Traffic Engineering in Software Defined Networks

2024年3月7日發(fā)(作者：我可愛(ài)的小狗)

2013 Proceedings IEEE INFOCOM1Traf?cEngineeringinSoftwareDe?nedNetworksSugamAgarwal1MuraliKodialamBellLabs,Alcatel-LucentHolmdel,NJ,USA{muralik,lakshman@}anAbstract—SoftwareDe?nedNetworkingisanewnetworkingparadigmthatparatesthenetworkcontrolplanefromthepacketforwardingplaneandprovidesapplicatallycentralizedcontrollerthathasaglobalnetworkviewisresponsibleforallthecontroldecisionsanditcommunicateswiththenetwrecentlyannounced[5]thatitisusingaSoftwareDe?nedNetwork(SDN)tointerconnectitsdatacentersduetotheea,ef?ciencyand?exibilityinperformingtraf?ctstheSDNarchitecturetoresultinbettributionofthispaperisontheeffectiveuofSDNsfortraf?cengineeringesicular,weshowhowtoleveragethecentralizedcontrollertogetsigni?cantimprovthattheimprovementsarepossibleeveninculatetheSDNcontroller’soptimizationproblemfortraf?cengineeringwithpartialdeploymentanddevelopfastFullyPolynomialTimeApproximationSchemes(FPTAS),bybothanalysisandns-2simulations,theperformancegainsthatareaUCTIONSoftwareDe?nedNetworking(SDN)isanewnetworkingparadigm[1],[13],[14]nctionalparationandtheim-plementationofcontrolplanefunctionsonparatecentralizedplatformshasbeenofmuchrearchinterestduetovariouxpectedoperationalbene?tinginterdomainroutingfromindividualroutersandusingalogicallycentralizedRout-ingControlSystemwaspropodin[2],[4]asameanstomakeroutingsystemsmoremanageable,tRouterarchitecture[12]propodthedisaggregationofroutersintopacketforwardingelements,withopenstandardizedinterfaces,spropodasameansforquickerintroductionofnetworkfunctionssuchastraf?cengineering,newVPNfeatures,otoachievevariousothercost,operationalbene?-architectingthecontrolplaneintoadisminationplaneandadecisionplanewaspropodin[10].Thedisminatonplanereliablydisributesinformationtothenetworkelementsandthedecisionplanemakesalldecisonsthataffectthenetwork.1WorkdonewhileatBellLabsThisre-architecturemakesiteasierforthedecisionplanetohaveaglobalviewofnetworksttoalltheaboveapproachesisthatanSDNcomprisoftwomaincomponents:?SDNController(SDN-C):Thecontrollerisalogicallycentralizedfunction[3],[8].trollersdeterminestheforwardingpathforeach?owinthenetwork.?SDNForwardingElement(SDN-FE):icforforward-ingthepacketsisdeterminedby?ow[14]isastandardizedinterfaceth?owisinitiatedinthenetworkthefollowingactionsaretaken:(i)The?rstpacketofthe?owisntbytheSDN-FEtotheSDNcontroller,(ii)Theforwardingpathforthe?owiscomputedbytheSDNcontroller(SDN-C),(iii)TheSDN-CndstheappropriateentriestoinstallintheforwardingtablesateachSDN-FEonthepathfromthesourcetothedestination,(iv)Allsubquentpacketsinthe?controllerisresponsibleforpathsopossibletoimplementcentralizedorpartiallycentral-izedtraf?tance,GooglehasbuiltaSDNwithOpen?owrouterstointerconnectitsdatacenters(G-Scale)[5].Googleixpectinganimprovementinnetworkutilizationof20-30%aswellasimproveddelayandlossperformance[5]f?cengineeringproblemthatweconsiderismoti-vatedanetwork,notallthetraf?aybemultiplecontrollersfordifferentpartsofthequestiontheniswhetheritisstillpossibletodoeffectivetraf?cengineeringwhenallthetraf?paper,weconsidertraf?cengineeringinthecawheretofthenetworkdoeshop-by-hoproutingusingastandardroutingprotocollikeOSPF.978-1-4673-5946-7/13/$31.00 ?2013 IEEE2211

