uplink是什么意思ink的用法读音典

,A,nalytilcalerform-anceEvaluationMi1xed

ServiceswvithariableDataRatesfortheUpink

UM-TS

LarissaPopova

andWolfgangKoch

LehrstuhlfurMobilkommunikation,

UniversithtErlangen-Niurnberg,Germany

Email:{popova,koch}@

Abstract-Thecontributionofthis

paper

,we

introduceanewefficientanalyticalmethodforuplinkperfor-

manceestimationinmulti-serviceUMTSnetworksbytakingthe

dynamicbehaviouroftheusertransmissionrateinto

account.

Ourperformancemeasuresaretheaverageuserinformation

,atraffic

managementtechniqueisappliedtoUMTSandincorporatedinto

theanalyticalapproachtospecifyhowresourceshavetobeade-

quatelyengineeredtomeetquantitativeperformanceobjectives.

Forthisweinvestigatethejointbehaviouroftheadmissioncon-

trolcharacteristics,scheduling,thresholdandreservationscheme

forUMTSTerrestrialRadioAccessNetwork(UTRAN)in

terms

oftheir

agreement

withGradeofService(GoS)requirements.

Theperformanceoftheproposedalgorithmsisevaluated.

UCTION

Asthecapacitiesandflexibilityofcommunicationnetworks

increase,andtheamountofservices,theyprovide,expand,

thedemandforradioresourcesbecomeslesshomogeneous.

Inordertoprovideserviceinthisheterogeneousenvironment

moreefficiently,itisdesirabletobeabletoestimatethe

blockingprobabilityforallthespecificclassesofdemands.

Traffictheorycurrentlyplaysaquiteminorroleinthe

k

dimensioningofUMITS

systems

isgenerallybasedonrough

estimatesratherthanonrealisticanalyticalmodelsforthe

erableeffortis

spentonthe

designofavarietyofqualityofservice(QoS)mechanisms.

Theroleofthelatteristoensureanappropriatelevelofservice

heless,

thenecessityofapplyingtheappropriatetraffic-performance

relation,iftheobjectiveistoensurethatQoSmeetsits

targets

foragivenpopulationofspecifiedkind

of

users,

isstill

remaining.

Theperformance

e我国传统节日的由来简短 valuationofmulti-servicewirelessnet-

worksisacomplexproblem,becausethenon-Poisontraffic

characteristics,dynamicbehaviourofusertransmissionrate

andtheeffectoftheradiochannelhavetobeconsidered.

Furthermore,multi-bandWCDMAallows

manyuserstotrans-

single-userperspective,allotheruser\'ssignals,locatedinthe

air-interface,arenoise.

Every

timeanewcall

isaccepted,the

interferencelevelincreases,consequently,decreasessignal-to-

rtopreservethequality

ofalreadyacceptedcalls,anewcallshouldnotbeacceptedby

the

system

ifthiscallincreasesthenoisefor

any

otherexisting

apacityislimitedbythe

amountofinterferenceandnoiseintheradiointerface,itisby

definitionasoftcapacity,sincethereisnosinglefixedvalue

,theUMTS

system

has

nohardlimitsoncallcapacity,whichmakestheestimationof

networkcapacitymuchmorecomplex.

SincetheactualloadoftheexistingUMTSnetworksisstill

toolowtoanticipatethereallybusyhourtrafficvalues,there

aretwoalternative

waystoestimatetheperformanceof3G

mobilenetworks:theanalyticalmodellingandthesimulation

ofthefunctionalityofUTRAN.

Inthefollowing,abriefovervie部编版五年级语文上册 wofthe-state-of-the-artof

trafficengineeringofUMTSnetworksispresented.

In1981KaufmanandRobertspublishedrecursiveequations

whichcanbeusedforcomputingtheblockingprobabilities

foranErlangMultirateLossModel[3].In[5]andlater[2]

statedep清明时节雨纷纷的意思是什么 endentblockingprobabilitiesforcalculation

of

the

performanceofcellular

systems

withWCDMAradiointerface

wasintroduced.

In[5],[2]theother-cellinterferenceismodelledasalog-

normalrandomvariableandtheblo归去来兮辞原文及翻译注释高中课本 ckingprobabilitiesare

computedusinganappropriatelymodifiedKaufman-Roberts

[6]anuplinkUTRANperformancewasesti-

rminetheblockingprobabilityofanewcall

appearinginthecellktheloadgeneratedbythecallinthe

neighbouringcellswastakenintoconsideration.

Hoowever,inabovementionedstudies,thespecificsofsuch

importantfeaturesofUMTSasvariableusertransmissionrate

ologies

toanalyzeperformanceofUTRANbytheuseofsuitablesim-

ulationtoolshavebeenintroducedin[I].

Althoughsimulation

toolsproducegoodresultstheyneed

manyinputparameters

andahugeamountofcomputationtime.

This

paperpresentsanewapproximationmethodfor

per-

formanceestimationofmulti-serviceUMTSnetworks,taking

intoaccountkeyfeaturesofUTRAN,suchastheinterference

1-4244-0398-7/06/$20.002006IEEE

limitedsoftcapacityofacellandvariableuser\'stransmission

rate.

