eventuallywcdma联通

QoS-awarePowerControlandHandoffPrioritizationin3GWCDMANetworksTajje-eddineRachidi,AmalYasmineElbatji,MehdiSebbane,andHichamBouzekriAlakhawaynUniversityinIfrane,PO.
Box1881,Ifrane53000,MoroccoT.
Rachidi@alakhawayn.
maRésumé—L'intégrationdanslessystèmesdecontrledepuissanceenboucleferméedanslesréseauxde3èmegénérationàbaseduWCDMA(telqueleUMTS)desparamètresdeQualitédeService(QaPC)peuxrésulterenunemeilleuredistributiondesressources(notammentradio)auniveaudesstationsdebase.
Delamêmemanièrel'intégrationdesparamètresQoSdanslaprioritédeshandoff(QaHO)peuxaussirésulterendemeilleureperformancecomparéauxtechniquesditesaveugles.
Cetarticleprésentedeuxtellestechniques(QaPCetQaHO)quiutilisentlaclassedeservice,ledébitainsiqueledescripteurdedégradationdeservice(SDD)[Lataoui2000].
Lesperformancesobtenuespourcesdeuxtechniquescombinéesàl'aidedusimulateurdéveloppéen[Elbatji2003]sousdiversesconditionsdecharges,detraficetdestratégiesd'admissionsontprésentées.
Lesrésultatsmontrentquecestechniquescombinéesaméliorentl'utilisationdesressourcesradiopar22%réduisentleblocagelorsdeshandoffspar12%.
Abstract--QualityofServiceawarepowercontrol(QaPC)mechanismsthatsupersedetothetraditionalclosedlooppowercontrolinWidebandCodeDivisionMultipleAccess(WCDMA)typeofnetworks,suchastheUniversalMobileTelecommunicationSystem(UMTS),providesignificantadvantagesoverblindchannelestimationmechanisms.
ThesemechanismsintegratespecificQoSrequirementsofusersinpowercontroldecisionyieldingoptimaluseofresourcesavailableatthebase.
SimilarlyQoSawareprioritizationofhandoffs(QaHO)thatleverageuserQoSprofilecanalsoyieldsignificantimprovementsoverblindprioritization.
ThispaperpresentstwosuchQaPCandQaHOmechanismswhicharebasedontheclassofservice,thebitrate,andtheServiceDegradationDescriptor(SDD)[Lataoui2000]asenablingQoSparameters.
TheperformanceofourcombinedQaPCandQaHOmechanismsobtainedusingthetestbeddescribedin[Elbatji2003],underavarietyofloadandtrafficscenarios,andadmissionstrategiesisalsopresented.
Theresultsshowthat,whenmeasuredagainstblindmechanisms,thecombinedQaPCandQaHOsignificantlyimprovescontractupholdingofpremiumservicemobileusers,aswellasimproveresourceutilizationbymorethan22%whileimprovinghandoffsfailuresby12%.
Keywords--QoSprovisioning,MultimediaQoSsupport,Closedlooppowercontrol,HandoffprioritizationI.
INTRODUCTIONTherationalbehindtheUniversalMobileTelecommunicationSystem(UMTS)evolutionisthedeliveryofmultimediaservicescharacterizedbystringentrealtimerequirements,greatsensitivitytodeliverydelayandpacketloss,andtheneedforconsiderablewirelessresources.
UMTS,therefore,supportsQoSprovisioningthroughfour(4)basicclassesofservice[3GPP2002a,3GPP2002b]:Class1:Conversational(highsensitivitytodelayandjitter);Class2:Streaming(mediumsensitivitytodelay,andhighsensitivitytojitter);Class3:Interactive(lowsensitivitytodelay,highsensitivitytoroundtripdelaytimeandBitErrorRate(BER));Class4:Background(nodelaysensitivity,highsensitivitytoBER).
EachoftheseclassesimposesdifferentQoSrequirementontheUMTSnetworkwhichmustbemaintainedduringthelifetimeofthecorrespondingconnections.
ProvisioningQoSoverWCDMA-basedairinterfacecannotbefulfilledsolelybyproperAdmissionControl[Zheng1997]andefficientScheduling[Wong2003].
