Factors Affecting the Performance of Ad Hoc Networks

Transcription

1 Factors Affecting the Performance of Ad Hoc Networks Dmitri D. Perkins, Herman D. Hughes, and Charles B. Owen Department of Computer Science and Engineering Michigan State University East Lansing, MI 88-6 Abstract- Mobile Ad Hoc NETworks (MANETs) are an emerging class of network architectures that are characterized by their highly dynamic topology, limited resources (i.e., bandwidth and power), and lack of fixed infrastructure.the primary motivation for such networks is increased flexibility and mobility.random node mobility along with various other factors such as network size and traffic intensity may be very dynamic, resulting in unpredictable variations in the overall network performance.this study centers on investigating and quantifying the effects of various factors and their two-way interactions on the overall performance of ad hoc networks. This study will contribute to the modeling and development of adaptive ad hoc protocols (routing, medium access control, scheduling and buffer management).using k r factorial experimental design, we isolate and quantify the effects of five factors: node speed, pause-time, network size, number of traffic sources, and type of routing (source versus distributed), that affect the performance of ad hoc networks.specifically, this paper evaluates the impact of these factors on the following performance metrics: throughput, average routing overhead, and power consumption. I. INTRODUCTION AND MOTIVATION A Mobile Ad Hoc NETworks (MANET) [] is an selforganizing system of mobile routers (and associated hosts) connected by wireless links. Ad hoc networks may operate autonomously, or may be connected to the larger Internet. The goal of mobile ad hoc networking is to provide a rapidly deployable means of communication (and computing), independent of a pre-existing infrastructure (e.g., base stations). Such networks will utilize a wireless physical layer consisting of relatively low bandwidth, time-varying links. In current wireless networks, the wireless mobile node is never more than one hop from a base station that can route data across the communication infrastructure. In mobile ad hoc networks, there are no base stations and because of a limited transmission range, multiple hops may be required for nodes to communicate across the ad hoc network. Routing functionality is incorporated into each host. Thus, MANETs can be characterized as having a dynamic, multi-hop and, constantly changing topology. While mobile ad hoc networks can be used without a fixed infrastructure, their use is also being considered as part of the vision for a truly ubiquitous communications environment (e.g., Wireless Internet). The future success of ad hoc networking will depend on its ability to support existing and future Internet applications and protocols. Such adynamic environment poses tremendous protocol design challenges at every layer of the network architecture, ranging from physical layer issues to distributed medium access control to routing. Several factors will affect the overall performance of any protocol operating in an ad hoc network. For example, node mobility may cause link failures, which will negatively impact routing and quality-of-service support. Network size, control overhead, and traffic intensity will have a considerable impact on network scalability. These factors along with inherent characteristics of ad hoc networks may result in unpredictable variations in the overall network performance. The primary objective of this study is to evaluate and quantify the effects of various factors (and their two-way interactions) that may influence network performance. While there have been performance evaluations of ad hoc networks [, ], none have actually quantified the effects of the influential factors. Using a factorial experimental design [6, 7], we determine the impact of five factors: () node speed, () node pause time, () network size, () number of traffic sources and () routing protocol (source vs. distributed) on the performance of ad hoc networks. We examine the impact of these five factors on three performance metrics: () average throughput, () average routing overhead, a nd () power consumption. Quantifying the effects of these factors will help guide the design choices and tradeoffs. For example, suppose node mobility is shown to have a greater impact on average control overhead than any other factor. This would suggest that designing algorithms that adapt to node mobility would have the greatest impact on network performance. The factorial design model used in this work is based on a linear regression model and thus, makes the assumption that the effects of the factors are additive. Further, our model assumes that experimental errors are independent and normally distributed. The remainder of this paper is organized as follows. Section II describes the simulation environment and methodology. A discussion of the performance metrics and experimental factors is presented in Section III. In Section IV, the simulation results and design analysis are presented, followed by asummary in Section VI. II. METHODOLOGY, SIMULATION AND EXPERIMENTAL DESIGN Again, the goal of our experiments is to examine and quantify the affects of various factors and their interactions on the overall performance of ad hoc networks. To achieve this goal, //$7. IEEE 8