2013 Proceedings IEEE INFOCOM2TheobjectiveofthepaperistodevelopaSDNdeploymentschemethatcanadaptivelyanddynamicallymanagetraf?cinanetworktoaccommodatedifferenttraf?ncontributionsinthispaperarethefollowing:?Flex Node2512116???Toourknowledge,thisisthe?rstpapulatetheSDNcontroller’soptimizationproblemandoutlineaFullyPolynomialTimeApproximationScheme(FPTAS)byanalysisandsimulatonstheconsiderablegainsindelayandlossperformanceevenw?xednumberofSDN-FEsweoutlineanalgo-rithmfordeterminingtheplacementoftheforwardingelements.317131014Flex Node4Controller9815Flex wardingElementTheSDN-FEsperformthefollowingfunctions:Forwarding:earemultiplenexthopsforagivendestination,thentheSDN-FEscansplittraf?ctothedestinationinapre-speci?lativelyeasyforthecontrollertocomputemultiplenexthopsandloadtheroutingtabletotheSDN-FEs[12].Thereareveralwaysofsplittingtraf?conmultiplenexthops[15]whileensuringthatagiven?theapproachesneedextrameasurementsanditisrelaoreintherestofthepaper,weassumethetheSDN-FEscansplittraf?assumethattheSDN-FEscanperformtraf?cmeasurementsusingtechniquessuchasthoin[15].Measurement:TheroutingtableattheSDN-FEsismodi?edslightly,comparedtoastandardroutingtable,inordertoaidtraf?aticreprenta-tionshowingthedifferencebetweenastaatthereisanextraacketisprocesdbytheSDN-FEitdoesalongestpre?incredoneinordertodeterminetheamountoftraf?thatnode15(IPaddress45.67.2.5)e11withinterfaceIPaddress43.2.34.7,umncorrespondingtothetraf?ctracksthenumberofbytesroutedfromnode2tonode15forthedestinationpre?syfortheSDN-FEtoalsocomputethetotaltraf?pacelimitations,wedonotshowextensionstotheapproachinthispapertothepDESCRIPTIONWeconsideranetworkwhereacentralizedSDNcontofthenom-FEsforwardpacketsandthelogicforctiontoforwardingpackets,theSDN-FEsdosomesimpletraf?trollerusthistraf?cinformationalongwithinformationdisminatedinthenetworkbyOSPF-TEtodynamicallychangetheroutingtablesattheSDN-FEsinordertoadapttochangingtraf?trolleralsoexploitsthefactthatitcanmakeco-ordinatedchangesacrossthedifferentSDN-FEsthatitcontrolstomanagechangingtraf?ularnodesinthenetwork,alsoreferredtoasthenon-SDN-FEs,bridnetworkscenariowithtraditionalnetworksintermixedwithSDNsisalsoconsideredin[11].Ourobjectivesareconsiderablydifferentfromtheobjectivesin[11].Weassumethatnochangesaremadetothenon-SDN-FEs-FEs2,9,uthisnetwomethatalltualalgorithmthatarerunatthecontrollerisoutlinedinSection3.2212

2013 Proceedings IEEE INFOCOM3PrefixNodeNext

HopTraffic135.23/trollerTheSDN-Chasalltheroutinglogicanditcoordinatestrollerdoesthefollowingfunctions:Peering:TheSDN-Cpeerswiththeothernodesinthenet-workexchanginglinkweightsandothertopologyinformationusingOSPF-TE.(See[9]foranexample.)NotethatinOSPF-TE,thenodesorethecontrollerknowsthecurrentOSPFweightsaswellastheamountoftraf?c?owoneachlink(averagedoversometimeperiod).RouteComputation:Thecontrollerisresputestheroutingtablestakingintoconsiderationtheroutingdonebynon-SDN-FEs(badonOSPFlinkweights),thetraf?catthelinks(derivedfromOSPF-TEinformationor)andthecurrenttraf?cpattern(inferredfromthemeasurementsattheSDN-FEs).ThealgorithmforcomputingtheroutingtablefortheSDN-FEshastoensurethatroutingwidescribetheSDNcontroller’CONTROLLER’SPROBLthatthenet?NdenotethetofSDN-FEsandD=(e)andc(e)denotetheOSPFlinkweightandcapacityrespectivelyofalinke∈(e)toreprentthetraf?c??owonalllink∈sdtoreprentthetraf?cratefromnodes∈Ntosomeothernoded∈NandWudtoreprentthetotalamountoftraf?cfordestinationd∈NthateitheroriginatesorpassthroughSDN-FEu∈atingeneralWud≥-FEucanmeasureWudforueofTsdforallnodepairs(s,d)tingtableatnodeu∈H(u,d)rwords,NH(u,d)isthe?emainderofthispaper,weassumethatthenexthopisuniqueforallthenon-SDN-FEs,i.e,NH(u,d)hasonlyoneelementforallu∈hniquesinthispaperextenddirectlytothecawheretherealternateshortestpathsbetweentwonodesandtraf?atwhileNH(u,d)iscomputedbadonshortestpathsforallnodesu∈D,NH(u,d)canbetarbitrarilywhenu∈methatalllinkwethetreethatthatnodes2,9,atNH(6,13)=10,NH(1,13)=temple,node2cansplitthetraf?ctonostPathTreetoNode13De?nition1:GivenatofSDN-FEnodesC,apaths=u0,u1,u2,...uk=dfromasourcenodestoades-tinationnodedwillbetermedfeasibleifforj=1,2,...,k,(uj?1,uj)∈Eanduj=NH(uj?1,d)ifuj?1∈blepathwhereu0,u1,...,ede?nition,notethatapathisfeasibleifthenexthoptoagivendorewehavetoensurethatallthetraf?cbetweensanddhastoberoutedonP∈mple,inFigure3,3?2?5?12?atthisisnottheshortestpathwhichis3?2?11?h3?6?11?13isnotadmissiblesincethenexthopfornode3whichisanon-SDN-FEhastobethenexthopontheshortestpath,whichisnode2.2213