Thisisdue,ontheonehand,totheinherentcharacteristicsofthewirelesslink[Forman1994,Satyanarayanan1996],thatis,usermobilityandfadingchannel(timevariations)[Rappaport2001,Stuber2001,Proakis2000],higherrorrates,inherentinterferencelimitedcharacteristicsofWCDMA[Ericsson2001,Dahlman1998],andlowandvaryingbandwidth(2Mbpsatmost);andtotheunexpectedSoftHandoffs(SHOs)requestsontheotherhand.
Theformereffectshavebeen,untilrecently,cateredforusingclosedlooppowercontrolmechanisms,thatoperatesolelyonthebasisofchannelgain,butthatarenotawareofQoSrequirementsofunderlyingconnections.
Thisblindmodeofoperationdoesnotnecessarilyyieldoptimalpowerutilization,especiallywhenothernonpremiumconnectionsinthesystemarewillingtobedegraded;thatis,theyarecapableofadaptation,andwillingtohavetheirrequiredbitrate/powerreduced.
Whilethelater,thatisunexpectedSHOs,havebeentackledusingeitherreservationorpredictiontechniques[Soh2003].
Ourworkaimsatshowingthatuserwillingnesstobedegradedcanbeusedtoaugmentbothtraditionalclosedloopcontrolmechanismforcongestion(theeffectsinherentwirelesslinkeffectsdescribedabove)handling,aswellas,toimprovehandoffbyreducingtherateofdroppingofSHOrequests.
ALucentpatentedframeworkformodelinguserwillingnesstobedegradedasanewQoSparameterhasbeenpresentedin[Lataoui2000].
Therein,theServiceDegradationDescriptor(SDD)isanumberbetween0and5;thelargertheSDDisthemorewillingistheuser/connection1tobedegraded,andeventuallydropped.
WeadoptSDDinthisworktoo,anduseittogetherwiththeserviceclass(1-4)andthebitrateasenablingQoSparametersforthemechanismsweseektodevelop,thatisasystemthatcombinesbothQoS-awareclosedlooppowercontroltechniquesandQoS-awareprioritizationofSHOsinWCDMA-based3Gnetworks.
WealsoseektoquantifythebenefitsofthecombinationofourtwoQoSawareschemasfromtheperspectivesofboththenetworkprovider(thatisresource/bandwidthutilization)andtheuser(forcedterminations,andrateofacceptanceofSHOs).
Specifically,weaimatbuildingmechanismsthat1.
copewiththeinherentcharacteristicsofthewirelesslink,and2.
minimizetheprobabilityofdroppingofSoftHandoffs(SHOs),while3.
maintainingQoSrequirementsofactiveconnections,and4.
achievinghighsystemutilization.
1Thewordsuser,subscriberandconnectionwillbeusedinterchangeablyhereafter.
In[Bhatti1998],aQoSinformationmodelformakingadaptationdecisionsisdescribed,andin[Choukair2003]aruntimeadaptationofUMTSservicestoavailableresourcesispresented.
Nonetheless,toourknowledge,therehasn'tbeenmuchworkspecificallyonQoS-awareclosedloopsandQoS-awareprioritization.
Wehad,however,toucheduponthisissuein[Abid2001].
Furthermore,theEuropeanTelecommunicationsStandardInstitute(ETSI)specificationsforWCDMAairinterfacesuggestfiveactionstobetakenrespectivelyinthepresenceofcongestion[Ericsson2001,Dahlman1998],thatisfiveactionstocopewithlinkdegradation.
Theseare:Action1:congestioncontrolisactivatedwhichreducesthebitrateofnon-realtimeapplicationstodecreasethecongestionlevel.
Action2:ifthe1stactionisnotsufficient,congestioncontroltriggerstheinter-frequencyhandoverthatmovessomesubscriberstolessloadedfrequencies.
Action3:ifthe2ndactionfails,somesubscriberscanbehandedovertoadifferentoperator.
Action4:ifthe3rdactionapproachfails,somesubscriberswillbehandedovertoadifferentsystemsuchastheGlobalSystemforMobileCommunication(GSM).
Action5:consistsofblockingsubscribersoflowerprioritytoprotectthequalityoftheremainingones.
Actions1and5aim,specifically,atrenderingpowercontroldependentontheQoSrequirements.
Nospecificsaregivenhowever.
Furthermore,theseactionscanbeconsideredasboundaryconditionsofourmoregeneraladaptationstrategythataimsatredistributingsystemresourcesusingextrainformation,i.
e.