2 we used a k r (k=, r=) factorial design methodology [6], thus conducting separate experiments referred to as design points. Each experiment was replicated times, resulting in 8 simulation runs. This section provides a brief discussion of the simulation environment and the factorial experimental design. Our simulation study was conducted using the librarybased Global Mobile System Simulator (GloMoSim) [8] for sequential and parallel simulation of wireless networks. It was designed using the parallel discrete-event simulation capability provided by Parsec [9], a C-based parallel simulation language. GloMoSim was developed based on a layered approach similar to the OSI seven layer network architecture. Our model is simulated for seconds of simulated time. The radio transmission range of each node is approximately meters and the channel capacity is Mbits/sec. The free space propagation model [8] was used to determine if a node is reachable. The free space model predicts received signal strength when the transmitter and receiver have a clear, unobstructed line-of-sight path between them. Received power decays as a function of the T-R separation distance. The IEEE 8. Medium Access Control Protocol was used as the MAC protocol. In this study, we used constant bit rate CBR sources that continuously transmit -byte data packets at a rate of packets per second for the duration of the simulation. As it is our goal to investigate the impact of routing in ad hoc networks, we utilize two different routing protocols: Dynamic Source Routing (DSR) [] and Ad-hoc On demand Distance Vector (AODV) []. Both routing protocols are ondemand protocols and, thus, do not transmit periodic routing messages. These protocols differ with regard to the method by which routes are computed. DSR uses source routing in order to deliver packets to any destination in a mobile ad hoc network. Source routing requires that the headers of all data packets carry an ordered list of nodes through which the packet must traverse. AODV uses a distributed (e.g., hop-byhop) technique to deliver packets and uses sequence numbers (e.g., to avoid routing loops) for each route entry. The random waypoint mobility model [8] is used in our evaluations. In the random waypoint model, each node is placed randomly in the simulated area (6Xm ). After remaining at the location for a specified pause time, the node randomly selects another destination from the physical terrain. The node then moves to the new location at a speed uniformly chosen between aminimum and maximum speed (meters/sec). After reaching the destination, the node stays there for amobility-wp-pause time period. III. PERFORMANCE METRICS AND EXPERIMENTAL FACTORS To maintain consistency with factorial design terminology, we will refer to the variables that affect the outcome of an experiment as factors and the actual outcomes as performance responses or metrics. The following performance metrics are examined in this study []:. Throughput: throughput measures the effectiveness of the network in delivering data packets. That is, how well does the network deliver packets from the source to the destination?. Average routing overhead: the average number of control packets produced per node. Control packets include route requests, replies and error messages.. Average power consumption: measures the average power consumptions per node, as energy is a limited resource in ad hoc networks. Table shows the levels for each of the factors examined in this study. Each factor is examined at two different factor levels. We also examine the effects of the two-way interactions. That is, we want to determine whether the effect of one factor is dependent on the level of another. It should be noted that there are certainly many other factors (i.e., transmission range, MAC protocol, link bandwidth, size of roaming area, etc.) that may have an effect on network performance. The following factors were chosen as a starting point for this investigation. The main effect of a factor is the average change in the throughput due to changing the factor from its - level to its + level, while holding all other factors fixed. This average is taken over all possible combinations of the other (k-) factors. The two-way interaction effect is the difference between the average throughput when two factors are at the same level and the average throughput when they are at opposite levels. Table Factors Examined Label Factor Level (-) Level (+) Speed (m/s) Pause-time (sec) Network size 8 # of sources Routing Source Distributed IV. SIMULATION RESULTS ANDDESIGN ANALYSIS This section presents the main effects and two-way interaction for each factor. For brevity and convenience, each factor is denoted by its label (see Table ) and each two-way interaction, say and, by(). Before examining the main effects and their two factor interactions, we first observed the average performance results of the three performance metrics. Figures - show the performance metrics averaged over four replications at each design. By closely examining the following graphs and the experimental design matrix, we can make three key observations: For control overhead, there are fluctuations at each design point (one "high" followed by one "lower" value). 9