2013 Proceedings IEEE INFOCOM4De?nition2:Givenshortestpathroutingatthenon-SDN-FEs,traf?cthatgoesfromsourcetodestinationwithouttran-sitingthroughaSDN-FEwillbereferredtoasuncontrollabletraf?ourceofapacketisaSDN-FE,orifitpassthroughatleastoneSDN-FEbeforeitreachesitsdestinationthenthistraf?cwillbecalledcontrollabletraf?rwords,controllabletraf?ccomprisofpacketsthatpasatleastanopportunityattheSDN-FEstomanipulatethepathofcontrollabletraf?mplethetraf?cfrom6to13isroutedbyOSPFalong6?10?13,andsinceneither6nor10areSDN-FEs,traf?rast,traf?cfromnode8to13passthroughnode9whichisaSDN-FEandhencethistraf??nition3:WesaythataSDN-FEu∈Cinjectsapacketif??ationoftheSDN-C’sProblemSincetheonlytraf?cthatwecanmanipulateisthetraf?cthatpassthroughtheSDN-FEs,wejustfocusonthetraf??cIudisinjectedbySDN-FEu∈nlydosoalongoneoftheadmissiblepathsP∈(e)denotetheuncontrollable?owonlinke.(Notethatg(e)iasytocomputedifthesource-destinationtraf?edearlier,r,westillgivetheformulationbelowinordertomotivatetheactualdynamicroutingproblemsolvedbytheSDN-C).TheobjectiveoftheSDN-Cistoroutethecontrollabletraf?ayandpacketlossatthelinksareincreasingfunctionsofthelinkutilizationandthereforeweuthelinkutilizationasasurrogateforthedelay/uralobjectiveforormulation,thevariablesarex(P),whichisthe?henumberofpathsinthenetworkcanbeexponentialinthenumberofnodesandarcs,erthispathtothemorecompactnode-arcformulationsinceitlend-Csolvesthefollowingoptimizationproblem:f?cthatisinjectedbySDN-FEu∈Ctosomedestinationnoded∈ore,forallcontrollabletraf?cthereisauniqueSDN-FEthatinjectsthistraf?attheSDN-FEmayormaynotbethesourceofthetraf??gure,thenumbernexttothenodereprentsthetraf?mple,thetraf?cfromnode1tonode13(T1,13)?nition3,notethatthetraf?aluesofTsdareknownforallsource-destinationpairs(s,d),thenthevalueofIudcanbecomputedasfollows:RemovethelinksgoingoutoftheSDN-FEsanf?cthataccumulatesattheSDN-FEsisthetraf?mpleinFigure3,I2,13=9,I9,13=13andI14,13=edearlier,ymeasurementsavailableattheSDN-CarethevaluesofWudwhichisthetraf?cfordestinationdthatpassthroughnodeu∈C.12251minimizeθsubjecttog(e)+??P:P??ex(P)x(P)x(P)≤≥≥θc(e)?e∈EIud?u∈Cd∈N0?P(1)(2)(3)??P∈ndentlyRoutableTraf?cattheSDN-FEsThe?rsttofinequalitiensurethatthetotal?owonthelinkwhichisthesumoftheuncontrollable?ow(reprentedbyg(e))andthecontrollable?ow(whichisthecondsumtermontherighthandside)islessthantheproductofthemaximumlinkutilization(θ)andthecapacityofthelink(c(e)).?Thecondtofinequalitiensuresthatthetotalinjectedtraf?cisroutedinthenetwork.?Thethirdtofinequalitiensuresthatthe?imumvalueofθatiftheoptimumvalueofθ<1,eSDN-Csolvesthisoptimizationproblem,itiasytocomputethenexthoormulationabove,weassumedthatvaluesIudandg(e)ityboththequantitiesIudaswellasg(e)havetobecomputedbytheSDN-Cbadon?2214

