SCALABLE MODEL FOR THE SIMULATION OF OLSR AND FAST-OLSR PROTOCOLS

SCALABLE MODEL FOR THE SIMULATION OF OLSR AND FAST-OLSR PROTOCOLS 1 Hakim Badis 1,2 ad Khaldou Al Agha 1,2 1 LRI Laboratory, Uiversity of Paris XI, Orsay, Frace 2 INRIA Laboratory, Rocquecourt, Frace {badis,alagha}@lri.fr ABSTRACT Techological advaces ad rapid developmet of the IEEE 2.11 stadard have facilitated the growth of wireless local area etworks (WLAN) ad mobile computig i public domais. So, the routig protocols must support a very large ad fast mobility of odes over a very large ad-hoc etwork. I this paper, we preset a scalable simulatio model ad results for the Optimized Lik State protocol (OLSR) ad the Fast-OLSR extesio. OLSR is a proactive protocol, thus it periodically seds cotrol packets to build ad update the topology. Fast-OLSR extesio is desiged to meet the eed for fast mobility i Mobile Ad-hoc NETworks (MANETs). The aim of this article is to evaluate the performace of Fast-OLSR i a very large ad-hoc etworks by applyig a extesible simulatio model close to a real ad-hoc etwork. The simulatio results were obtaied with a IEEE 2.11 medium access cotrol ad physical layer model. Results show that the loss rate ca be miimized for the Adhoc etwork of odes with fast mobility by implemetig OLSR ad Fast-OLSR protocols. I. INTRODUCTION Growig iterest has bee give to the area of Mobile ad-hoc etworkig sice the apparitio of powerful radio devices allowig the coectio of mobile odes. A mobile ad-hoc etwork (MANET) [1] is a collectio of odes, which are able to coect o a wireless medium formig a arbitrary ad dyamic etwork with wireless liks. Implicit i this defiitio of a etwork is the fact that liks, due to ode mobility ad other factors, may appear ad disappear at ay time. This i a MANET implies that the topology may be dyamic ad that routig of traffic through a multi-hop path is ecessary if all odes are to be able to commuicate. A key issue i MANET is the ecessity that the routig protocols have to respod rapidly to topological chages i the etworks. At the same time, due to the limited badwidth available through mobile radio iterfaces, it is imperative that the amout of cotrol traffic, geerated by the routig protocols is kept at a miimum. Differet routig protocols are proposed i the MANET divided ito the followig categories: proactive, ractive ad hybride protocols. OLSR [2, 3] is a proactive, lik-state routig protocol, employig periodic message exchage to update topological iformatio i each ode i the etwork. Topological iformatio is flooded to all odes, providig routes immediately available whe eeded. However, whe a ode is movig fast, its eighbors are ot stable ad chage quickly. The OLSR protocol caot detect the broke liks quickly. So, the packets trasmitted o a ivalid lik are lost. I order to miimize packet loss, a extesio of the Optimized Lik State Routig protocol (OLSR), deoted Fast-OLSR [4] is proposed. The odes i a movig fast mode activate the fast movig mode by applyig Fast- OLSR protocol. I this paper, we evaluate the performace of a large Ad-hoc etwork of odes implemetig OLSR ad Fast-OLSR protocols. This paper is orgaized as follows: i sectio II, we provide a brief overview of the IEEE 2.11 access techique ad briefly describe the simulatio model. I sectio III, we describe the three families of routig protocols (i.e., reactive, proactive ad hybrid protocols) discussed i the MANET workig group. I sectio IV, we briefly preset OLSR, a proactive protocol suitable for dese ad large mobile ad hoc etworks. I sectio V, we show the Fast- OLSR extesio that takes ito accout fast mobility. I sectio VI, we evaluate the performace of the Fast-OLSR extesio o the basis of the simulatio results. We preset the simulatio model that we have developed to represet a mobile etwork. II. THE IEEE 2.11 STANDARD AND THE SIMULATION MODEL A. IEEE 2.11 physical layer We use the IEEE 2.11 [5] direct sequece (DS) system. The physical layer ca offer a throughput of 1 or 2 Mbps, which takes ito accout the exact protocol overhead. We have used the followig assumptio: the broadcast packets ad the poit to poit packets are set at 1 Mbps. Our simulatio model takes ito accout the exact overhead caused by the physical layer of IEEE 2.11 stadard. For further detail refer to [5,6]. B. The IEEE 2.11 MAC scheme With radio sigals, it is ot possible to directly detect collisios i a radio etwork. Ideed, it is ot possible to liste to alie trasmissio while actually trasmittig. Packet collisios must therefore be detected by aother meas. The IEEE 2.11 stadard uses a ackowledgemet for a poit-to-poit packet, broadcast packets are ot ackowledged. The receiver seds this ackowledgemet packet just after receptio of the packet. The MAC scheme of the IEEE 2.11 is primarily based o a CSMA (Carrier Sese Multiple Access) scheme. The mai priciple of this access techique is a prevetive listeig of the chael to be sure that o other trasmissio is o the way before trasmittig its packet. If the sesig of the chael

