Performance comparison of BER-based routing protocols under

36th Annual IEEE Conference on Local Computer Networks

LCN 2011, Bonn

Performance comparison of BER-based routing protocols under realistic conditions Tiguiane Yélémou∗‡ , Jonathan Ledy∗§ , Benoˆıt Hilt§ , Anne-Marie Poussard∗ , Philippe Meseure∗ ∗ University

of Poitiers, XLIM-SIC CNRS Lab, France University of Bobo Dioulasso, Burkina Faso § University of Haute Alsace, MIPS Lab, France (contact: [email protected])

‡ Polytechnic

Abstract—To face QoS constraints of multimedia applications, many researches focus on improving the standard MANET routing protocols. Nevertheless, only few works relating to the comparison of enhanced heuristics are available. In those literature, the comparison parameters are measured through simulations performed mainly under unrealistic conditions. In this paper, we compare BER-based enhancements of AODV and OLSR when they have to deal with mobility and multicommunication. We point out that link quality, influenced by obstacles in the propagation field, should be taken into account in MANET routing protocols enhancement simulations. Index Terms—Wireless networks, routing algorithm, BERbased, an enhanced NS2.

allows to take into account the channel quality when setting up the routes between the wireless nodes. The strengths and suitability for deployment in different scenarios of these QoS protocols are pointed out in this article. The rest of the paper is organized as follows: In section II, we present and analyze previous works, in section III, we present our BER-based approach to enhance AODV and OLSR. Simulation results will be studied in section IV. Then we will conclude with the privileged fields of application of these QoS protocols.

I. I NTRODUCTION

Many researches compare MANET standard routing protocols: AODV, DSR, DSDV, OLSR, etc. Major trends in the choice of reactive or proactive protocols as routing protocol in MANET are known. Most papers [4][5] show that reactive protocols perform better than proactive ones for main MANETs performance metrics. However most studies are based on simulations conducted under unrealistic conditions. Very often, obstacles in the propagation field are not taken into account in the radio propagation model. This leads to biased results due to the fact that environment interactions affect significantly link quality [6][7]. In mobility situations, unrealistic mobility models with constant speed are very often used and interactions between mobile entities are not taken into account. Authors of [5][8][9] show that mobility model may drastically affect protocol performances. When Haerri et al. [2] test OLSR and AODV in an urban environment under realistic node mobility model [1], they show that OLSR globally outperforms AODV, contrary to most opinions. Multi-communication effects must also be evaluated in a convenient way. Interference is an inherent property of wireless networks which affects routing protocol performance as well as network efficiency. Interferences due to other communications, leading to a saturation of access links, impact the probability of successful packet reception in wireless networks. Many authors [10][11] show that MANET performance degradate when the number of nodes increases because each node has to share the radio channel with its neighbors. Recent years have seen proposals for heuristics addressing the issue of QoS in MANET routing protocols [12][13][14].

The need to communicate anytime and anywhere, the miniaturization of communication equipments and implementation of multimedia applications on mobile devices, have generated a renewed interest in research on QoS in Mobile Adhoc NETworks (MANET). Due to their low cost, ease of deployment and agility, MANETs give new perspectives for the future of telecommunications. However, due to their intrinsic properties such as the drawbacks generated by different simultaneous transmissions, the effect of mobility, the limited bandwidth, these networks fail to provide a quality of service capable of fulfilling the strict requirements of multimedia applications. Considering this challenge, routing protocols play a crucial role, especially when one must take into consideration the quality of radio links and the effects of mobility and multicommunication. After intense researches to adapt the protocols used in wired networks to wireless networks, it is now time to improve their performance by taking into account their surrounding characteristics. Research efforts have not much focused on evaluating these QoS protocols and compare their performance in realistic conditions. Very often the effect of mobility and interference remains either untested or simulated with unrealistic propagation and mobility models. To overcome this, we use both a realistic mobility model taking into account the interaction of the moving object with surrounding obstacles and with other moving objects (VanetMobiSim) [1][2] and a propagation simulator taking into account obstacles in the propagation medium (Communication Ray Tracer) [3]. Under these conditions, we test and analyze performances of BER-based approach of QoS enhanced AODV and OLSR protocols.This enhancement