3 Examining the experimental design matrix, we observe that this pattern follows the level changes of factor, node speed. The fluctuations become more pronounced for higher design points (i.e., 6-) suggesting astrong interaction of node speed, routing and, potentially, network size. For throughput, the second and fourth blocks of eight design points experience significant fluctuations (one "high" followed by one "low" values), indicating a factor interaction. Looking at the design matrix, we observe that these fluctuations correspond to level changes of multiple factors, namely factors one (node speed), and five (routing protocol), and, possibly, factor three (network size). For power consumption, the design points are consistently grouped in blocks of four. There are minor fluctuations (one "high" followed by one "low" values) within each block, again suggesting that node speed has significant influence on power consumption. The consistent groupings of four suggest some interaction among factors. Figures - and the above observations are certainly beneficial to understanding the main effects of each factor as well as their two-way interactions, but such observations are only qualitative. To confirm the results, we must now quantify the effects. Next, we formally substantiate the observations discussed above. Figures -6 show the 9% confidence intervals for the expected main effects and twoway interaction on control overhead, throughput and power consumption, respectively. For control overhead and throughput, the graphs show that the main effects of factors,,, and are real (significantly different from zero) and thus impact network performance. The results also show that the main effect of factor two (pause-time) does not significantly influence either performance metric. Moreover, the two-way interactions and the main effects of factors and have no real impact on power consumption. Additional observations and discussion are presented for each metric in the following sections. A. The Effects on Control Overhead As shown in Figure, the effects of factors,, and result in a significant increase of control data. Interestingly, this result suggests that using distributed routing gives rise to an increase in control overhead. Finally, we see that the amount of control overhead is negatively influenced by several twoway interactions: (x), (x), and (x). Thus, the previous observations regarding control overhead are valid. Average Packet Overhead Control Overhead Design Points Figure. Average control overhead for each experiment. Average Throughput (bps) Figure. Average throughput for each experiment. Average Power Consumption Average Throughput Design Point Power Consumption Design Point Figure. Average power consumption for each experiment.

4 Effects on Cntrl data (pkts/router) - - Effects on Control Overhead x x x x x x x x Effect Labels Figure. 9% confidence intervals for the expected main effects and two-way interaction effects on Control Overhead (packets/router). Effect on Throughput (kps) Effects on Throughput x x x x x x x x Effect labels Figure. 9% confidence intervals for the expected main effects and two-way interaction effects on average throughput in kbps. Effect on Power (mwhr) Effects of Power Consumption x x x x x x x x Effect Labels Figure 6. 9% confidence intervals for main effects and two-way interaction effects on power consumption in (mwhr). B. The Effects on Average Throughput The effects of factors and (and their two-way interactions) are significantly negative for throughput. The effect of factor, network size, is positive, indicating that larger network size results in improved throughput. Notice, that the type of routing used has a negligible effect on throughput performance. Combining this fact with the results (using distributed routing results in an increase in control overhead) from the Subsection IV.A suggest that distributed routing is less efficient than source routing. Thus, source routing tends to be a"better" design choice if efficiency is the primary concern. C. The Effects on Power Consumption Power consumption is strongly influenced by two factors. The effect of factor, network size, is positive, decreasing the average power consumption as the network size is increased. The effect of factor, number of sources, is negative, increasing the average