idicates a ogoig trasmissio, the the ode waitig to start its trasmissio draws a radom backoff delay. At the ed of the outgoig trasmissio, this backoff will be decremeted wheever the chael is free (o carrier sesed). The ode starts its trasmissio whe its backoff delay reaches 0. This mechaism is preseted i figure 1. Fig. 1. Statio A Frame Statio B Statio C Decremeted backoff Defer Defer DIFS Remaiig backoff Staio C Frame IEEE 2.11 backoff mechaism C. The simulatio model DIFS Statio B Frame We use a simulatio model ear as that used i [6]. 1) Physical layer model: The mai assumptio of our physical layer model is that we have a liear superpositio of sigals set by potetial trasmitters. This model aturally leads to the itroductio of the sigal stregth set by ode j to ode i deoted CS i, j. The sigal stregth Pow(i) received by ode i is therefore Pow(i) = N j=1 a jcs i, j where a j = 1 if ode j is trasmittig, or a j = 0 otherwise. Simple propagatio laws of radio sigals usually have the followig expressio CS i, j = P j ri, α where: j P j deotes the power set by ode j; r i, j deotes the distace betwee ode i ad ode j ad α deotes the sigal decay, usually 2 α 6. It should be oted that the oly importat assumptio is the liearity of the model. We ca actually use this liear model with all existig propagatio models or precomputed figures. We eed to itroduce the carrier sesig parameter. This parameter is a threshold above which the chael is assumed to be busy. I a CSMA protocol, this threshold makes it possible to decide whether the chael is idle or busy. We will call this parameter the carrierseselevel. We eed the to precise coditios to esure the correct receptio of packets. We will assume that a packet set by ode j to ode i i the trasmissio iterval [t b,t e ] is correctly received by ode i if : t [t b,t e ] CS i, j (t) data-level. CS t [t b,t e ] i, j k j a k (t)cs i,k (t) capture-level. There are two parameters: the data-level ad the capturelevel. The data-level correspods to the sigal stregth ecessary to successfully trasmit a sigal. The capturelevel correspods to the miimum value of a sigal to oise ratio to successfully decode a trasmissio. 2) Medium access scheme simulatio model: This model is very close to real operatios. However, it cotais two simple approximatios that simplify the ackowledgemet ad the RTS/CTS schemes. A complete descriptio of this model ca be foud i [6]. III. MANET ROUTING PROTOCOLS Ad-hoc etworks are self-orgaizig, rapidly deployable, ad require o fixed ifrastructure. Nodes i a adhoc etwork may be highly mobile, or statioary, ad may be very widely i terms of their capabilities ad uses. The primary objectives of this ew etwork architecture are to achieve icreased flexibility, mobility ad ease of maagemet relative to ifrastructured wireless etworks. Ad-hoc etwork is itself mobile. Several protocols exist, addressig the problem of routig i mobile ad-hoc etworks. We ca classify the routig protocols o the basis of their cotrol behavior i the followig categories: proactive, reactive ad hybrid. Proactive protocols use a adaptive system of routig based o periodic exchage of cotrol messages. There may be various kids of cotrol messages: those which are set locally (broadcast to oe-hop) to eable a ode to discover its local eighborhood; ad those which are set to be diffused i the etwork ad which permit to distribute the topology iformatio to all the odes i the etwork. I a proactive approach, the routig protocol periodically updates the reach ability iformatio i the odes routig table. Thereby a route is immediately available whe eeded. The cost for it is a use of substatial badwidth for the periodic cotrol traffic to acquire iformatio, some of which may ever be used. Proactive protocols iclude DSDV [7], OLSR [2,3] (a optimizatio of the lik state algorithme OSPF [8])ad TBRPF [9]. Reactive protocols do ot take ay iitiative for fidig a route to a destiatio, before the iformatio is eeded. The protocol attempts to discover routes oly o demad by floodig its query i the etwork. Durig route discovery, the data packet is put o wait util the route becomes available. The drawback of this techique is that the braod cosumptio of the badwidth for its global search (floodig) process, as well as addig large delays before sedig data packet. Examples of reactive protocols iclude AODV [] ad DSR [11]. Hybrid protocols as ZRP [12] ad CBR [13], employ a proactive scheme for oe scope or type of requiremets ad fuctios as a reactive protocol for the other. IV. OPTIMIZED LINK STATE ROUTING PROTOCOL (OLSR) OLSR [2, 3] is a proactive routig protocol, iherits the stability of a lik state algorithm [14] ad has the advatage of havig the routes immediately available whe eeded due to its proactive ature. I a pure lik state protocol, all the liks with eighbor odes are declared ad are flooded i the whole etwork. The OLSR protocol is a optimizatio of the pure lik state protocol for the mobile ad-hoc etworks. First, it reduces the size of the cotrol packets: istead of all liks, it declares oly a subset