978-1-61284-928-7/11/$26.00 ©2011 IEEE

II. BACKGROUND AND RELATED WORK

259

But very little research has focused on the comparison of these QoS-protocols under realistic conditions. III. BER- BASED ROUTING PROTOCOLS To study and compare BER-based approach on on-demand and table driven MANET routing protocols, we focus on two standard ones, OLSR and AODV protocols. Theses two protocols do not basically take into account QoS metrics. Indeed with AODV protocol, route discovery is initiated when a source requires it. An intermediate node may respond if it knows a more recent path to the destination, otherwise the destination respond to the first route request. This means that the lower number of hops remains, the lower delay we get if realistic propagation model is not used. Yet, if link quality is bad, a node may repeat its transmission several times. This increases delay but not the number of hops. In our QoS approach of AODV [15], any node that receives a route request message, does not forward it if the BER of its link with the transmitter neighbor is over a threshold value. Indeed, if this node retransmit this request, it contributes to establish a bad path which may cause high packet loss and high delay due to possible several retransmission attempts. With OLSR, the routing computation mechanism is also based on the lowest number of hops. The Multi Point Relays (MPR) computation mechanism is heavily based on the number of 2-hop-neighbors that MPR candidates can reach. This MPR mechanism can lead to bad network capacity, since a node have a partial knowledge of the network [16]. This can result in low PDR (Packet Delivery Ratio) and important end-to-end delay due to retransmissions. To correct this, we modify the selection criteria and base it on the quality of the links. This avoids data travels on bad paths (in term of BER). The basic idea in OLSR with BER consideration consists in selecting the path which offers the best BER during both the choice of MPR nodes and the path computation in routing tables [16]. Respecting the overall structure of the original MPR selection, our BER-based MPR selection mechanism mainly consists in considering that a 2hop-neighbor is not covered until the 1hop-neighbor which provides the best route in term of BER is found. For routing table calculation, a variant of Djisktra algorithm where shortest path is primarily based on BER metric is therefore used [16]. IV. P ERFORMANCE EVALUATION A. Experimental setup Most research evaluations rely on simulation to show the effectiveness of new QoS approaches for protocol enhancements. Most of the time, they do not take into account any environment parameters when modeling the propagation channel. Furthermore, other effects such as multiple paths induced by the environment are not taken into account although they highly influence the quality of the received signal. If the environment is not considered, the obtained results may be biased and rather optimistic. The influence of bad links is thus highly underestimated. To compute more real simulations, we must use a realistic model of wave propagation

taking into account the environment. Therefore, we enhanced NS2 with a raytracer simulator which has been developed at the XLIM-SIC laboratory [3]. Our BER-based protocols directly rely on the BER value computed by this improved NS2. Concerning mobility, VanetMobiSim [1] tool has been used to generate realistic nodes mobility. This model takes into account interaction of mobile nodes with surrounding obstacles and with other mobile nodes. CBR 512 bytes data packets size are sent on a 24 Mbps rate channel. The MAC layer used is 802.11a. Three important performance metrics for MANET are measured: packet delivery ratio (PDR), average end-to-end delay of data packets and routing overhead. B. Simulation results 1) Multi-communication situations: Many works have focused on interference effects on wireless network efficiency. It is shown that interferences directly affect routing protocols performance [10][11]. To highlight the impact of multicommunication on our BER-based protocols, in fixed nodes scenario, we increase the number of simultaneous sourcedestination packets transmission from 3 to 20, and measure the effects on performance metrics. The number of nodes in the area is fixed to 40. Fig. 1 shows that, for all the fourth protocols, PDR values decrease when the the number of simultaneous data connections increases. For this metric, standard and BER-AODV are better than BER-OLSR for small number of simultaneous transmissions. But when this number increases, this margin progressively decreases until BER-OLSR outperforms both AODV protocols. This reversed tendency is observed from 5 data connections for standard AODV and 10 data connections for BER-AODV (Fig. 1). According to our trace files, this situation is explained by the congestion generated by AODV in nodes concerned by the communications. Concerning control packets, Fig. 2 shows that their number dramatically increases according to the number of sourcedestination connections for standard AODV. This standard ondemand routing protocol emits from 2200 control packets for 3 source-destination connections up to 8000 for 20 sourcedestination connections, whereas for OLSR protocols, the number of control messages remains constant around 2200 messages. Additional connections have no effect on OLSR overhead since control packets exchanges are periodical. Notice that BER-AODV outperforms standard AODV in routing overhead. This is due to the non-participation of certain nodes to the packets broadcasting when BER-AODV is used: these nodes have bad links (in terms of BER) with the sender of the message to forward. Secondly, these bad links are more deteriorated in multi-communication situations and lead to several repetitions of route request and route error processes due to broken links. For its part, the end to end delay increases with the number of simultaneous source-destination transmissions for all protocols but BER-based OLSR is better than other protocols. Routes are already established for OLSR. BER-OLSR permits