,userwillingnesstobedegraded.
Theremainderofthepaperisstructuredasfollows:Wefirstdescribecalladmissionandhandoffstrategiesweuse,thenwedescribethecombinedSDD-based(QaPCandQaHO)mechanism.
Wethenpresentthesimulationscarriedoutusingthetestbedof[Elbatji2003]forperformanceevaluation.
ResultsofthecombinedschemearecomparedtoablindmechanismforcongestionhandlingasspecifiedinUMTS[Choukair2003],whichdoesnotuseotherQoSattributesthantheclassificationoftheclassintorealtime(RT)(i.
e.
,classes1and2)andnon-realtime(NRT)(i.
e.
,classes1and2)asinactions1and5describedabove.
Finally,wepresentourconclusionsandfutureworks.
II.
ADMISSIONCONTROLANDHANDOFFSTRATEGIESUserrequestsareprocessedonaFCFSbasis.
ThedecisionofacceptingorrejectingarequestisbasedontheQoSprofileattachedtotherequestontheonehandandthemaximumavailablepower2inthesystemPmax.
Variousadmissionstrategiesareavailable:A.
StrictadmissionstrategyInthisstrategy,aconnectionnewisacceptedinthesystematinstanttonlyif∑Pi(t)+Pnew≤Pmax,wherePi(t)isthepowerrequiredbyexistingconnectioni,andPnewisthepowerrequiredbyconnectionnew.
B.
NRTOverloadadmissionstrategyInthisstrategythebase/systemisallowedtoacceptconnectionsevenif2Othersystemresourcessuchasspreadingsequencesandbuffersareassumedtoexistinsufficientnumbers/quantities.
thetotalpowerrequiredbyallconnectionsexceedstheavailablepower.
Inthiscase,NRTconnectionswillhavetobedelayedbythescheduler.
Specifically,aconnectionnewisacceptedinthesystematinstanttifandonlyifbothconditionshold:-∑Pi/RT(t)≤PmaxwherePi/RT(t)isthepowerrequiredbyexistingrealtimeconnectioni,(thatisclass1and2connectionsinthesystemincludingeventuallythenewconnection).
-∑Pi(t)≤(1+α)Pmax,where0<α<1indicatesthemaximumoverloadallowedforNRTconnections,andispansacrossallconnectionsincludingtheoneunderadmissiondecision.
Fig.
1showstheNewConnection(NC)Admissionhandlingprocess.
Aqueuedconnectioncanbedroppedfromthequeueifitreachesitstimeout(setto2unittimebydefault).
Fig1alsoshowsthatnegotiationofQoSrequirements(currentlybandwidthonly)takesplaceintheAdmissionControlentity.
IfnegotiationfailsandanNRToverloadstrategyischosen,theNCcanbeacceptedifitbelongstoaNRTclass,andreconsideredwhenpowerallowsso.
NCQueuingNCTermination/MovementTimeoutCongestionCheckNCBlockedCongestionPresentNoCongestionAcceptNCRejectPossibleNCNegotiationPossibleNRTOverloadTimeoutNoPossibleNRTOverloadNoPossibleNegotiationTimeoutNCNeededNCCongestionCheckNCCACDecisionNeededNCAcceptanceNCQoSNegotiationNCQueingNCTermination/MovementNCBlockedNCDroppedNCNRTOverloadFigure1.
Newconnection(NC)admissionhandlingSHOSHOTermination/QueuingMovementTimeoutCongestionCheckSHOBlockedCongestionPresentNoCongestionAcceptPossibleNRTOverloadSHOPossibleRejectSHOSHOCACReconsiderationNegotiationTimeoutNomoreAdaptationPossibleNoPossibleDegradeNegotiationBlockTimeoutNoPossibleNRTOverloadTimeoutSHONeededSHOQueingSHOCACDecisionNeededSHOAcceptanceSHOQoSNegotiationQoSAdaptationPowerAvailableSHOTermination/MovementSHODroppedSHOCongestionCheckSHOBlockedinQueueSHONRTOverloadFigure2.
HandoffrequesthandlingFig.
2showstheSoftHandoff(SHO)Admissionhandlingprocess.
ASHOundertakethesameprocessasanNC.
However,unlikeanNC,thereisnonegotiationandQoSAdaptation(seenextsection)istriggeredtoprovidetheSHOwiththenecessarybandwidthattheexpenseoftheexistingconnections.