2013 Proceedings IEEE INFOCOM5themeasingIudForeachdestinationd∈etheroutingorderR(d)?rstnodeuinR(d)tIud=neunitof?owfromutodandtβv(u,d)bethefractionofthisunit?owthatreachesnodev∈hsuccessivenode??winR(d)SetIwd=Wwd?u?dwβw(u,d)neunitof?owfromwtodandcomputeβv(w,d)forallv∈evaluesofIudareknownforallu∈Cforalld∈N,weuthistocomputethevaluesoftheg(e)whichistheuncontrollabletraf?cthat?owsonlinke∈doneasfollowsWeinjectoneunitof?owatnodeu∈Cfordestinationd∈Nandcomputesαe(u,d)whichisthefractionofthisunit?eknowIudforu∈C,wecancompute????αe(u,d)Iud?e∈E.g(e)=f(e)?u∈ingIudandg(e)DynamicallyTheonlymeasuredquantitiesthatareavailabletotheSDN-Care??Thelinkloadf(e)foralllink∈ntitiesWudforallu∈Cforalld∈hetwoquantities,theSDN-Chastocomputethevaluesofg(e)foralle∈EandIudforallu∈Cforalld∈?era?sallthenexthopsforallnodesinDandatalltheSDN-FEsitnotknowsallthenexthopsforthedestinationandthetraf??nition4:Givenadestinationdandthecurrentroutinginthenetwork,theroutingorderofthenodesinCwithrespecttothisdestinationdisde?nedasanorderingofthenodesinCdsuchthatifu∈Cappearsbeforev∈Cinthislistthenthereisnotraf?tetheroutingorderfordestinationnodedasR(d)andthefactthatuappearsbeforevinR(d)asu?utingorderiswellde?nedforanydestinationnodedsincetherecannotbeanyroutingloopsinthetraf?c?owingtodestinationd.(InfactitispossibletoorderallthenodesinthenetworknotjustthenodesinC,butweareinterestedonlyintheorderingofthenodesinC.)canotethatthereistraf?ore9?tingorderis(2,9,14).Otherorderingsarepossiblebutnode14shouldappearafternode9inanyordering.521211TheSDN-CnowknowsthevaluesofIudforallnodesu∈Cforalld∈Naswellasthevaluesofg(e)foralle∈atingtheDynamicRoutingProblemTheSDN-Croutestraf?valentproblemthatismoreconvenienttosolveistokeepthecapacitiesofthelink?xedbutscaletheinjectedtraf?csothatitstill?oblemisthefollowing:maximizeλsubjectto??P:P??ex(P)x(P)x(P)≤≥≥c(e)?g(e)=b(e)?e∈EλIud?u∈Cd∈N0?P(4)(5)(6)??P∈raptimalλ>1thenthecurrenttraf?ccanberouteattheoptimalsolutiontothisscalingprobleeofthefactthattheproblemhasanexponentialnumberofvariables,wecansolvertowritetheduallinearprogramtothedynamicroutingproblemshownabove,weassociatedualvariablesl(e)witheachlinkcapacityconstraint(4)andzudforthe2215

