Experimental Performance Evaluation of AODV Implementations

Size: px

Start display at page:

Download "Experimental Performance Evaluation of AODV Implementations"

Shannon Lucas
5 years ago
Views:

1 Experimetal Performae Evaluatio of AODV Implemetatios Koojaa Kuladiithi, Asaga Udugama, Niko Fikouras, Carmelita Görg Uiversity of Breme Cmuiatio Networks (CN) Otto-Hah-Allee NW Breme {koo adu iko g}@.ui-breme.de Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

2 Overview Motivatio Implemetatio Cosideratios Testbed Set UP UDP Results Aalysis TCP Results Aalysis Further Ivestigatios Colusio Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

3 Motivatio Lak of performae evaluatio i real AODV etworks Simulatio (Easy to develop, Simple to get/aalyze/repeat Results) Implemetatio (Diffiult to develop, ode should be deployable, results should be realisti) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

4 Implemetatio Cosideratios Uiast ad broadast/multiast muiatios RREQ, RERR Multiast/Broadast RREP, RREP-ACK, RERR uiast Itereptig IP pak that require a route (to mee the Route Disovery proess) Iformatio of IP pak that utilize the existig routes (to update route lifetime) Maipulatio of the Routig Table (Add, Delete & Modify routes) Support for a Timer Mehaism (Route disovery proess, Route maiteae, Route deletio, Hello Messages, et.) Paket Queuig (Bufferig) Lik break detetio (Hello Messages, Lik Layer) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

5 Implemetatio Cosideratios, ot. Possible Arhiteture Models Iterfae Moitorig Arhiteture (Purely o User Spae) Netfilter based Arhiteture (User Spae + Kerel Spae) Kerel based Arhiteture (Kerel Spae) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

6 Implemetatio Cosideratios, ot. User Spae User Spae AODV User Proess User Apps AODV User Proess User Apps Kerel Spae Sok Kerel Spae Sok/IOCTLs Filter Lik Layer Driver Lik Layer Driver Network Iterfae Network Iterfae Iterfae Moitorig Arhiteture (Purely o User Spae) Filter based Arhiteture (User Spae + Kerel Spae) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

7 Implemetatio Cosideratios, ot. User Spae User Apps Kerel Spae Sok AODV Kerel Proess Lik Layer Driver Network Iterfae Kerel based Arhiteture (Kerel Spae) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

8 Cparisio of Implemetatios UU-AODV implemetatio by Uppsala Uiversity, Swede Filter based arhiteture o Liux platform Notebooks & PDAs Versio 0.7 JAdho- AODV implemetatio by Uiversity of Breme, Germay Iterfae moitorig arhiteture for Liux/Widows Platforms Notebooks & PDAs Versio 0.2 Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

9 Testbed Set-Up Node 1 Node 2 Node 3 Node 4 Node 5 Node 6 Stati (o-mobile) Ad-ho /w (oferee ro, leture ro) Same frequey CH MACKill Filter (at the /w layer) 3 modes of ofiguratios UU (versio 0.7) - AODV implemetatio by Uppsala Uiversity, Swede JAdho (versio 0.2) - AODV implemetatio by Uiversity of Breme, Germay Stati Set routes maually i ad-ho etwork, (No AODV proess is ruig) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

10 Testbed Set UP, Cot.. Hardware 6 Soy Vaio otebooks (Mobile AMD Duro proessors ruig at 1.1 GHz ad 256 MB of RAM) For wireless oetivity, Ciso Airoet IEEE b wireless ards (set i ad-ho mode o hael 1, o WEP eryptio) Software Liux Madrake 8.2 distributios ad the Liux Kerel I order to perform the experimetal evaluatio while maitaiig the otebooks i lose physial distae to eah other the MACKill software was istalled o eah otebook to filter IP Pak at the lik layer Iperf traffi geerator software to geerate UDP ad TCP flows Web100 to measure a seleted set of TCP parameters (RTO, CWND) TCPDUMP to idetify DupAks The WLAN iterfae of eah otebook has bee alloated a differet IP address fr a differet sub etwork Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

11 Testbed Set-Up - UDP Measurem Foud the maximum data rate that a be used without ay paket loss at the lik layer Use Stati Mode with 2 odes Max of 3.6 Mbps of UDP stream without ay paket loss [ Whe usig 6 odes, ~ 0.5Mbps UDP stream a be used to sed data without ay lik layer paket loss] Ui-diretioal data flow at 0.5Mbps is used to evaluate upper layer performae Atual Load Paket Delivery Fratio Out of order Pak Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

12 Testbed Set Up TCP Measurem TCP throughput is measured Throughput variatios are aalysed usig the followig parameters: RTO CWND DupAks Both UDP & TCP measurem are take With the irease of odes fr 2 to 6(Fr 1 5 hops), while takig ode 1 as the seder For all 3 modes (Stati, UU, JAdho) For the duratio of 60 seods All UDP measurem ad TCP throughput are take at the reeiver side, exept TCP parameters Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

13 UDP Results Aalysis Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

14 UDP Results Aalysis Paket Delivery Ratio vs Number of odes UDP Results Less with the irease of odes (hop outs) Lies betwee % Most paket loss ouurs whe releasig buffered pak (just after the route disovery time explaied i the ext slide) Therefore, PDR remais lose to 100% for loger UDP sessios Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