III.
COMBINEDSDD-BASEDQAPCQAHOPROVISIONINGMECHANISMInourapproach,powerisconsideredtobetheonlylimitingresource.
Othersystemresourcesuchasspreadingcodes[Proakis2000]andbufferingcapacityareconsideredtobeavailableinsufficientquantities.
Thecostofaconnection(i)atagiventimeiscomputedaccordingtothefollowingformula[Mueckenheim2002]:Ci(t)=Eb/No.
1/w.
Ii(t)/Hi(t)(1)EnergytoNoiseratioEb/Noissetbydefaultto18dB,butcanbesettoadifferentvaluetoaccountforqualityofUserequipment.
Intercellinterferenceisnottakenintoconsiderationinthecurrentversionofthetestbed.
TheChiprate(W)issetto3.
84Mchips.
TheinterferenceatagiventimeIi(t)isthesumofinterferencesexertedbyexistingusersonthetargetuseratagiventimewithinthesamecell.
ThecentrallimittheoremisusedtomodelIi(t)asaGaussianprocesswithzeromeanandagivenvarianceσ2[Rappaport2001].
Theσ2wasinitiallysetto0.
5.
However,itcanbesettootherwisetoaccountformulti-path.
ThechannelgainatagiventimeHi(t)followsaRayleighdistribution.
Thisismodeledusingarandomprocessinthefrequencydomain[Rappaport2001].
Ci(t)isthecost(powerperbit)formaintainingconnectioniinaninterferencelimitedenvironement;itembodiestimevariationsofthechannel[Elbatji2003].
ThetotalaveragepowerrequiredbyconnectionioperatingatbitrateRiistherefore:Pi(t)=Ci(t)*Ri,(2)OurnewcombinedapproachcomplieswithWCDMA[Ericsson2001,Dahlman1998]and3GPPspecifications[3GPP2002a,3GPP2002b].
Therationalbehinditistoprovideabasisfor:1.
HandlingchanneldegradationintheWCDMAradioaccessnetworkbydynamicallytriggeringaQoSAdaptationAlgorithmthatsupercedestotheclosedlooppowercontrol,and2.
ProvidingtheincomingSHOrequests,whichwouldotherwiseberejectedbytheCallAdmissionController(CAC)duetolackofresources,withthenecessaryresourcesbytriggeringthesameQoSAdaptationAlgorithm.
Fig.
3describesourcombinedmechanismforhandlingcongestionandSHOs.
TheQoSadaptationalgorithmisattheheartofthecombinedmethod,andistriggeredtomakeroomforanincomingSHO,andinthepresenceofcongestion(thatisthetotalpowerrequiredbyexistingconnectionsislessthantheavailablepoweratthebase:ΣiPiAlternativelycongestionisalsodefinedasthelackofpowerforrealtimeconnections(RT)namelyforclass1andclass2connectionsonly.
Thatis,ΣiPiinRTItisworthmentioningthatbothmodesaresupportedinthetestbedusedforevaluation,andthatcongestionisdeclaredafter2unittime(ut)persistenceofcongestionsymptoms(lackofpower).
Thisconferstothecongestionhandlingprocessstabilitywithrespecttotemporaryshortfades.
A.
QoSadaptationalgorithmTheadaptationalgorithmresolvescongestionintwophases.
ThetwophasesareapplieddifferentlyincaseofcongestionhandlingandincaseofSHOadmission.
Inmanyways,itisanimprovementtothealgorithmsuggestedin[Abid2001].
InaccordancewiththeQoSframeworkdefinedin[Lataoui2000],eachconnectionrequestbytheUserEquipment(UE)includesaQoSprofile.
TheprofilecomprisestherequiredbitrateRi,thetrafficclassCLi,andtheServiceDegradationDescriptorSDDi.
Thelattertakesvaluesbetween0and5.
ThelargertheSDDis,themorewillingisamobileusertogetdegraded/dropped.
Figure3.
ProcessestriggeredtohandlecongestionandSHOsarebaesdonacoreQoSadaptationalgorithmwhichusesSDDQoSdescriptor,aswellasclassofserviceandbitrate.
1)TheDegradationPhase:ThisphaseissolelybasedontheSDD.