Ouralgorithmisnewforitsanalyticalconsiderationoftwo

quitedifferent

aspects,

whichareassociatedwithQoSinterms

ofacceptableinterferenceleveland

averageuserinformation

rateattheolnehand,andblockingprobabilityoflnewcalls

esforevaluationofrealisticcases

rasttothewidespreadview,thatanalytical

approachesformodellingaradiointerfacearetoosimplified

andyieldinaccurateresults,theanalyticalmethodproposed

inthispaperdemonstratestheflexibilityandtheaccuracy

se,somedegreeofsimplificationis

sthedominantas丽春酒多少钱一瓶 pectsoftherealquantities

wereidentifiedandrepresentedintheselectedmodeland

unimportantaspectswereneglected.

ionII

somebasic

relationsregardingadmissioncontrolinUTRANaregivenand

the

system

ionIIItheproposed

methodforblockingprobabilitycalculationforUTRANis

esofincorporationoftraffic

management

strategy

intothemodelarepresentedforrealisticUMTS

scenarioinSection

IV.

Theeffectsofthis

strategyonthe

y,SectionV

concludesthe

paper.

MODEL

sicrelations

InaUMTS

systemthenumberofeffectivechannelsis

obtainedintermsofthenumberofavailablecodes,the

processinggain,andtherequiredEbINo(Eb:averagereceived

bitenergy,No0powerspectraldensityofwhiteGaussian

noise).Accuratesignalreceptionofdifferenttrafficstream

types

ispossibleonlywhentherelationof

energyper

bitto

NOforthejth

useriscalculatedas

Wpi

(EbJN)j=Rj.

vjItot,.1-Pj1

wherePjisreceivedsignalpower

fromthejth,W-chiprate

ofspreadingsignal,vj-activityfactorofthejth

user,Rj-

bitrateofthejthuser,It,tll-totalreceivedwideband

power

includingthermalnoise

power.

Beforeadmittinganewconnection蜂唐罗隐古诗视频 inthe

system,connec-

tionadmissioncontrolensuresthatanewcallwillnotsacrifice

issionofanew

connectionofaspecificserviceclasskintheconsideredcell

is

arandomvariabledependingonthecurrentstateofthe

cell,thenoise-riseofthenewconnectionandtheother-cell

interference.

B.

Modelcharacteristicsand

assumptions

The

system

underconsiderationisacellularUMTSnet-

methattheoverall

system

ishomogeneousin

hasystemany

cellisstatisti-

callyidenticaltoanyothercellThetotalcapacityofeachcell

initionofachannelusedin

ourmodel

isgivenin[2].

Thefollowing

parametersandassumptionsareused:

*

Thenetwork

servesnindependentclassesofPoissondis-

tributedtrafficstreamswitharrivalrates:Al,A2*.A,

*

Theservicetimeforthecallsofparticularservicesclasses

2

..,n,respectively.

*Eachcallisassignedoneormultiplechannelsdepending

,achannelisdefinedbya

viceclasskischarac-

terizedbyadatarateofdk=mkx12.2kbps(mkis

positiveinteger).

.Themeantrafficofferedtothesystembyserviceclassk

isak=Ak/[k

.

Callsofdifferentserviceclasseshavedifferentresource

requirements

(user

transmissionrateandservicetime),

whichdependonthetotalnumberofalreadyoccupied

channels,lloftheservicekcanbeadmitted

inthesystemwithotherthatinitialtransmissionrate.

*

Theservice

processes

inthecellsaredependent,and

anactualtrafficloadinonecellinfluencetheservice

probabilityinothercells.

Theobjectiveistodeterminethecallblockingprobabilityand

averageusertransmissionrateforeachserviceclass.

EDALGORITHM

Inordertodeterminethecallblockingproba励志名言名句优美句子摘抄 bilityand

averageusertransmissionrateforeachserviceclass,werevisit

theanalysisfrom[2],wherethefollowingrecurrentformulais

proposedforestimationoftheuplinkperformanceofUTRAN:

(2)G(r)=rLakTnkG(r-Tnk)(l-br-m,k),

k=l

whereG(r)istheprobabilityofrbusychannelsinthe

system

andbisthestatedependentblockingprobability(softblock-

ing),whichcharacterizesthecurrentother-cellinterference

situation.

Weappropriatelyextendthisformulatostudyarealistic

ingly,letusconsiderapart

of

thetwo-dimensionalMarkovchaindiagramconstructedfora

systemsupportedtwotrafficclasses(Fig.1).

Thisshouldbeunderstoodasfollows.A

newcallofservice

classkwithresourcerequirementdkisonlyacceptedifall

onally,

weassume,

thatinUTRANmultiplechannelsrequiredto

establishanewconnectionareoccupiedsimultaneouslyand

thereforetheother-cellinterferenceremainsthesameduring

thisallocation

process

forallchannels.

