SANE is based on the observationthat we validate on real world tracesthat individuals with similar interests tend to meet more often In our approach individuals network members are characterized by their interest pro64257le a compact representation ID: 22886
Download Pdf The PPT/PDF document "SocialAware Stateless Forwarding in Pock..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
III.INTERESTSPACEANDPROFILESWeassumeeachindividualinthenetworkcanberepre-sentedthroughherinterestprole,acompactrepresentationofherinterestswithintheinterestspace.Werepresenttheinterestspaceasanm-dimensionalunitcubeC=[0;1]m,wheremisthetotalnumberofinterestsinthenetwork.Interestsareintendedinaverybroadsense,theymightrepresentdegreeofinterestinacertaintopic(e.g.,cinema,literature,etc.),thefactthatanindividualbelongstoacertainphysicalorvirtualcommunity(e.g.,livinginacertainneighborhood,memberofaFacebookinterestgroup,etc.),andsoon.Giventheabovedenitionofinterestspace,itisnaturaltorepresenttheinterestproleofanindividualAwithanm-dimensionalvectorreporting,foreachpossibleinterestdimen-sion,A'sdegreeofinterestintheparticulartopic/community(eitherarealnumberorabinaryvalue).Toexpresssimilaritybetweenindividualinterests,andthusquantitativelymeasurehomophilydegreeofinterestsimilarity[11]weusethewell-knowncosinesimilaritymetric[4]:Denition1:Giventwom-dimensionalvectorsAandB,thecosinesimilaritymetric,denoted(A;B),isdenedasfollows:(A;B)=cos(\AB)=AB kAkkBk;wherekXkrepresentthelengthofvectorX.Notethat,giventhedenitionofinterestspace,0(A;B)1inourmodel,withhighervaluesof(A;B)correspondingtoahigherhomophilydegree.Ourstatelessprotocolsarebasedonasimpleandnaturalobservationfromeverydaylife:Ourmovementsareguidedbyourinterests.Tovalidatethisintuition,weusetracescollectedduringanexperimentdonewithrealBluetoothcommunicatingdevicesdistributedtoparticipantsoftheInfocom2006con-ference[6],[7].Thisdatatracecontainsnotonlycontactlogs,alsodoesitreportinformationonparticipants'nationality,residence,languagesspoken,afliation,scienticinterests,etc..Fromthisinformationwegenerateinterestsproleasdenedabove.Intheprocess,wediscardparticipantsthathavenotdeclaredanyinterest,inordertoremoveerroneousproles.Thisway,thenumberofparticipantsreducesto61nodes.Tosupportourintuition,werstcalculatethecosinesim-ilaritybetweentheinterestprolesforeverypairofpartici-pants.Then,wecomputethePearsoncorrelationindexamongthisvalueandthetotalmeetingduration/meetingfrequencyamongeverycouple.Wethencomputethecorrelationcoef-cientsamongprolesimilaritiesandmeetingduration/meetingfrequency,onlyforpairsofindividualswhospend,ontheaverage,morethanacertainamountoftimetogether.Thiswaytheeffectofcasualshortmeetingsisattenuated.TheresultsarepresentedinTableI.Ascanbeseen,whenwefocusonlongermeetings,thecorrelationofmeetingfrequencyandsimilarityofinterestprolesisconsiderablyhigh,reaching0.67.Theseresultssupporttheconclusionthatourintuitionissoundandthatitcanbeusedasthebasicmechanismofsocial-aware,statelessforwardingprotocols. AVGmeettime Cd Cf Nodes 0(min) .28 .08 61 5(min) .55 .57 53 10(min) .67 .67 26 TABLEICORRELATIONBETWEENINTERESTSPROFILESANDPARTICIPANTS'ENCOUNTERS.CdANDCfINDICATETHEPEARSONCORRELATIONCOEFFICIENTBETWEENPARTICIPANTS'COUPLESPROFILESANDRESPECTIVELYTOTALMEETINGDURATIONSANDMEETINGRATES.IV.SOCIALAWARENETWORKING(SANE)Inthissection,wedescribeSocialAwareNEtworking(SANE),aprotocolsuitethatenablestheefcientdeliveryofinformationtorelevantdestinationsinPSNs.SANEsupportsanovelcommunicationservice,thatwecallinterest-cast(seeSectionIV-B),besidesthetraditionalunicast.Weassumethateachnodecanbeaforwarderandthereforemaintainsabufferofmessagesthatmustberelayedtotherespectivedestinations.EachmessageMhasaheaderthatcontainsatargetinterestprolethatwecallmessagerelevanceprole,anintegervalueNreplicasrepresentingthenumberofreplicasofthemessagethatthenodeisallowedtoforwardtootherrelays,andatime-to-livevalueTTL.InPSNsnodescanexchangeinformationasacommunicationopportunityarises.Accordingly,SANEproceduresaretriggeredeachtimeanode,sayA,enterswithintheradiocoverageofanothernode,sayB.Initially,nodesexchangetheirinterestprole(IP),theneachnodestartscanningitsbufferformessagestorelay.A.UnicastIntheunicastcasethemessagerelevanceproleissetequaltotheinterestproleofthedestination.Accordingtoourinterest-basedapproach,amessageMshouldpreferablybeforwardedtoindividualswhoseinterestprolecloselyresemblestheoneofthedestination.Morespecically,asin[14],weassumethatinordertokeepthecommunicationoverheadundercontrol,thesamemessagecanberelayedatmostforNreplicastimes.Messagerelayingobeythefollowingrules:MessageMshouldberelayedtoanodeBifanonlyifboththetwofollowingconditionshold:thecurrentvalueofNreplicasishigherthan1.thecosinesimilaritymetricbetweentherelevanceofmessageM,denotedasR(M),andtheIPofB,denotedIP(B),ishigherthanagiventhresholdthatwecallrelayingthreshold.Incase,thevalueofNreplicasinthemessageheaderishalvedandacopyofthemessageissenttoB.ObviouslythemessageistransmittedtonodeBregardlessofthevalueofNreplicasifBthedestinationofthemessage.Inthiscase,nodeAremovesthemessagefromthebufferafterthisisrelayedtoB.Notethat,asthethresholddecreases,theforwardingstrategybecomesmoreaggressive.Thisresultsinthedecreaseofthedeliverydelay,andintheincreaseofboththedeliverysuccessprobabilityandthecommunicationoverhead(cost)incurredforthedeliveryofthemessageM,thatwedenote allthecompetingprotocols,atalowercostandatasimilarorsmallerdelay.Quiteinterestingly,bufferlimityieldsthesurprisingresultthatUN-SANEEPhashighersuccessratethanEpidemicitself.2)Interest-cast:Here,weshowresultsrelatedtothetwointerest-castversionsofourprotocol:SANESW,andSANEEP.SincethereisnoimmediatewayofextendingBUBBLEintoaninterest-castprotocol,wecompareSANEprotocolsonlytoEpidemicandBinarySW,whoseinterest-castversionsarestraightforward(simplydeliversacopyofthemessagetoallrelevantdestinations).ThewaywegeneratemessagesandtheinputtuningparametersofBinarySWandSANESWarethesameasinunicast.TheresultsareshowninFigure2.Inthiscase,coveragereferstothepercentageofrelevantdestinationsholdingacopyofthemessagewhentheTTLexpires.Asseenfromthegures,SANEprotocolsperformverywell,providingcomparablecoverageofrelevantdestinationstothatofEpidemic(forTTLsvalueslargethan30min),butwithamuchreducedcost(asmuchas10-foldcostreductionwithrespecttoEpidemic,incaseofSANESW).Thebenetsofsocial-awareforwardingareevidentcomparingtherelativeperformanceofBinarySWandSANESW:withacomparablecost,SANESWprovideshighercoverageandlowerdelaythanBinarySW.B.SANEwithSyntheticTracesThesynthetictracesweuseforevaluationhavebeenob-tainedfromtheSWIMmobilitymodel[12],[13].InSWIM,nodesareassignedahomepointinthenetworkarea,assumedtobeasquare.Eachtimeanodehastochooseitsnextdesti-nation,ittradeoffsdistancefromitshomepointandpopularityofthepossibledestinations.Thus,nodeswithrelativelyclosehomepoints(neighbors)tendtogotothesamelocationsandgetincontactmoreoften.InordertorunSANEonSWIM'straces,wedothefollowingsetup:First,wegenerateathenetwork,andagivennumberofnetworknodes.Foreachnode,a4-dimensionalinterestprolevectorisrandomlygenerated,withentrieschosenindependentlyanduniformlyatrandomin[0;1].Eachprolevectoristhennormalizedto1thisway,wemakesurethatnonodehasverylowinterestsornointerestsatall.InSWIM,neighborstendtohaveahighermeetingrate.TheamountofcorrelationbetweenvicinityofhomepointsandmeetingrateinSWIMiscontrolledbyaparameterthatherewedenoteas(see[12],[13]fordetails):Thehigherthisparameter,thehigherthecorrelation.Therefore,wecangenerateSWIMmobilitytracescontrollingtheresultingcorrelationbetweennodeprolesimilarityandtheirmeetingfrequencybytuningSWIM'sparameter.Duetospacelimitation,inthefollowingwewillonlyshowresultsforaSWIMsimulationwith200nodesscatteredinasquareareaof500m500mandwithsetinsuchawaythatthecorrelationbetweeninterestprolesimilarityandpairwisemeetingratesisabout.7.Unfortunately,duetolackofspace,herewedonotpresentSWIM-basedcomparisonresultsofSANEwiththeafore-mentionedwell-knownforwardingbasedprotocols.Stillwewanttostressthatduetothehighcorrelationbetweennode-prolesandpairwisemeetingratestheadvantageoftheSANEprotocolsoverthecompetitorsbecomesevenmoreevidentthaninInfocom06simulations.Figures3and4weshowthesuccessrate,averagecostandaveragedelayperreceivedcopywhentherelevancethresholdis=:95,versusthevalueoftherelayingthreshold.Asexpected,thecommunicationcostincreasesasthevalueofdecreases.VI.CONCLUSIONSInthispaper,wehaverstvalidatedtheintuitionthatindividualswithsimilarintereststendtomeetmoreoftenthanindividualswithdiverseinterests,andthenusedthisintuitiontodesigntherstsocial-aware,statelessforwardingmechanismforopportunisticnetworks,calledSANE.AnicefeatureoftheSANEforwardingapproachisthatitcanbeusednotonlyfortraditionalunicastcommunication,butalsoforrealizinginnovativenetworkingservicesforPSNs,suchasinterest-casting.Whencollectivelyconsidered,theexperimentalresultsclearlyshowthesuperiorityofSANEprotocolsoverbothsocialoblivious,statelessandsocial-aware,statefulapproaches.Quiteastonishingly,SANEprovidesbet-terperformancethancompetitorsevenwhenthedegreeofcorrelationbetweeninterestprolesimilarityandpairwisemeetingratesismodest,asintheInfocom06scenario.Ifthiscorrelationishigher,asitmightbeexpectedinpracticalsituations,weexpectthattheadvantagesofSANEprotocolsovercompetitorsbecomesubstantial.REFERENCES[1]C.Boldrini,M.Conti,A.Passarella,ContentPlace:Social-AwareDataDisseminationinOpportunisticNetworks,ACMMSWiM2008.[2]P.Costa,C.Mascolo,M.Musolesi,G.P.Picco,Socially-AwareRoutingforPublish-SubscribeinDelay-TolerantMobileAdHocNetworks,IEEEJSAC,Vol.26(5),2008.[3]E.Daly,M.Haahr,SocialNetworkAnalysisforRoutinginDiscon-nectedDelay-TolerantMANETs,ACMMobiHoc2007.[4]M.M.Deza,E.Deza,EncyclopediaofDistances,Springer,Berlin,2009.[5]W.Gao,Q.Li,B.Zhao,G.Cao,MulticastinginDelayTolerantNetworks:ASocialNetworkPerspective,ACMMobiHoc2009.[6]P.Hui,J.Crowcroft,E.Yoneki,BUBBLERap:Social-basedForward-inginDelayTolerantNetworks,ACMMobiHoc2008.[7]P.Hui,E.Yoneki,S-Y.Chan,J.Crowcroft,DistributedCommunityDetectioninDelayTolerantNetworks,ACMMobiArch2007.[8]P.Hui,A.Chaintreau,J.Scott,R.Gass,J.Crowcroft,C.Diot,Pocket-SwitchedNetworksandHumanMobilityinConferenceEnvironments,ACMWDTN2005.[9]S.Ioannidis,A.Chaintreau,L.Massoulie,OptimalandScalableDistributionofContentUpdatesoveraMobileSocialNetworks,IEEEInfocom2009.[10]F.Li,J.Wu,LocalCom:ACommunity-BasedEpidemicForwardingSchemeinDisruption-tolerantNetworks,IEEESecon2009.[11]M.McPherson,Birdsofafeather:HomophilyinSocialNetworks,AnnualReviewofSociology,vol.27,n.1,pp.415444,2001.[12]A.Mei,J.Stefa,SWIM:ASimpleModeltoGenerateSmallMobileWorlds,IEEEInfocom2009.[13]S.Kosta,A.Mei,J.Stefa,SmallWorldinMotion(SWIM):ModelingCommunitiesinAd-HocMobileNetworking,IEEESecon2010.[14]T.Spyropoulos,K.Psounis,C.S.Raghavendra,EfcientRoutinginIntermittentlyConnectedMobileNetworks:TheMulti-copyCase,IEEETrans.onNetworking,vol.16(1),2008.[15]A.Vahdat,D.Becker,EpidemicRoutingforPartiallyConnectedAdHocNetworks,TRCS-200006,DukeUniv.,2000.