Iteratively,theactiveconnectionthathasthehighestSDDistheconnectionthatgetsdegradedintermsofitsbandwidthrequirementsasfollows:onesuchconnectionwith384Kbpsbitraterequirementwillbedegradedto144Kbps.
Similarly144Kbpsisswappedfor64Kbps,and64Kbpsisswappedfor16Kbps.
2Mbpsand16Kbpsconnectionsarenotdegradedinthisschema.
2)TheDroppingPhase:Thedroppingphaseisinvokedonlywhenwillingconnectionsweredegraded,butcongestionpersists.
Inthisphase,droppingisbasedon:Fi(t)=SDDi*Pi(t),(3)wherePi(t)isthepowerrequitedbyconnectioniattimet.
Fi(t)ishighforconnectionsrequiringmuchpowerandatthesametimemorewillingtobedegraded3.
Iteratingthroughclass4,3,2,then1,connectionswithhighFi(t)aredroppeduntilcongestiondisappears.
Toconferfairnesstothedroppingphase,connectionswithsimilarFi(t)areconsideredaccordingtotheircostCi(t)first,thenaccordingtotheirarrivalstime.
AsmentionedearliertheQoSadaptationalgorithm,justdescribed,isinvokedtoprovidethenecessarybandwidthforSHOrequeststhatnormallywouldnotbeacceptedduetolackofresources/power.
TwocasesaredistinguisheddependingontheclassofserviceoftheSHOrequest:-NRT:ifthedegradationofexitingconnectionsisnotenoughtocollectthenecessaryresources/power,theSHOrequestisrejected.
-RT:thenecessaryresources/powerwillalwaysbeprovidedbydegradationfirstthendropping.
Fromanimplementationpointofview,theQoSadaptationalgorithmdoesn'tdegradeordropagivenconnectionunlessitissurethattheuserundergoingSHOwillbeaccepted.
Actualdegradation/droppingofconnectionstakesplaceatomically.
B.
NumericalResultsandDiscussionTheevaluationoftheproposedcombinedapproachiscarriedoutusingtheWCDMAcompatibletestbeddescribedin[Elbatji2003].
Thistestbedallowsforavarietyofuser/connectionarrivalpatternswithUMTScompatibleQoSprofiles(classes,bitrates,speed,3Theconnectionthatismuchmorewillingtobedegradedistheconnectionthatshouldbefirstconsideredinthedroppingphase.
etc.
)aswellasSDD,tobeinjected.
Italsoallowsforavarietyofadmissionandcongestionsignalingstrategiestobesetup.
Usingthistestbedwebenchmarkourproposedcombinedapproach,againstabasiccombinednonQoS-awarecongestionhandlingmechanismthatconformsUMTSclassicaldropping[Choukair2003],andanonQoS-awareSHOprioritizationmechanism.
Actionsinthisbasiccombinedmechanism(BA)aretriggeredsolelybypower4availabilityregardlessofQoSattributesofcurrentconnections(seeFig.
4).
CongestionpersistenceSortconnectionsaccordingtotheirPowerDroptheconnectionconsumingthehighestpoweryesnoFigure4.
QoSprovisioninginBAmechanism.
SHOrequestsaretreatedlikenewconnections.
IterativelyconnectionsconsuminglargestamountsofpoweraredroppeduntilthetotalpowerofremainingactiveconnectionsbecomeslessthanPmax(themaximumavailabletransmitpowertoaUMTSBase).
Itisworthnoticingthatthecost(power/bit)isnotusedinBAeither.
Specifically,wemeasureaveragedroppingperclass,SHOacceptancerate,andaveragebandwidthutilizationforBAversuscombinedQaPCandQaHOundertwoloadscenarios:asteadyincrease(A),andsuddenincrease(B)asinFig.
4.
Tothisend,twoseriesofexperiments:seriesAandseriesB(54Othercellresourcessuchasspreadingsequencesandbuffersareassumedtobeavailableinsufficientquantities.
runsexactlyineachseries)werecarriedoutonaPentiumIII(728MHz)with128MBofRAMrunningWindowsXPusingthetestbed.
Theseexperimentsconsistinlaunchingthetestbedsimulationsfor300unittime(ut),correspondingto20minutesrealtime,foreachrun.
Thenaveragingallcollectedmeasurementsforeachseriesseparatelyoverthefiveruns.
TABLEI.
MAINEXPERIMENTALSETTINGPARAMETERS.