2013 Proceedings IEEE INFOCOM6demandconstraints(5).Thedualcannowbewrittenas??e∈Eminimizesubjectto??e∈Pb(e)l(e)l(e)≥≥≥zud?P∈Pud?u∈C?d(7)10?e∈E.(8)(9)inphaswhereineachprimalpha?owisroutedtoagivendestinationfromallSDN-FEsalongthelightestpermissiblepath(usingthedualvectorl(e)asthelinkweight).OnceeachSDN-FEuhasshippeda?owofIudtodestinationd,the(dual)weightsofthearcsonwhich?ocessofaugmenting?owaorithmisgiveninmoredetailbelow:AlgorithmCOMPUTETHROUGHPUT:DL←0l(e)←δ/b(e)e∈ERsd←0?(s,d)whileDL<1doforeachdestinationd∈Nd??(u)=Iud?u∈CwhileDL<1andd??(u)>0forsomeu∈CdoPud:Shortestadmissiblepathusingl,?uwithd??(u)>0c=mine∈∪sPsdb(e)ρ(e)istheutilizationofe∈Eρ=max{1,maxe∈∪uPudρ(e)}f(u)=min{d??(u),c}?uu)Routef(ρ?owfromeachutod.u)d??(u)=d??(u)?f(ρu)Rud=Rud+f(ρl(e)=l(e)(1+??ρ(e))??RecomputeDL=e∈Eb(e)l(e)endwhileendforendwhile????u∈Cd∈NIudzudl(e)Assumethatwetl(e)tobetheweightoflinke∈E.(WeuthetermweightinsteadofcostinordertoavoidconfusionwiththeOSPFlinkcosts).Notefromthe?rsttofconstraintszudisthelightestpathfromutod.(AgainweuthetermlightestpathtoavoidconfusionwiththeshortestpathusingOSPFcosts).LetLuddenotethelightestpathfromutodusingthelinkweightsl(e)lcannowbere-writtenas??e∈Eminimizesubjectto????u∈Cd∈Nb(e)l(e)IudLudl(e)≥≥10?e∈E.(10)(11)Inotherwords,givenanynon-negativetoflinkweights??bE(e)l(e)??isanupperboundonthel(e),notethat??e∈u∈Cd∈utlinethesolutionofthedynamictraf?gtheDynamicRoutingProblemWeuaFullyPolynomialTimeApproximationScheme(FPTAS)sonforsolvingtheproblemasanFPTASinsteadofastandardlinearprogrammingproblemisthattheFPTASisverysimpletoimplementandrunssigni?cantlyfasterthanageneraSprovidesthefollowingperformanceguarantees:forany??>0,thesolutionhasobjectivefunctionvaluewithin(1+??)-factoroftheoptimal,andtherunningtimeisatmostapolynomialfunctionofthenetworksizeand1/??.maldualalgorithmforourproblemworksasfollows:Thealgorithm?rstcomputesavalueδthatisafunctionofthedesiredaccuracylevel??,lweightofeachedgee∈Eisinitializedtol(e)=b(δe).Theprimaldualalgorithmoperatesudλ=minRTudOutputλThenextresultgivestherunningtimeofthealgorithmandtheproofofthisresultisalmostidenticaltotheoneinKarakosatas[7].Theorem1:Set????111?????δ=1???m(1+n??)??thentherunningtimeoftheprimal-dualalgorithmisO(???2m2logO(1)m)marks:orithmfollowsinthesameveinasKarakostas[7].Thecorrectnessofthealgorithmaswellastherunningtimrehoweversomekeydifferencesintheimplementationofthealgorithm.2216

2013 Proceedings IEEE theKarakostas[7]paper,criticallyimportantforussincewecancomputetheroutingfrheroutingatthenon-SDN-FEsisbadonthedestination,iteration,wehavetocore6,wegivetheOSbersnexttothelinksreprentsthedualweights(notOSPFcosts).thatwearenowcomputingthelightestadmissiblepathstonode13fromalltheSDN-FEs.(2,9,14).SincetheadmissiblepathshavetoutheOSPFshortestpathatallthenon-SDN-FEs,odatthearcwith0.8fromnode2tonode13reprentsthepath2?11?13andthearcwithweight0.4fromnode2tonode13isthepath2?5?12?isnewgraphisformed,theligucedgraphisshownexplicitlyonlyforillustrativeearemaximizingthethroughputofthenetwork,thduetothefacr,mal-dualalgorithmtypically?dagainstloopsintheoptimalsolution,wepostprocesstheoptimalsolutionusingabreadth-?ningtimeofdGraphforLightestPathtoNode130.10.511120.330.560.3130.2FElocationshavebeendetermined,theSDN-Csolvesthedynamicroutingproblemperiodicallytodeterminetheroutingoftraf?cattheSDN-FEsbadonthetraf?ticallyanysystemwithSDN-FEswilloutperformasysteticetheimprove-mentinimumlectionofnodeswilldependonthetraf??rstistopickthenodesindependentofthetraf?cmatrixandthecondapproachistousomeestimateofthetraf?mptedmethatweknowthenumberofSDN-FEsinthenetworkandwearegivenatentativetraf?cmatrixTwhereTsdisthetraf?cbetweennodess∈Nandd∈ualtraf?ccan,andingeneralwill,deviatefromthistraf??nition5:Thethroughputofatraf?cmatrixToveratofSDN-FEsCisde?nedasthelargestscalarλsuchthatλtethethroughputvaluebyλ(T,C).NotethatifC=?correspondstotewhereC=ede?nition,itiasytoeλ(T,?)≤λ(T,C)≤λ(T,N)?T,tethatλ(T,N)canbecomputedbysolvingastandardmaximumconcurrent?oregivenTandf,theobjectivethenistodetermineC:|C|=h0.5170.20.3100.20.31440.580.190.3150.50.3maxλ(T,C).INGTHELOCATIONOFSDN-FESGivenanetworktopology,the?stepofthealgorithm,we2217