power consumption as the number of sources increases Intuitively, this is reasonable since increasing the network size while maintaining a constant traffic load essentially increases the number of routers eligible to forward packets, resulting in a reduced load. On the other hand, increasing the number of traffic sources simply increases the routing load of each mobile host, resulting in increased power consumption. D. Quantifying the Effects: Allocation of Variation The importance of a factor can be determined by the proportion of variation in the performance metric that is explained by the factor. The proportions of variation explained by each factor and the two-way interactions are shown in Table. The last row, labeled EE, is the proportion of variation attributed to experimental error. Notice that main effect of factor, the number of sources, is responsible for 69 percent, percent, and percent of the variation in power consumption, control over, and throughput, respectively. Based on these results and analysis, it appears that factor is the most influential factor, followed by node speed and network size. The two-way interaction of node speed and the number of traffic sources (x) is also significant. Table Allocation of Variation for each Performance response Percentage of Variation Explained by Factor Power Consumption Control Overhead Factor Throughput x x...7 x.8.. x x..6.7 x x..9.7 x.7..8 EE

5 V. SUMMARY This paper has presented a comprehensive analysis of five factors: node speed, pause-time, network size, number of traffic sources, and type of routing (source versus distributed), that affect the routing performance of ad hoc networks. A factorial experimental design was used to isolate and quantify the main effects as well and two-way interactions of these factors on three performance responses: throughput, average routing overhead, and power consumption. The main results and observations of our analysis are as follows: For the experimental design used in this study, source routing was more efficient. That is, it achieved approximately the same performance as distributed routing but used less control overhead. As the result show, the main effect of factor, number of traffic sources, has the strongest impact on the performance responses followed closely by node speed and network size. Increasing the network size, while maintaining the traffic load results in increased throughput, decreased control overhead and decreased power consumption. Increasing traffic load and increasing the number of traffic sources may not result in the same performance results. For example, adding more traffic sources will certainly increase the control overhead, while increasing the transmission rate at a single source does not necessarily result in increased control overhead. It is important to note that, while we were very sensitive to the selection of factor levels, our results and conclusions (e.g., estimates of effects and interactions) are based upon the factor levels used in this design and may vary if different factor levels are used. To reduce the potential variations when different factor levels are used, we executed several simulations and selected the factors most appropriate. That is, we took special care not to select factors levels too far apart to provide any meaningful or useful results. Thus, we believe the results of this study will greatly contribute to the modeling and design of adaptive protocols for ad hoc networks. [] J. Broch, J, D. B. Johnson, a nd D. A. Ma ltz, The Dynamic Source Routing protocol for mobile ad hoc networks. Internet Draft, draft-ietf-manet-dsr-.txt, October 999. Work in progress. [] C. E. Perkins, Ad Hoc On Demand Distance Vector Routing. Internet Draft, draft-ietf-manet-aodv-.txt, November 998. Work in progress. [] J.Broch,D.A.Maltz,D.B.Johnson,Y.C.Hu,andJ. Jetcheva. "A performance comparison of multi-hop wireless ad hoc network routing protocols". Proceedings of the ACM/IEEE MOBICOM '98, Dallas, Texas, October 998. [] D. D. Perkins and H. Hughes, A performance comparison of routing protocols for mobile ad hoc networks. Proceedings of SPECTS, Vancouver, B.C. Canada, July. [6] R. Jain, 99. The Art of Computer System Performance Analysis. New York: John Wiley & Sons Inc. [7] A. M. Law and W. David Kelton,. Simulation, Modeling, and Analysis. New York: McGraw-Hill Higher Education. [8] L.Bajaj,M.Takai,R.Ahuja,K.Tang,R.Bagrodia,and M. Gerla, "GloMoSim: a scalable network simulation environment". UCLA Computer Science Department, Technical Report 997, May 999. [9] R. Bagrodia, R. Meyer and M. Takai, Parsec: a parallel simulation environment for complex System. IEEE Computer, Vol. no., October 998, pp [] M. S. Corson. and J. Macker, Mobile ad hoc networking: routing protocol performance issues and evaluation considerations. Internet RFC, January 999, REFERENCES [] M. S. Corson, J. P. Macker, and G. H. Cirincione, Internet-based mobile ad hoc networking. IEEE Internet Computing, Vol., No., July/August 999, pp. 6-7.