PhysicalLayerParametersE/N=18DBW=3.
84McpsPmax=35W(for100sofusers)Cmax=2.
5mW/bitTrafficParametersCL=1,R=2Mbps,CD=60ut,V=0km/hCL=2,R=384Kbps,CD=30ut,V=60km/hCL=2,R=144Kbps,CD=30ut,V=100km/hCL=3,R=64Kbps,CD=4ut,V=120km/hCL=1,R=16Kbps,CD=64ut,V=160km/hQueuingtimeout:2utforallconnections.
SDDrandomQoSadaptationtriggering2utcongestionpersistenceDegradation384Kbps→144Kbps144Kbps→64Kbps64Kbps→16Kbps.
2Mbpsand16KbpsconnectionsarenotdegradedCACstrategyNRToverloadis10%ofthetotalavailablepowerAdmissionnegotiationNoIneachrun,thetestebedisloadedwith100initialconnectionstobringittoaninitialclosetocongestionstate.
Pmax(seeTable1)aswellasotherphysicallayerparametershavebeencarefullychosentoyieldcongestionaround100connections.
Subsequentconnectionsarethrowninaccordingthefollowingtrafficmodels:callrequestsaregeneratedforseriesAaccordingtoPoissondistributionwitharateof2connections/utduringthe300ut.
AsforseriesB,aburstof5connections/utisgeneratedbetweenut50and100(seeFig.
4).
Theinitialpositioninthecellofanewcall,aswellasitsclassofserviceCL,anditsSDDaregeneratedrandomly.
Foreachcall,thebitrateR,thespeedV,andthecallduration(CD)areassignedaccordingtotheclass.
Forthepurposeofallsimulationsa10%overloadcorrespondingtoNRTtrafficisused.
Asummaryofthetrafficmodelaswell,physicallayermainparameters,andtestbedadmissionstrategyisgiveninTable1.
AlthoughFig.
5showsthearrivalpatternsforbothseriesAandB,duetospaceshortagewewillshowonlythegraphscorrespondingtoseriesA,thatissimulationunderasteadyarrivalpattern.
TheresultsobtainedforseriesBwillnonethelessbegiven.
0100200300400500600T-0T-25T-50T-75T-100T-125T-150Time(ut)steadyload(A)sharpload(B)Figure5.
ConnectionarrivalscenariosA:steadyandB:unexpectedsharploadat50ut.
Fig.
6summarizestheaverageconnectiondroppingperclassforthecombinedQaPCandQaHOversusBA,forseriesA.
InBA,premiumtrafficofclass2isheavilypenalizedwhencongestionoccurs,whilewiththecombinedQaPCandQaHO,premiumtrafficthatisclasses1and2experiencelessdropping,thusmaintainingQoSrequirementsforcriticaltraffic.
SimilarresultsareobtainedunderseriesB.
Moreover,asshowninFig.
7,animprovementof12%isobtainedforthecombinedQaPCandQaHOforSHOrequests.
Thisimprovementreaches19%forseriesB.
AsillustratedinFig.
8,combinedQaPCandQaHOgivesatotal22%moreresource/powerutilizationthanBAwithseriesA,andastaggering25%moreutilizationisobtainedforseriesB.
23,9743,4514,6117,970,5329,9734,5334,96051015202530354045501234Dropping%BAQaPC+QoHOFigure6.
PercentageaveragedroppingrateperclassobtainedforseriesA,forcombinedQaPCandQaHOversusBA.
QaPC+QaHO,98,86%BA,86,36%80828486889092949698100AcceptanceRate%Figure7.
SHOacceptancerateforcombinedQaPCandQaHOvs.
BAobtainedwithseriesA.
010000000200000003000000040000000500000006000000070000000110192837465564738291100109118127136145BandwidthConsumptionBAQaPC+QaHOFigure8.
ResourceutilizationcombinedQaPCandQaHOversusBAacrosstime.
TheresultsclearlyshowthatourcombinedQoSawarecongestionandSHOhandlingissuperiortothemechanismsuggestedin[Choukair2003],inallrespects.
IV.
CONCLUSIONANDFUTUREWORKSWehavepresentedaQoSawaremechanismforpowercontrolandHandoffin3GWCDMAnetworks.
Wehaveusedbitrate,serviceclassandServiceDegradationDescriptorasenablingQoSparameters.
