem. Data transmission was followed by an ACK. For radio propagation model, a two-ray path loss model was used. We did not consider fading in our simulation. The traffic sources were Constant Bit Rate(CBR) with 512 bytes per packet. We modified the node structure to include a power table. The structure of packet header was modified to carry data about the transmit power and the threshold power of a node. Route cache was modified to store additional information about the ’expected life’ of the nodes for a path. Routing decision logic of DSR was changed to energy aware decision logic. The energy model was modified to implement equation 5 for energy consumption while transmitting packet. A static scenario of 40 mobile nodes were randomly distributed initially in an area of 200m by 200m (a square area). The source and destination pairs were spread randomly over the network but the number of pairs were kept
0.25
Transmit Power (watt)
0.2
0.15
0.1
0.05 Fixed transmit power Variable transmit power 0 60
80
100
120
140 160 Distance(meter)
180
200
220
240
Figure 3. Transmit Power Variation with distance 70
60
50 % of Energy Saved
quest, route reply) are transmitted at default power. Moreover, in order to maintain normal operation of MAC layer, all MAC layer packets (RTS,CTS,ACK) are transmitted at that default power too. To conceive how much energy saving can be achieved by adjusting transmit power as mentioned above, let us consider a two nodes scenario where there are only a source node and a destination node. The source node adjusts its transmit power according to Equation 4. In order to observe the effect of power adjustment on energy consumption, we use a model mentioned in [8]. The energy consumption per data-packet of size D bytes over a given link can be modeled as(5) E(D, Pt ) = K1 Pt D + K2
40
30
20
10
0 60
80
100
120
140 160 Distance(meter)
180
200
220
240
Figure 4. Energy saving per packet in two node situation
constant during each scenario. Each CBR source started randomly at the beginning 0 to 10 seconds of the simulation and each simulation was run for 250 seconds. The number of connections was 20 in our case. We varied the area to 300m by 300m, 400m by 400m and 500m by 500m while keeping the number of mobile nodes and connections constant. We measured total energy consumptions of the nodes and the number of dead node at the end of simulation. An initial energy of 1.0 Joules were assigned to each node. To compare the performance of the proposed ESDSR protocol and the traditional DSR protocol, we focus on the following performance metrics: • Capacity: It is measured as total number of useful data packets reached the destination at the end of simulation.
11000
40 DSR ESDSR
35 10000
% of Energy saving per packet
30
packet
9000
8000
7000
25
20
15
10 6000
5 5000 40
60
80
100
120 140 160 180 Area (thousand sq. meter)
200
220
240
Figure 5. Total packet reached to destination nodes
0 40
60
80
100
120 140 160 180 Area (thousand sq. meter)
200
220
240
Figure 7. Percentage of energy saving per packet
2.5 DSR ESDSR
12 DSR ESDSR
2
10
8
dead node
mJ/packet
1.5
1
6
4 0.5
2
0 40
60
80
100
120 140 160 180 Area (thousand sq. meter)
200
220
240
Figure 6. Energy consumption per packet reached to destination
• Energy Consumption per packet: It is measured as the ratio of the total energy consumed for the network to the number of packets successfully reached the destination. • Number of dead nodes: It is measured as the number of nodes became out of battery battery at the end of the simulation.