15 UDP Results Aalysis Throughput betwee 4 odes UU & JAdho have higher throughput (tha the give throughput) at the begiig, due to the release of buffered pak Causes pak to be lost due to the higher data rate at the lik layer, while releasig the buffered pak Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

16 UDP Results Aalysis Atual load for a offered load of 0.5Mbps vs Number of odes Average throughput of both UU & JAdho is ~ 0.5 Mbps JAdho has less throughput with a 6 ode setup Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

17 UDP Results Aalysis Paket Sequee Number Vs Time Buffered pak i JAdho Liear i Stati mode UU releases buffered pak before allowig other pak (pak go just after settig routes) JAdho has out of order pak due to user spae bufferig (Other pak e first before the buffered pak) Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

18 UDP Results Aalysis UDP Performae i geeral Stati > UU > JAdho JAdho Route disovery time is higher due to ruig at the user spae Issues to be improved with JAdho Bufferig Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

19 TCP Results Aalysis Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

20 TCP Results Aalysis Throughput vs Number of odes Less with the irease of odes Cotrast to UDP throughput, JAdho shows the highest throughput for TCP Closer aalysis of TCP throughput of both UU ad Stati shows that UU has a higher throughput tha Stati Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

21 TCP Results Aalysis Throughput betwee four odes Reasos for the Throughput variatios are ivestigated by aalyzig TCP parameters dupaks, CWND & RTO for eah mode of muiatio Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

22 TCP Results Aalysis TCP Flow paket trae, with dupaks for Stati mode Paket Trae TCP Results JAdho Paket Trae DUPACKS - Higher i Stati > JAdho Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

23 TCP Results Aalysis TCP Flow paket trae, with dupaks for UU mode Paket Trae DUPACKS are higher i Stati > UU > JAdho Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

24 TCP Results Aalysis (CWND & RTO) CWND (Slides 34 & 35) - Variatio of CWND Stati > UU > JAdho - Frequey of throughput variatio is higher i Stati RTO (Slides 36 & 37) -Variatios of RTO JAdho > UU > Stati - I Stati, more pak set to lik layer - Therefore, higher paket loss at the lik layer Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

25 TCP Results Aalysis JAdho, beig a user spae program, RTO is higher for TCP muiatio TCP itself adjust with RTO variatios ad put less pak, ausig less losses at the lik layer TCP throughput (i geeral) is mostly degraded i ad ho mode, due to the Exposed Node Problem A B C D 1. B seds data to A 2. B seds RTS to A, C also reeives it 3. C s NAV as arrier busy 4. C a t sed data to D simultaeously 5. C is a exposed ode durig data trasmissio betwee A & B Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

26 Additioal Observatios 1) Effet of Hello Iterval*Hello Loss HelloIterval*HelloLoss is Lower, the probability of Route Breaks are higher at higher data rates Solutios: Better to use Lik layer ifo to detet eighbor Use Higher value for Hello Loss i stati /w that uses higher data rates 2) Lifetime of the reverse route MiimalLifetime = (urret time + 2*NET_TRAVERSAL_TIME - 2*HopCout*NODE_TRAVERSAL_TIME) Route Lifetime = 140ms (HopCout=3,NODE_TRAVERSAL_TIME=10, Net Diameter =10) Cosider the proessig delay for the MiimalLifetime Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

27 Further Ivestigatios Cosider the mobility i the same set up Variatios of results with the bakgroud traffi Effet of AODV parameters Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

28 Colusios Arhiteture ifluees performae, due to followig, the way paket bufferig is doe i User Spae implemetatios - results i iitial delays ad uordered pak Paket Proessig delay Trasport layer performae a be improved by fie tuig AODV parameters to suit the topology (stati or mobile) UU Liux Platform Performae is better due to use of Netfilter JAdho Bufferig has to be improved Ca easily be made to work aross differet platforms Works o Liux/Widows XP & 2000 Possibility to be adopted to ru o Java based mobile devies Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

29 Rest of TCP Results Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

30 TCP Results TCP throughput variatios betwee 4 odes Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

31 TCP Results TCP Flow paket trae, with dupaks for Stati mode Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

32 TCP Results TCP Flow paket trae, with dupaks for UU mode Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

33 TCP Results TCP Flow paket trae, with dupaks for JAdho mode Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

34 TCP Results Variatio of wd for Stati & UU modes Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

35 TCP Results Variatio of wd for Stati & JAdho modes Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

36 TCP Results RTO Variatios for Stati & UU modes Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

37 TCP Results RTO Variatios for Stati & JAdho modes Uiversität Breme Kmuikatiosetze Experimetal Performae Evaluatio of AODV Implemetatios

Optimal Placement of Mesh Points in Wireless Mesh Networks 1

Optimal Placement of Mesh Points in Wireless Mesh Networks 1 Optimal Plaemet of Mesh Poits i Wireless Mesh etworks Suk Yu Hui 2, Kai Hau Yeug 2, ad Ki Yeug Wog 3 2 Departmet of Eletroi Egieerig, ity Uiversity of Hog Kog syhui@ee.ityu.edu.hk, eeayeug@ityu.edu.hk