260

to outperform the standard one thanks to its lower number of retransmissions. BER-AODV has the worst delay. Due to bad link reject strategy, BER-AODV produces longer path that increases the delay.

Fig. 3. Evolution of delay according to the number of simultaneous transmissions.

Fig. 1. Evolution of PDR according to the number of simultaneous transmissions.

for PDR (see Fig. 4) and overhead (see Fig. 6) are, this time, not significant. This is explained by the fact that in the mobility situation, MPR selection mechanism does not perform well as a result of rapid change in node neighborhood. Also, the delay variation between standard OLSR and BER-based one never goes beyond 3ms (Fig. 5). These performance (see Fig. 4 and Fig. 5) allow us to conclude that QoS approach for routing protocol enhancement is not so effective as widely shown in realistic urban mobility situations. In other words, it is maybe not worth trying to select paths with better links since the instability of links is the real issue of mobility.

Fig. 2. Evolution of overhead according to the number of simultaneous transmissions.

2) Mobility situations: In this section we analyze the impact of mobility on these protocols. Node speed increases from 4m/s to 20m/s. The number of nodes remains fixed to 40. Fig. 4, Fig. 5 and Fig. 6 show the same performance measures as in previous section when speed increases. Fig. 4 shows that, globally, speed has not an important effect on the PDR. For every protocol, the difference of PDR value never goes beyond 8 points. BER-AODV and BER-OLSR have approximately the same trend in terms of PDR. Between standard OLSR and BER-OLSR we observe that the difference

Fig. 4.

Evolution of PDR according to nodes speed.

V. C ONCLUSION AND PROSPECTS We modified AODV and OLSR protocols to take into account the quality of radio links in routing path computation.

261

From these results, we conclude that OLSR protocols (standard and BER enhanced one) are suitable for real-time applications and may be preferred to AODV ones in dense networks with a lot of source-destination connections. R EFERENCES

Fig. 5.

Fig. 6.

Evolution of delay according to nodes speed.

Evolution of routing overhead according to nodes speed.

These new processes should allow the choice of better routes for data transmission. As QoS metric, we use the bit error rate (BER). We conducted simulations on the standard AODV and OLSR and the BER-based ones under realistic radio propagation and mobility models. Multi-communication and mobility impact are studied for these protocols. The results show that BER-based AODV and BER-based OLSR are better than standard ones for the PDR. BER-based OLSR always induces better end-to-end delay than BER-based AODV, then seems suitable for situations with delay and jitter constraint. In high load context, BER-based OLSR performs better than BER-based AODV for PDR. In mobility situations, the four protocols have nearly the same performance for PDR metric. We also point out that in an urban environment with realistic mobility and propagation conditions, the impact of the speed is limited when dealing with multimedia communications.