2013 Proceedings IEEE INFOCOM8115nodesExodusAbovenet0.9thevaluesofTsdareknown,thethroughputproblemcanbewrittenasNormalized

Throughput0.8maximizeλsubjectto??P:P??e0.70.60.5x(P)x(P)x(P)≤≥≥c(e)?e∈EλTsd?s∈Nd∈N0?P(12)0.4??(13)(14)Fig.8.0.312345678P∈Psd9122Number of Flex NodesEffectofNumberofSDN-FEsSincetheproblemisstructurallythesameastheSDN-Csproblem,MENTALRESULTSWerantwogroupsofexperimentstochecktheeffectivenessofthealgorithmusingthefollowingthreetopologies:(i)The15nodetopologyshowninFigure1,(ii)TheExodus(Europe)pologyhas22nodesand74links,(iii)15nodetopologyallthelinkcaROCKETFUELtopologies,thelinkweightsaregivenandthelinkcapacitiesareassumedtobetheinverofthelinkcosts.(Wehaveconsolidatedmultiplelinksbetweentwonodesintoasinglelinkintheexperiments).?rstisthestaticperformancePerformanceMeasurementTheexperimentswereperformedtocomputetheexpectedperformanceimprovementduetothedynamicroutingalgo-rithmifwearegivenatraf?heplotsforstaticperformancemeasurement,malizedthroughputforagiventofSDN-FEsCisde?nedasλ(T,C).λ(T,N)atforOSPFroutingwithnoSDN-FEsthevalueofC=?.Weperformedtwotsofexperiments.?NormalizedThroughputversusNumberofSDN-FEsForallthreetopologies,emstobethereasosubquentexperiments,the?numberofSDN-FEsforthe15nnessoftheChoiceofSDN-FEsInthecondtofexperiments,wetestedthensitivityoftheperformancewithrespecttotherealtraf?cmatrix(asoppodtothetraf?cmatrixthatisudtopicktheSDN-FEs).Inotherwords,sincethelocationoftheSDN-FEsis?xedassumingsometraf?cmatrix,wewantedtoehowwellthischoiceperformedifthetraf?ore,inthecondtofexperiments,we?xedtheSDN-FEsfortheExodustopologytofourand?xedtheirlocationbadonsomeestimatedtraf?cho20randomtraf?cmatricestochecktheperformanceimprovementthatwegetwiththedifferenttraf?eplotthenormalizedatthenormalizedthroughputofSDNroutingissigni?cantlybetterthanOSPFforalltheexperiments.0.9OSPF RoutingFlex Routing0.850.8Normalized

Throughput0.750.70.650.60.550.5111213Experiment manceforDifferentTraf?-2SimulationExperimentsHeretheobjectiveistomeasurelinkdelaysandkstateroutingprotocolwasmodi?edtoallowSDN-FEswhhofthethreenetworksv-eraltraf?cpatternsweregeneratedwitheachsourcending2218