of liks with its eighbors that are its multipoit relay selectors (see Sectio IV-A) [15]. Secodly, it miimizes the floodig of its cotrol traffic by usig oly the selected odes, called multipoit relays, to broadacst its messages. Therefore, oly the multipoit relays of a ode retrasmit the packets. This techique sigificatly reduces the umber of retrasmissios i a floodig or broadcast procedure [16, 17]. OLSR protocol performs hop by hop routig, i.e. each ode uses its most recet iformatio to route a packet. Therefore, whe a ode is movig, its packets ca be successfully delivered to it, if its speed is such that its movemet could be followed i its eighborhood, at least. A. Multipoit relay The idea of multipoit relays is to miimize the floodig of broadcast packets i the etwork by reducig duplicate retrasmissios i the same regio. Each ode m of the etwork idepedetly selects a set of odes i its oehop eighbors, which retrasmits its packets. This set of selected eighbor odes, called the multipoit relay (MPRs) of m ad deoted MPR(m), is computed i the followig maer: it is the smaller subset of oe-hop eighbors with a symmetric lik, such that all two-hop eighbors of m have symmetric liks with MPR(m). This meas that the multipoit relays cover (i terms of radio rage) all the two-hop eighbors. Figure 2 shows the multipoit relay selectio by ode m. Fig. 2. Multipoit relays of ode m m Multipoit relays of ode m Each ode m maitais the set of its multipoit relay selectors (MPR selectors). This set cotais the odes that have bee selected by m as a multipoit relay. Node m oly forwards broadcast messages received from oe of its MPR selectors. B. Neighbor sesig Each ode must detect the eighbor odes with which it has a direct ad bi-directioal lik. The ucertaities over radio propagatio may make some liks ui-directioal. Cosequetly, all liks must be checked i both directios i order to be cosidered valid. For this, each ode periodically broadcasts its hello messages, cotaiig the list of eighbors kow to the ode ad their lik status. The hello messages are received by all oe-hop eighbors, but are ot forwarded. They are broadcast at a low frequecy determied by the refreshig period Hello iterval (the default value is 2 secods). These hello messages permit each ode to lear the kowledge of its eighbors up to two hops. O the basis of this iformatio, each ode performs the selectio of its multipoit relays. These selected multipoit relays are idicated i the hello messages with lik status MPR. O the receptio of hello messages, each ode ca costruct its MPR selectors table. C. Topology iformatio Each ode with a o-empty MPR selector set periodically geerates a topology cotrol message (TC message). This TC message is diffused to all odes i the etwork at least every TC iterval. A TC message cotais the list of eighbors that have selected the seder ode as a multipoit relay. The iformatio diffused i the etwork by these TC messages will help each ode to build its topology table. Based o this iformatio, the routig table is calculated. The route etries i the routig table are computed with Dijkstra s shortest path algorithm [18]. Hece, they are optimal as cocers the umber of hops. The routig table is based o the iformatio cotaied i the eighbor table ad the topology table. Therefore, if ay of these tables is chaged, the routig table is re-calculated to update the route iformatio about each kow destiatio i the etwork. V. FAST-OLSR Fast-OLSR [4] protocol, a extesio of OLSR, is desiged to eable a fast movig ode ad keep the coectivity with other odes i the etwork by quickly discoverig a small umber of eighbors ad selectig amog them a small umber of multipoit relays. To do that, the Fast-OLSR uses a higher hello frequecy to detect quickly its eighborhood chages ad establishes a small umber of symmetric liks. Whe a ode detects that it is movig fast by a mechaism suggested i sectio 6.2.3, it activates the Fast-movig mode. I this mode, the ode seds a Fast-Hello messages at a high frequecy to