JIANetal.:BETA/M/1MODELFORMACHINETYPECOMMUNICATION587 out.Numericalanalysisshowsthatifthearrivalpattern followsBetadistribution,itcouldpotentiallyincreasethe meansojourntimeandmeanwaitingtimeofthesystem.And increasedmeanserverrateordecreasedserverutilizationmay helptosolvetheproblemthoughthisisnotwhattheoperators want.Beyond3GPPsproposals,ifwewanttochoosean appropriateBetadistributionfordifferentMTCapplications,it isbettertoletα<βanddetermineareasonablegapbetween αandβ. APPENDIXA BETADISTRIBUTION TheprobabilitydensityfunctionoftheBetadistribution,for x∈[0,1]andshapeparametersα>0andβ>0,isapower functionofthevariablexanditsre?ection1?x: Fig.3.NumericalsolutionofBforgeneralBeta/M/1model. Γ(α+β) α?1β?1 f(x)=x(1?x)(9) A Γ(α)Γ(β) whereΓ(z)istheGamafunction.Themeanandvarianceof variancetoestimateBetadistribution’sshapeparameters,these xis: conclusionscanbeusedasapreliminaryguidetochoosethe ααβ appropriatemodelfordifferentMTCapplicationscenariosbe- E[x]=,Var[x]=(10) 2 α+β(α+β)(α+β+1) yond3GPPTR37.868andpromptlyjudgetheeffectivenessof thechosenmodel.AsBeta(1,1)equalsuniformdistribution REFERENCES andisusedtomodelthescenariothatMTCdevicesaccess [1]3GPP,StudyonFacilitatingMachinetoMachinecommunicationin3GPP networkinanon-synchronizedmanner,intuitivelyifMTC network,TR22.868,Mar.2007. devicesaccessthenetworkinahighlysynchronizedmanner, [2]3GPP,StudyonprovisionoflowcostMTCUEsbasedonLTE,TR itsmeansojourntimeandmeanwaitingtimewillbelarger, 36.888,June2012. [3]T.TalebandA.Kunz,“Machinetypecommunicationsin3GPPnetworks: thereforethecurveofBwillbeonthetop-leftofthatof potential,challenges,andsolutions,”IEEECommun.Mag.,2012,pp. Beta(1,1).Thismeanswhenchoosingtheshapeparameters 178–184,Mar.2012. ofBetadistributionforMTCapplications,itisbettertolet [4]3GPP,StudyonRANImprovementsforMachineTypeCommunication, TR37.868,Sept.2011. α<β.Inaddition,ifthedifferencebetweenαandβis [5]M.Zukerman,IntroductiontoQueueingTheoryandStochasticTeletraf?c larger,tosomeextent,itrepresentsmoresynchronizedaccess Models,2000.Available:ee.cityu.edu.hk. behavior. [6]S.M.Ross,IntroductiontoProbabiltyModels.Elsevier,2010,pp.497– 568. [7]O.C.Ibe,MarkovProcessforStochasticModeling.Elsevier,2008,pp. IV.CONCLUSION 105–148. [8]J.W.Cohen,AppiedMathematicsandMechanics-TheSingleServer IncontextofMTCapplications,Beta/M/1modelwouldbe Queue.NorthHolland,1986,pp.159–329. amoreappropriatetraf?cmodelthantraditionalM/M/1model[9]L.ShidaandL.Shishi,SpecialFunction.ChinaMeteorologicalPress, 1988,pp.537–551. torepresentthesynchronizedaccessbehaviorofMTCdevices. [10]S.ZhangandJ.Jin,ComputationofSpecialFunctions.NanJing ByassumingtheshapeparametersofBetadistributionto UnivercityPress,2011,pp.208–215. beintegerandusingtheavailablesolutionofG/M/1mode,[11]A.Amokrane,A.Ksentini,etal.,“Congestioncontrolformachinetype communication,”inProc.2012IEEEICC. theperformanceanalysisofBeta/M/1modelcanbecarried |
|