Packet Loss in Mobile Ad Hoc Networks(6)

Abstract — We investigate packet loss in mobile ad hoc networks via simulation. Ad-hoc on-demand distance vector (AODV) and destination-sequenced distance vector (DSDV) are chosen as representatives of the on-demand and proactive routing protocols respect

Fig.3.PacketlossforTCPconnections

forpacketloss.Speci cally,integratingcongestioncon-troltechniqueswithDSDVwillsigni cantlyimprovethethroughput,asshownin gure3.Foron-demandroutingprotocolslikeAODV,fastrediscoveryofnewrouteswillreducemobility-relatedpacketloss,andgainhigherthroughputconsequently.S.R.Dasetal.proposedadhocon-demandmultipathdistancevector(AOMDV)protocoltodecreasetheroutediscoverylatency[16].TheirresultshowedthatAOMDVlosesfewerpacketsthanAODV(3-5%less).

V.DISCUSSIONS

Thesimulationresultsbringoutsomeinterestingfactsandgiverisetoseveralimportantproblems.Theyarediscussedinthissection.A.ShortestPathandCongestion

Figure1and2showthatDSDVlosesmuchmorepacketsduetocongestionthanAODVdoes.Sincetheperconnectiontraf cloadismuchlighter(lessthan8packets/s=32Kb/s)thanthecommunicationcapacityofahost(2Mb/s),theoccurrenceofcongestionindicatesthatconnectionsconvergeonheavilyburdenedhosts.Theconvergedtraf cloadexceedsthecapacityofthosehosts.Thisdifferencemayresultfrom,withaverygreatchance,thedifferentroutemaintenanceschemesusedbyDSDVandAODV,becausebothprotocolsusedistancevectortorepresentroutinginformationandchoosetheroutesbasedontheshortestpaths,.Inamobileadhocnetwork,hostskeepmoving.Theshortestpathbetweenasourceandadestinationmaychangeas

6

timepasses.DSDVrequiresperiodicalupdatesofroutinginformation.Everyhosthasthemostrecentknowledgeaboutroutes.Itislikelythatthepathchosentoforwardpacketsisthecurrentlyshortestone.IncontrasttoDSDV,AODVpicksupapath(usuallytheshortestone)whenahostinitiatesaroutediscovery.Thehostkeepssendingpacketsviathispathuntilitbreaks,evenifshorterpathsbecomeavailableafterroutediscovery.

2

2

Fig.4.Shortestpathandcongestion

Thedifferencebetweenthesetwostrategiescanbeillustratedwith gure4,inwhichSisasetofsourcehostsandDisasetofdestinationhosts.PshortestpathsbetweenSandD.Originally,1andP2aretwobothDSDVandAODVsendpacketsfromStoDthroughthesetwopaths.Attimet,ahostHmovesinbetweenSandD,andashorterpathisavailable.AODVstillsendspacketsviaP1andPpath2.DSDV,however,sendsallpacketsthroughthenewonceit ndsoutthenewoneisshorter.CongestionmayoccurathostHwhentraf cloadexceedsitscapacity.Thisexampleshowsthatkeepingsendingpacketsthroughtheshortestpathmaycausecongestion.

B.PacketLossDistributionoverTime

The gurespresentedinsectionIVprovidestatisticalresultsofpacketlossoverthesimulationtime.Wefurtherinvestigatethepacketlossproblembyexploringanswerstothefollowingtworesearchquestions.

Whatisthedistributionofpacketloss?Whatarethecharacteristicsofthedistribution?

Ispacketlossevenlydistributedoverallhosts?Whatisthedistributionofpacketlossataspeci chostoverthetime?

Anotherexperimentisconductedtostudythepacketlossdistribution.Sampledataiscollectedevery10seconds.20CBRconnectionswitharateof4packets/sareused.Pausetimeis50seconds.DSDVisusedastheroutingprotocolbecauseithascomparablecongestion-relatedandmobility-relatedpacketlosses.Togetenoughsampledata,thesimulationrunsfor2500seconds.Figure5showsthedistributionsoftotalpacketlossatmobilehostswithID0,10,and20(everyhostisgiven