establish Fast liks. A Fast-Hello message has the same format with hello message, but its size is smaller because the ode i fast movig must detect a reduced umber of its eighbors ad select over them its MPRs. Oly odes i Default Mode ca be selected as MPRs. For further detail refer to [4]. VI. SIMULATION I this sectio, we evaluate the performace of Fast- OLSR. We have carried out simulatios to aalyze OLSR ad Fast-OLSR i differet cofiguratios ad scearios. We simulate OLSR ad Fast-OLSR by OPNET simulator [19] as described i sectios 4 ad 5. A. Simulatio model 1) Assumptios: Our simulatio model is based o the followig mai priciples: The etwork is represeted by a radom graph model, i which odes are placed radomly i a give regio, iitially;

% of odes implemet OLSR ad Fast-OLSR. % of odes implemet oly OLSR protocol; All the odes are idetical (havig the same capabilities) but they fuctio idepedetly of each other; If the odes are mobile, each ode idepedetly decides its movemet: its speed ad the directio; The ode ca either receive or trasmit at a time, e.g., it ca ot receive aythig while it is trasmittig; The odes access the trasmissio chael by usig CSMA/CA protocol (described i sectio II); There are o turs aroud time betwee trasmittig ad receivig: the odes ca switch over betwee trasmit ad receive mode istatly; Mobility is ucorrelated amog the odes of a etwork ad liks fail idepedetly. 2) Topology geeratio model : We geerate the topology of the etwork by radomly distributig (radom graph) the odes i a give regio (area max x,area max y). Each ode is represeted by a subqueue ad placed i the regio by radomly selectig its x ad y co-ordiates. The umber of odes is give as etwork parameter, it ca reach 0000 odes. We select the type of mobile (implemet OLSR ad Fast-OLSR or oly OLSR) by radomly distributig i a way that % of these mobiles implemet OLSR ad Fast-OLSR ad % implemet oly OLSR. Iitially, all the mobiles are i Default mode. Figure 3 shows the ode s model represetaio thus their positios via co-ordiates x ad y. of which is characterized by its directio θ i ad distace. Fig. 4. 6 R θ 5 θ 6 R5 R4 θ3 Iterval radom mobility vectors θ 1 2 R θ4 θ 2 1 R To obtai a balace betwee the arrivals ad th departures i our area (area max x,area max y), all the odes leavig the zoe of periphery are reijected i the zoe which is symmetrically opposed to them, thus elimiatig the board s effects. figure 4 shows the reijectio of the odes M1, M2 ad M3. M2 M1 M1 R 3 After Subqueue 0 (X0,Y0) Subqueue 1 (X1,Y1) M3 M2 M3 area_max_y Fig. 3. Node s model area_max_x Subqueue 2 (X 2,Y 2) Subqueue 1 (X 1,Y 1) Subqueue (X,Y) DATA, TC, Hello Each subqueue has a profile cotaiig the idetity of the correspodig ode, the co-ordiates x ad y, the maximum speed, the miimum speed ad other useful iformatios for the implemetatio. 3) Ad-Hoc Mobility Model : The radom ad-hoc mobility model [] proposed i this sectio is a cotiuoustime stochastic process. Each ode s movemet cosists of a sequece of radom legth itervals, durig which a ode moves i a costat directio at a costat speed. To calculate the co-ordiate of ode at t durig a iterval i of duratio T i, agle θ i ad speed V i, we calculate at the first time the distace D covered by, D = V it i. The, we calculate the x, y local co-ordiates, x = Dsi(θ i), y = Dcos(θ i ). At the ed, we calculate the global co-ordiates by chagig scale. Figure 3 illustrates the movemet of ode over six mobility itervals, each Fig. 5. Way-Roud model There are three importat parameters: λ, Speed Max, Speed Mi, for calculatig the iterval legths, directio ad speed. The iterval legths are expoetially distributed with mea 1 λ. The directio of the mobile ode durig each iterval is uiformly distributed over (0, 2Π). The speed durig each iterval is uiformly distributed over (Speed Mi, Speed Max). 4) The traffic ad queuig model : Data packets are geerated at the odes accordig to the Poisso distributio. The packet arrival rate at differet odes is idepedet from each other. I our simulatio, all the odes geerate the same amout of load for the etwork. We have take the mea packet size as 1Kbytes. We limited the maximum packet size to 64Kbytes. The cotrol packets also follow this maximum size limitatio. Each ode is a subqueue to queue up the ew arrivig packets. The ew packets are queued up as log as there is a space i the subqueue. Whe the subqueue is full, the ew packet is simply rejected. Whe a packet is successfully trasmitted, the packet subqueue is freed from the trasmit subqueue ad that space is made available for a ew packet. I our simulatios, the destiatio for a data packet is radomly selected amog all the destiatios i the etwork, at each selectio.

