Packet Loss in Mobile Ad Hoc Networks(5)

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

3)PacketLossComparisonforAODVandDSDV:ThecomparisonofdifferentpacketlossesforAODVandDSDVisasfollows.

Totalpacketloss:ThetotalpacketlossforDSDVisalways10%to20%higherthanthatofAODV,regardlesspausetimeornumberofconnections.Formoderateandheavycommunicationrequests,totalpacketlossforDSDVismorestablethanthatofAODVwiththeincreaseofpausetime.

Congestion-relatedpacketloss:DSDVlosesmorepacketsduetocongestionthanAODVdoes.Thegapofcongestion-relatedpacketlossbetweenDSDVandAODVdecreaseswiththegrowthofcommunicationrequest.

Mobility-relatedpacketloss:AODVhasmoremobility-relatedpacketlossthanDSDVdoes.B.VaryingTraf cLoadandTraf cType

Thesecondsetofexperimentsillustratetheeffectoftraf cloadandtraf ctype.Pausetimerangesover{0,50,100,200,300,500}seconds.10,20and30connec-tionsareused.Bothunresponsivetraf candresponsivetraf carestudied.ThepacketrateforCBRconnectionsis8packets/s,whichinjectsareasonableheavyloadtothenetwork.Weusethesamemobilityscenariosandconnectioncon gurationsforthissetofexperimentsasfortheprevioussetofexperimentstocomparetheresultswiththepreviousones.

1)CBRconnectionswith8packets/s:Asshownin gure2,eachcurvethatrepresentsadifferenttypeofpacketlosshassimilarshapeasthecorrespondingoneinthepreviousexperiments,but atter(i.e.,increaseanddecreasearemoregradually).

ForAODV,mobilityisstillthemajorcauseforpacketloss.CongestionplaysamoreimportantrolecomparedwithCBRconnectionwith4packets/srate.Increasingnumberofconnectionshaslesseffectinthissetofexper-imentsthaninthepreviousone.From10connectionsto20connections,totalpacketlossincreasesbyonlyabout3times.From20to30connections,paring gure2awith gure1a,totallossincreasesby660%for0secondpausetime,andby200%for500secondspausetime.Formoderateandheavycommunicationrequests,totalpacketlossisonlytripledordoubledaspacketrateincreasesfrom4to8packets/s.

ForDSDV,congestiondominatespacketlossevenwhenthereareonly10connections.TotalpacketlossincreasesthesameamountasthatforAODVwhencom-municationrequestincreases,withrespecttopercentage.Totalpacketlosswithtraf cloadisalmostthesameforDSDVandAODV.

Fig.2.Packetlossfor8packets/sCBRconnections

ForbothAODVandDSDV,increasingcommunicationrequesthassimilarimpactontotalpacketloss(i.e.,morelosses)asincreasingtraf cload.Theincreaseofeitherparameterwillresultindecreasingtheimpactoftheotherparameter.Heaviercommunicationrequestortraf cloadintroducesmorecongestion-relatedpacketloss.

2)TCPconnections:Numberofbytes(thetotalsizeofalllostpackets),insteadofnumberofpackets,isusedinexperimentswithTCPconnections.BecausebothapplicationdataandACKpackets,whichhavedifferentsizes,aretreatedasdatapacketsbytheroutingprotocol,thenumberofbytesismorecomprehensivethanthenumberofpackets.

Figure3demonstratesbytelossinTCPconnec-tions3.Itshowsthatthecongestion-relatedlossforbothprotocolsisgreatlyreducedbythecongestioncontrolmechanism.Totallossdecreaseswiththedecreaseofmobility.DSDVoutperformsAODVintermsoftotalloss.TotallossofDSDVisonlyhalfofthatofAODVinalltestcases,becausetheeffectofthemajorcauseforDSDVtolosepackets(i.e.,congestion)isoffsetbythecongestioncontrolmechanism.

ForAODV,withthedecreaseofcongestion-relatedloss,morethan90%oftotallossismobility-related.Thetotaleffectofmobilityandcongestionislessthan20%forDSDV.

Toimprovethroughput,differentroutingprotocolsrequiredifferentmechanismstoremedythemajorcauses

ThedifferenceoftheamountsofbytessentbyAODVandDSDVissmallerthan5%.

3