5
Simulation results and Analysis
In this section, we present our simulation results and analysis of those results. The ESDSR and DSR protocols are tested in scenario with differing node concentration per unit area. With those tests, we want to compare capacity
0 40
60
80
100
120 140 160 180 Area (thousand sq. meter)
200
220
240
Figure 8. Number of dead node at the end of simulation
of the network (i.e. how many packets were successfully reached to destination), the energy consumption per packet, how much energy saved in ESDSR and how many node are ’dead’ at the end of simulation. In order to do that we simulated network scenarios consisting of 40 mobile nodes. Figure 5 shows the number of packets successfully reached the destination. In ESDSR, the packet reached the destination is higher than that of DSR. The reason is that some nodes were out of batteries during the simulation and hence they could not forward any more packet to a destination. They could not transmit their own packet either. But in ESDSR the nodes have longer lives. So the nodes were capable of transmitting their own packet and they were able to forward packets for longer period of time. The energy consumption per packet is depicted in Figure 6. From that figure
it can shown that energy consumption per packet is around 0.75mJ when the network is operating in an area of 200m by 200m using ESDSR protocol. But the energy consumption per packet is around 1.25mJ for that same network topology but using DSR protocol. The energy consumption per packet gets larger when we increased the network area. Since packets are traveling for more hops in larger network area, the energy consumption per packet increases. When the network area is 500m by 500m, the energy consumptions per packet in ESDSR and DSR are almost equal. The percentage of energy saving in ESDSR is depicted in Figure 7. In that figure it is shown that 37% of energy used in DSR can be saved in ESDSR protocol. The number of dead nodes at the end of simulation is shown in the Figure 8. The number of dead nodes is 1 in ESDSR compared to 5 dead nodes in DSR. The number of dead nodes are almost same when the network area is 500m by 500m. When the network area gets larger, the transmit power in ESDSR tends to be equal to that of DSR. Hence the number of dead nodes are almost same in ESDSR and DSR.
6. Conclusions In this paper, we proposed an energy saving routing protocol for mobile ad hoc network. The minimum energy routing protocol is designed and implemented by making changes in the minimum-hop fixed-transmit power version of DSR. Although network life is maximized using our protocol, including extra information in the packet header changes packet structure and increases packet size. Modifying the IEEE 802.11 MAC layer protocol will cost a change in the network card and firmware. Introducing transmit power control over the transmission medium may cause changes in the power circuits in the radio hardware. Since packets are not sent via minimum hop, the average number of hop will increase. Hence delay may be higher in ESDSR in compare to DSR. We are also investigating how such delay can be reduced in our protocol. The comparison of ESDSR with other energy aware routing protocol needs to be done in future to fairly judge the performance of ESDSR.
Acknowledgement This work was partly supported by Natural Sciences and Engineering Research Council (NSERC) of Canada.
References [1] C.E. Perkins, “Ad Hoc On Demand Distance Vector (AODV) routing:, IETF Internet-Draft, draft-ietfmanet-aodv-00.txt, November 1997.
[2] J. Broch, D. B. Johnson, and D. A. Maltz, “The Dynamic Source Routing Protocol for Mobile Ad Hoc Networks”, IETF Internet-Draft, draft-ietf-manet-dsr00.txt, March 1998 [3] K. Fall and K. Varadhan, “Ns Notes and Documentation Technical Report”, University of California Berkeley, LBL, USC/ISI, and Xeron PARC, 2003. [4] S. Ivan and X. Lin, “Power-aware localized routing in wireless networks.” , Proc. of the 14th International Symposium on Parallel and Distributed Processing, May 1-5, 2000. pp.371 [5] S. Singh, M. Woo, C.S. Raghavendra,” Power-aware Routing in Mobile Ad Hoc Networks”, , Proc. of the Fourth Annual ACM/IEEE International Conference on Mobile Computing and Networking,October, 1998, pp.181-190 [6] C.K. Toh, “ Maximum Battery Life Routing to Support Ubiquitous Mobile Computing in Wireless Ad Hoc Networks”, IEEE Communication Magazine, June 2001,pp.138-147 [7] Woo K, Yu C, Youn HY, Lee B.” Non-Blocking Localized Routing Algorithm for Balanced Energy Consumption in Mobile Ad Hoc Networks”, Proc. of International Symp. on Modeling, Analysis and Simulation of Computer and Telecommunication Systems(MASCOTS 2001), pp 117-124 [8] Sheetalkumar Doshi, Timothy X Brown ,” An ondemand minimum energy routing protocol for a Wireless Ad Hoc Network”, Proc. of ACM SIGMOBILE Mobile Computing and Communication REview. Vol. 6, Issue.3 , July 2002, pp. 50-66