B. Implemeted Algorithm I this part, we show the OLSR ad Fast-OLSR protocols implemetatio. 1) Mobile i Default mode: The implemetatio of OLSR protocol is represeted i the figure 6. The source layer seds a DATA packet to the OLSR layer with a specified iterarrival time. This DATA packet cotais the address of the ode, which has origially geerated this message. This field should ot be cofused with the source address from the UDP header, which is chaged each time to the address of the itermediate ode witch is retrasmittig this message. It cotais also the destiatio address ad the packet legth. The OLSR layer icludes the route i DATA packet. The ext hop router is idetified by the etry of the destiatio i the host routig table. The OLSR layer seds the hello (with the specified Hello iterval), TC (with the specified TC iterval)ad DATA messages to the MAC layer. The MAC layer trasmits the packet by applyig CSMA/CA protocol to its eighbor odes ad seds the packet well received to the OLSR layer. This layer uses a ackowledgemet for a poit-to-poit packet, broadcast packets are ot ackowledged. This ackowledgemet packet is set by the receiver just after receptio of the packet. a ode i Fast-Movig mode detects that it is o loger movig fast, it switches to the default mode. Fast-OLSR layer fuctios i the same way that OLSR cocerig the DATA packets. It seds the Fast-Hello messages to the MAC layer with the specified Fast hello iterval. There is o trasmittig of TC messages because oly odes i Default Mode ca be selected as MPRs. The switchig module is described i the ext paragraph. 3) Switchig to the Fast-Movig/Default Mode: I order to detect a high or small umber of chages i eighborhood, we proposed the model depicted i figure 8. Time correspodig of 3 Hello or Fast Hello messages Nb1 Nb2 Nb3 Nb4 Nb5 After trasmissio or receptio of Hello or Fast Hello message Treatmet Time Source Layer Fig. 8. The switchig algorithm Fig. 6. DATA packets OLSR Layer (OLSR protocol) Hello, TC ad DATA messages MAC Layer (CSMA/CA protocol) Hello, TC, DATA ad Ack messages The geeric OLSR scheme i our simulatio Hello, TC ad DATA messages Fig. 7. OLSR protocol Source Layer switchig DATA packets MAC Layer (CSMA/CA protocol) Fast OLSR protocol Fast Hello ad DATA messages Fast Hello, Hello, TC, DATA ad Ack messages The geeric Fast-OLSR scheme i our simulatio 2) Mobile i Fast-Movig mode : The implemetatio of Fast-OLSR protocol is represeted i the Figure 7. Iitially, all the mobiles are i default mode. Whe a ode detects that it is movig fast, it switches to the Fast- Movig mode ad starts sedig Fast-Hellos. Also, whe A ode witch implemet OLSR ad Fast-OLSR must detect the umber of chages i its eighborhood. Iitially, this ode waits a time correspodig of oe-trasmissio ad tow receptios of Hello (if this ode is i a default mode) or Fast-Hello (if this ode is i Fast-movig mode) messages, or two trasmissios ad oe receptio of Hello or Fast-hello messages, to record the umber of ew eighbors. This time is the smaller time to establish a symmetric liks with other odes. After this time, the ode records the umber of chages i its eighborhood Nb i. Heceforward, after each trasmissio or receptio of Hello or Fast-Hello message, the ode will record the umber of chages (ew eighbors, lost liks) Nb i. Whe the ode records Nb i (i 4), if it is i the Default mode, it compares (Nb i Nb i 3 )/Nb i with the give threshold T 1. if (Nb i Nb i 3 )/Nb i T 1, the ode activates the mode Fast-movig. If the ode is i the Fast-Movig mode, it compares (Nb i Nb i 3 )/Nb i with the give