[1] J. Haerri, F. Filali, and C. Bonnet, “Mobility models for vehicular ad hoc networks : a survey and taxonomy,” in IEEE communications surveys and tutorials, vol. 11, pp. 19–41, IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC, 2009. [2] J. Haerri, F. Filali, and C. Bonnet, “Performance comparison of aodv and olsr in vanets urban environments under realistic mobility patterns,” in 5th IFIP Mediterranean Ad-Hoc Networking Workshop, (Lipari (Italy)), June 2007. [3] R. Delahaye, Simulation Réaliste et Efficace de la Couche Physique pour l’Aide au Routage des Réseaux Ad hoc. PhD thesis, University of Poitiers, 2007. [4] R. K. Jha, S. V. Limkar, and U. D. Dalal, “A performance comparison of routing protocols for security issue in wireless mobile ad hoc networks,” in International Conference on Emerging Trends in Engineering & Technology, pp. 825–830, 2010. [5] A. Rahman, S. Azad, F. Anwar, and A. H. Abdalla, “A performance comparison of proactive and reactive routing protocols of mobile adhoc network (MANET),” Journal of Engineering and Applied Sciences, vol. 2, pp. 891–897, 2007. [6] W. Hamidouche, R. Vauzelle, C. Olivier, Y. Pousset, and C. Perrine, “Impact of realistic mimo physical layer on video transmission over mobile ad hoc network,” in 20th Personal, Indoor and Mobile Radio Communications Symposium, 2009. [7] M. Guenes, M. Wenig, and A. Zimmermann, “Realistic mobility and propagation framework for manet simulations,” in NETWORKING 2007: Ad Hoc and Sensor Networks, Wireless Networks, Next Generation Internet, Proceedings, vol. 4479 of Lecture Notes in Computer Science, pp. 97–107, Springer Verlag, 2007. [8] S. Hamma, E. Cizeron, H. Issaka, and J.-P. Guedon, “Performance evaluation of reactive and proactive routing protocol in ieee 802.11 ad hoc network,” in Next-Generation Communication and Sensor Networks 2006, vol. 6387, pp. 38709–38709, 2006. [9] G. Marfia, G. Pau, E. De Sena, E. Giordano, and M. Gerla, “Evaluating vehicle network strategies for downtown portland: Opportunistic infrastructure and the importance of realistic mobility models,” in First International MOBISYS workshop on Mobile Opportunistic Networking, pp. 47–51, 2007. [10] K. Jain, J. Padhye, V. Padmanabhan, and L. Qiu, “The impact of interference on multi-hop wireless network performance,” in 2nd ACM MOBICOM 2003 Conference, San Diego, CA, SEPT 19, 2003, vol. 11, pp. 471–487, 2005. [11] R. Hekmat and P. Van Mieghem, “Interference in wireless multi-hop adhoc networks and its effect on network capacity,” Journal on Wireless Networks (Proceedings of Conference on Ad hoc networking), vol. 10, pp. 389–399, Aug. 2004. [12] N. Enneya, K. Oudidi, and M. Elkoutbi, “Enhancing delay in MANET using OLSR protocol,” Int’l J. of Communications, Network and System Sciences, vol. 5, pp. 392–399, 2009. [13] P. Esposito, M. Campista, I. Moraes, L. Costa, O. Duarte, and M. Rubinstein, “Implementing the expected transmission time metric for OLSR wireless mesh networks,” in Wireless Days 2008 (WD08), pp. 1–5, Nov. 2008. [14] F. Ingelrest and D. Simplot-Ryl, “Maximizing the delivery of MPR broadcasting under realistic physical layer assumptions,” Journal of computer science and technology, May 2008. [15] J. Ledy, H. Boeglen, B. Hilt, A. Abouaissa, and R. Vauzelle, “An enhanced AODV protocol for VANETs with realistic radio propagation model validation,” in 9th International Conference on Intelligent Transport Systems Telecommunications, pp. 398–402, 2009. [16] T. Yélémou, P. Meseure, and A.-M. Poussard, “A new ber-based approach to improve olsr protocol,” in Eighth IEEE and IFIP International Conference on wireless and Optical communications Networks Paris France, May 2011.

262