2013 Proceedings IEEE INFOCOM915nodeEXP115nodeEXP215nodeEXP3ExodusEXP1ExodusEXP2ExodusEXP3AbovenetEXP1AbovenetEXP2AbovenetEXP3MaxLossSDNRouting0.0000.0000.0000.0000.0002.0001.00052.0000.000MaxLossOSPF415.000391.000320.000140.000105.000106.000449.000555.000552.00015nodeEXP115nodeEXP215nodeEXP3ExodusEXP1ExodusEXP2ExodusEXP3AbovenetEXP1AbovenetEXP2AbovenetEXP3MeanDelaySDNRouting0.0030.0050.0020.0700.0600.0530.0100.0140.013MeanDelayOSPF0.0110.0100.0150.1060.1060.1000.0210.0290.025TABLEICOMPARISONOFMAXIMUMLOSSOVERALLLINKSTABLEIVCOMPARISONOFMEANDELAYOVERALLLINKS15nodeEXP115nodeEXP215nodeEXP3ExodusEXP1ExodusEXP2ExodusEXP3AbovenetEXP1AbovenetEXP2AbovenetEXP3MeanLossSDNRouting0.0000.0000.0000.0000.0000.0400.0120.6120.000MeanLossOSPF13.54311.89413.2284.9064.6934.1068.20011.5526.670TABLEIICOMPARISONOFMEANLOSSOVERALLLINKSdif?shownhowimprovednetworkperformancecaingevenafewstrategicaemedoesnotinvolvemakinganypro-tocolchangesattheremainingnodesinthenetworkwhichroutetraf?lperformancemeasurementsaswellasns-2simulationsshowthatthemethodcansigni?cantlyimproveoverallnetworkthNCES[1],an,,,n,r,”Ethane:TakingControloftheEnterpri”,ACMSIGCOMMCCR,37(4):1-12,2007[2],ll,er,d,,Merwe,”DesignandImplementationofaRoutingControlPlatform”,NetworkedSystemsDesignandImplementation,May2005.[3],n,,,,n,”Nox:TowardsaNetworkOperatingSystem”,ACMSIGCOMMCCR,July,2008[4]er,ishnan,d,,Merwe,”TheCaforSeparatingRoutingfromRouters”,FDNA2004.[5],”O(jiān)peningAddress:2012OpenNetworkSummit”,April2012.[6]”NetworkDevelopmentandDeploymentInitiative(NDDI)”,/network/o/.[7]stas,”FasterApproximationSchemesforFractionalMulti-commodityFlowProblems”,PrentedatACM-SIAMSODA2002.[8].,”O(jiān)nix:ADistributedControlPlatformforLargeScaleProductionNetworks”,OSDI2010,October,2010.[9]ento,berg,or,,na,aes”VirtualRoutersasaService:theRouteFlowapproachleveragingSoftware-De?nedNetworks”,CFI2011.[10].,”Network-WideDecisionMaking:TowardaWafer-ThinControlPlane”,HotNets-III,November2004.[11]nberg,,”RevisitingRoutingControlPlatformswiththeEyesandMusclesofSoftware-De?nedNetworking”,ACM-SIGCOMMHotSDNWorkshop,2012[12]an,opal,,i,,”TheSoftRouterArchitecture”,ProceedingofHotnets2004,November2004.[13]n,on,ishnan,ar,on,d,r,,”O(jiān)penFlow:EnablingInnovationinCampusNetworks”,ACMSIGCOMMCCR,April,2008.[14]TheOpen?owSwitch,open?[15]ran,,,”AchievingNear-OptimalTraf?cEngineeringSolutionsforCurrentOSPF/IS-ISNetworks”,IEEE/ACMTransactionsonNetworking,V.13,No.2,April2005.[16],,”O(jiān)ptimizingOSPF/IS-ISWeightsinaChangingWorld”,IEEEJournalonSelectedAreasinCommunications,traf?entedswereudtogeneraterandomtraf?15noIandIIshowthemaximumnumberofpacketslostoverallthelinksandthemeannumberatSDNRoutingdoessigni?eistrueforthemaximumdelay(TableIII)andmeandelay(TableIV).ThedifferenceinpSIONIncrementalSDNdeploymentwhereSDNsco-existwithtraditionaofparticularimportanceforlargenetworkswherecompletegreen?elddeploymentisMaxDelaySDNRouting0.1190.1530.0600.9130.7570.7170.2280.2880.297MaxDelayOSPF0.2830.2850.2771.6451.7321.7120.2420.3960.49615nodeEXP115nodeEXP215nodeEXP3ExodusEXP1ExodusEXP2ExodusEXP3AbovenetEXP1AbovenetEXP2AbovenetEXP3TABLEIIICOMPARISONOFMAXIMUMDELAYOVERALLLINKS2219