threshold T 2 (T 2 < T 1 ). If (Nb i Nb i 3 )/Nb i T 2, the ode activates the Default mode. C. Advatages of the simulatio model The proposed simulatio model is very extesible, there is several parameters: etwork parameters (umber of odes, regio,...), OLSR parameters (Hello iterval, TC iterval, use of MPRs,...), Fast-OLSR parameters (Fast hello iterval,threshold...), CSMA/CA parameters (Radio rage, oise ratio, RTS/CTS, sigal decay,...) ad mobility parameters (Speed mi, Speed max, iter arrival,...). The umber of odes ca reach 0000 odes. With our method (each ode is represeted by a

subqueue), the simulatio model is very optimized that eables to reduce the machie time ad cosequetly to icrease the time of simulatio. These, we ca passe to a few weeks of simulatio istead of few miutes that allows to refie the obtaied results. The simulatio model is very close to real Ad-hoc etwork operatios. At each time, we ca detect the positio of mobiles by our mobility model. D. Evaluatio results Figures 9 ad depict the loss of TC messages ad rate versus the mobility. Oe curve (Default mode) is draw with a etwork of odes i a regio of 0 2 m 2 where all the odes implemet oly OLSR protocol. The speed is icreased from meters/secod (36Km/hr) up to meters/secod (216Km/hr). We also keep the threshold T 1 as 0.8, T 2 as 0.2 ad the Fast hello iterval as 0 ms. Every 0 secods, the ode movemet is re-determied. Each mobile ode selects its speed ad directio, which remais valid for ext 0 secods. We chose a sigal decay α = 2 ad the trasmissio rage as meters. The secod curve (With Fast-Movig) is drow with the same etwork of the previous curve but % of odes implemet OLSR ad Fast-OLSR ad % of odes implemet oly OLSR protocol. Loss TC messages (%) 0 Fast-Movig mode Default mode ad Fast-OLSR ad % of odes implemet oly OLSR protocol. Ideed, the odes lose the topology iformatio of the etwork (Loss TC message). Cosequetly, it will have a great umber of the uavailable routes. we ote a variatio from 13% (m/s) to % (m/s) i figure 9 ad a variatio from 12% (m/s) to % (m/s) i figure. With Fast-OLSR protocol, there are less of loss TC messages; therefore, the loss of DATA will be smaller. Figures 11 ad 12 depict the loss of TC messages ad rate versus the mobility with a differet Fast hello iterval. For a speed of meters/s (144Km/hr) ad a loss rate less tha %, the oly possible Fast hello iterval is 0ms. However, for a speed limited to meters/s (8Km/hr) with a lost rate less tha %, there are several possibilities for the Fast hello iterval: 0ms, 0ms. We select the highest curve, which gives us a Fast hello iterval of 0ms. So, for a give maximum acceptable loss rate ad the maximum reachable speed, we ca determie the largest Fast hello iterval. Loss TC messages (%) Fig. 11. 0 0 ms 0 ms 0 ms Loss TC messages versus mobility Fig. 9. Loss TC messages versus mobility 0 0 ms 0 ms 0 ms 0 Fast-Movig mode Default mode Loss rate (%) Loss rate (%) 0 Fig. 12. Loss rate versus mobility 0 VII. CONCLUSION Fig.. Loss rate versus mobility As figure 9 ad show, the packet loss becomes greater as the speed icreases ad the loss of TC messages ad rate i a etwork where all the odes implemet oly OLSR protocol is greater tha the loss of TC messages ad rate i a etwork where % of odes implemet OLSR I this work, we have studied the performace of the OLSR ad the Fast-OLSR protocol (a extesio of OLSR dealig with fast mobility). We have proposed a scalable simulatio model close to real Ad-Hoc etwork. The umber of odes ca reach 0000 odes. This model is very extesible; we ca easily exted the simulatio s eviromet ad give a very large simulatio time. The