NumericalresultsobtainedusingaWCDMA-andUMTScompatibletestbed,showthatourproposedQoSawaremechanismsignificantlyimprovesQoScontractupholdingforpremiummobileusers,aswellasincreaseresourceutilization,whileimprovingSHOacceptance.
CurrentinvestigationsarefocusingonintegratingBERandqueuelengthasextraenablingQoSparameterforourapproach;aswellas,evaluatingthismechanisminpresenceofdistributedadmissionstrategies.
REFERENCES[Abid2001]M.
R.
Abid,T.
Rachidi,A.
Bensaid,S.
Gruhl,andM.
Soellner"AdaptivefuzzycalladmmisioncontrollerforUMTS",the5thworldconferenceonSysmetics,CyberneticsandInformaticsSCI,Vol.
XVI,2001,pp.
93-99.
[Bhatti1998]S.
BhattiandG.
Knight"NotesonaQoSinformationmodelformakingadaptationdecisions"1998.
[Choukair2003]Z.
Choukair,andS.
Sfar,"RuntimeadaptationofUMTSservicestoavailableresources",Proc.
ofthe17thInt.
Conf.
onAdvancedInformationNetworkingandApplications,IEEE2003.
[Dahlman1998]E.
DahlmanandP.
Bening"WCDMA–theradioInterfaceforfuturemobilemultimediacommunications",IEEETransactionsonVehicularTechnology,Vol47,No4,November1998.
[Elbatji2003]A.
Y.
Elbatji,T.
Rachidi,andH.
Bouzekri"AtestbedfortheevaluationofQoSprovisioninginWCDMAbased3GwirelessNetworks",inProceedingsoftheIASTEDInternationalConferenceonCommunicationSystemsandNetworks,,CSN,September8-10,2003,Benelmadena,pp.
31-36[Ericsson2001]EricssonRadioSystem,"BasicconceptsofWCDMAradioaccessnetwork",2001,ericsson.
com.
[Forman1994]G.
H.
Forman,andJ.
Zahorjan,"Thechallengesofmobilecomputing,"IEEEComputer,vol.
27,no.
4,pp.
38-47,April1994.
[Lataoui2000]O.
Lataoui,T.
Rachidi,andL.
GSamuel"QoSmanagementarchitectureforpacketswitched3rdgenerationmobilesystems",inproceedingsofINTEROP,Mai2000,p.
365.
[Mueckenheim2002]J.
Mueckenheim,andS.
Gruhl,QualityofServiceSchedulingMethodforUMTSDownlink,LucentTechnologies.
Personalcommunication,2002.
[Proakis2000]J.
G.
Proakis,"Digitalcommunications",McGraw-Hill;4thedition,2000.
[Rappaport2001]T.
S.
Rappaport,"Wirelesscommunications:principlesandpractice"PrenticeHallPTR,2001.
[Satyanarayanan1996]M.
Satyanarayanan,"Fundamentalchallengesinmobilecomputing",inProceedingsoftheFifteenthAnnualACMSymposiumonPrinciplesofDistributedComputing,ACM,1996.
[Soh2003]Wee-SengSohandHyongS.
Kim,"QoSprovisioningincellularnetworksbasedonmobilitypredictiontechniques",IEEEcommunicationsMagazine,Vol41,No.
1.
,January2003,pp.
86-92.
[Stuber2001]G.
L.
Stuber,"Principlesofmobilecommunications",KluwerAcademicPublishers;2ndedition,2001.
[Wong2003]WilliamK.
Wong,HaiyingZhu,andVictorC.
M.
Leung"SoftQoSprovisioningusingthetokenbankfairqueuingschedulingalgorithm",IEEEWirelessCommunications,Vol.
10,No.
3,June2003,pp.
8-16.
[Zheng1997]Q.
Zheng,T.
Yokatani,andT.
Ichihashi"Connectionadmissioncontrolforhardreal-timecommunicationinATMnetworks",Proceedingsofthe17thInternationalConferenceonDistributedComputingSystem,IEEE1997.
[3GPP2002a]3GPP"QoSconceptandarchitecture",http://www.
3gpp.
org,ETSI23.
107v5.
5.
0(2002-2006).
[3GPP2002b]3GPP"UTRANoveralldescription",http://www.
3gpp.
org,ETSI25.
401v5.
4.
0(2002-2009).