mobility model is very complete ad cotiuous-time, we ca detect the positio of mobiles at each time. We have also proposed a ew model, which permits to each ode the switch to the Fast-Movig/Default mode very quickly. The performaces of a etwork with mobiles implemeted OLSR ad Fast-OLSR i terms of loss rate are performed compared to a etwork of mobiles implemetig oly OLSR protocol. Results show that the loss rate ca be miimized for the Ad-hoc etwork of odes with fast mobility by implemetig OLSR ad Fast-OLSR protocols. REFERENCES [1] J. Macker, S. Corso. Mobile Ad hoc Networkig ad the IETF. ACM Mobile Computig ad Commuicatios Review, 1998. [2] A.Qayyum. Aalyse ad evalutaio of chael access schemes ad routig protocols for wireless etworks. Phd thesis, Computer Sciece Dept, Uiv. Paris Sud i Orsay, Novembre 00. [3] P.Jacquet, P. Muhlethaler, A.Qayyum, A. Laoutiti, L.Vieot, T.Clause. Optimized Lik state Routig protocol (OLSR). Iteret Draft, draft-ietfmaet-olsr-05.txt, 31 October 01, Work i progress. [4] M. Bezaid, P. Miet, K. A. Agha. Itegratig fast mobility i the OLSR routig protocol. INRIA research reports, Jui 02. [5] IEEE 2.11 stadard. Wireless LAN Medium Access Cotrol (MAC) ad Physical Layer (PHY) specificatios. Jue 1997. [6] P. Muhlethaler, A. Najid. A efficiet Simulatio Model for Wireless LANs Applied to the IEEE 2.11 Stadard. INRIA research reports 4182, http://www.iria.fr, April 01. [7] C. E. Perkis ad P. Bhagwat. Highly dyamic destiatio sequeced distace-vector routig (DSDV) for mobile computers, i SIG-COMM 91. [8] J. Moy. OSPF versio 2. Iteret Stadard, Request For Commets 2328, April 1998. [9] Richerd G. Ogier, Fred L. Templi, Bhargav Bellur ad Mark G.Lewis. Topology Broadcast based o Reverse-Path Forwardig (TBRPF). Iteret Draft, draft-ietf-maet-tbrpf-03.txt, November 28 01, Work i progress. [] C. E. Perkis, E. M. Royer ad S. R. Das. Ad hoc o-demad distace vector (AODV) routig. Iteret Draft, draft-ietf-maet-aodv-09.txt, November 01, Work i progress. [11] David B. Johso, Yih-Chu Hu, David A. Maltz ad Jorjeta G. Jetcheva. The Dyamic Source Routig Protocol for Mobile Ad Hoc Networks (DSR). Techical report, Rice Uiversity, AON Networks, Caregie Mello Uiversity, November 01. draft-ietf-dsr-06.txt, Work i progress. [12] Zygmut J. Haas ad Marc R. Pearlma. Zoe Routig Protocol (ZRP) for AD HOC Networks. Iteret Draft, draft-ietf-maet-zoe-zrp-02.txt, Jue 1999, Work i progress. [13] M. Jiag, J. Li, Y-C. Tay. Cluster Based Routig Protocol (CBRP). Iteret Draft, draft-ietf-maet-cbrp-spec-01.txt, August 1999, Work i progress. [14] E. Streestrup. Routig i Commuicatio Networks. Pretice Hall Iteratioal, Ic., Eglewood Cliffs, NJ, 1995. [15] A. Qayyum, L. Vieot, A. Laouiti. Multipoit relayig: A efficiet techique for floodig i mobile wireless etworks, Research Report RR- 3898, INRIA, http://www.iria.fr, February 00. [16] P. jacquet, P. Muhletaler, A. Qayyum, A. Laouiti, T. Clause, L. Vieot. Optimisatio lik state routig protocol. I IEEE INMIC, (Pakista), December 01. [17] A. Qayyum, L. Vieot, A. Laouiti. Multipoit relayig techique for floodig broadcast message i mobile wireless etworks. I HICSS: Hawai Iteratioal Coferece o System Scieces. Hawii, USA, Jauary 02. [18] A. S. Taebaum. Computer Networks. Pretice Hall, 1996. [19] OPNET simultor, http://www.opet.com/. [] A. Bruce ad T. Zati. A Mobility Based Framework for adaptive Clusterig i Wireless Ad-Hoc Networks. Appears i IEEE joural o Selected Areas i Commuicatios, Vol. 17, No. 8, August 1999.