Then,withtheabovementionedassumption,theservice

probabilityforanewcallwithresourcerequirementdkis

definedas:

1-bk

Pr{r/

rN-(r+2)...r\'

=Prfr\'

+mk-1)

(3)

entofstate-transitiondiagramofMarkovchaininamulti-

serviceUMTS

system

withasoftblocking

whereand

r\'

istheinterferencefromothercellexpressedas

channelequivalent.

Furthermore,toincorporateatransmissionratereduction

policy(TRRP)intoourmodel,weemployamodifiedan-

alyticalapproachforestimatingtheblockingprobabilityin

AsynchronousTransferMode(ATM)networksproposedin

[4].Inthisso-calledthresholdmodelforblockingavoidance

newcallsofcertainserviceclassescangetaconnectionwith

otherthantheinitialresourceandservicetimerequests,which

omemodificationstothefirstsum

andtheincorporationintotheexpressionofthesecondsum

weget:n

G(r)rakMk6k(r)G(r-Mk)+

r

~ak,Tnki

1k5(r)G(r

k=l1=1

Tnk)

forr

=1,...

IN,where3k(r)=1when

I_r

or,

when

r<_JA+

1r7k,

otherwise

6k-1,1isanumberofapossible

variationoftransmission

rate,Jklistheadjustedthreshold(s).

Forcalculationofthetotalblockingprobabilityforaservice

whiletakingintoaccountTRRPwemodifyproperlythelimits

ofthesummationsinthefollowingexpression:

N-Mkl

Bk=G-1G(r)brkk(5)

r=O

N

whereGEG(r)Thisistheprobabilityofacallofservice

classktobeblockedwithitslowestpossibletransmissionrate,

xxithtakingintoacounttheother-cellinterf民间鬼故事短篇超吓人 erence.

MANCEEVALUATIONANALYSIS

Somenumericalexamplesarepresentedinthissectionin

order.

*

todemonstratetheflexibilityandtheaccuracyoftheana-

lyticalmethodusedtoevaluatetheteletrafficperformance

oftheUMTSsystems.

*

tostudyhowvariousparametersaffecttheperformance

metricsofinterest.

.toquantifythebenefitsoftheproposedtrafficmanage-

mentstrategybycalculatingthemaximumnumberof

usersthataUMTScellularnetworkcanhandlefora

givenGoSundercertainsystemconditions,suchasnet-

worktopology,relevantcellscenario,realisticuser/traffic

distributionprofile,etc.

WeconsidertheUMTSnetwork,supportingtwotraffic

classes:speechcallswith12.2kbpsandvideostreamwith

128kbpsandtargetEbINoof4and1.5,

maximumnumberofchannelsin

onecellis315,theother-

cellinterferencefactori=0.55andtheofferedtrafficis

p=ltheother-cellinterferenceasalog-

normalrandomvariableasproposedin[5].Underrealistic

assumptions,themeanother-cellinterference

power

isabout

E{Iother1=5X10-15mWwithavariationcoefficientof

1,sbeenverified

byusingaUMTSradionetworksimulator.

Themostimportantparametersandrequirementsfordiffer-

entservice-classesofthe

system

consideredherearesumma-

rizedinTable1.

TypeottheservicekSpeeh

Data

Datarate,dk[kbps]12.2

128

EtLN[dB]4

1.5

Correspondingnumberotrequiredchannelsmk1

10

Chiprateotspreadingsignal[Mchip/s3.84

Maximumnumberotchannels/cell315

Table1:Mainsimulationparameters.

Itisevidentthatin

systemswithacompletesharingpolicy

theserviceclasseswithlowtransmissionratesprofitfroma

lowerblockingprobabilityincomparisontohighrateservice

classes.A

very

undesirablesituationis,however,whenthe

blockingprobabilityofhighrateserviceclassesexceedsthe

GoSconstraint,whiletheblockingprobabilitiesofother

thefollowingweexaminetheeffectofapplyingdifferent

admissioncriteriatodifferenttrafficstreamstogivesatis-

factoryperformanceintermsofboth,blockingprobability

andsatisfactoryusertransmissionrateforeachtrafficclass.

Specifically,blockingprobabilityforthespeechserviceclass

andblockingprobabilityforthedataserviceclassareexpected

tobeinthe

range

of

1-2%

and110-11%,s,

weassumethatspeechisaconstant-bit-rateservice,while

datacallsallowadaptivedatarates,

dependingontheloadof

the

system.

Eachnewuserofthedataclasshasaresource

requirementdk=128kbpsthatwillbesatisfiedif

r_Jk,

(inourcase

1-1andareducedratedk64khps,otherwise.

ByapplyingtheTRRPintothesystemunderconsideration,

itis

necessarytobearinmind,thatloweringthetransmission

nehand,lessre-

sourcesare

neededatagiventime

forhandlingtheconnections

therhand(and

in

contrasttorealtimeconnections),thetransmissionduration

ofsuchcallsbecomeslonger,sincetheamountofinformation

tobetransmitteddoesnotdependonthetransmission

rate,

,atrade-offbetween

systemperformanceandusersatisfactionmust

befound.

TheFig.2demonstratesthesystemperformanceversusthe

thresholdforaservice-mixscenariowith70%speechcalls

and30%

nothresholdisapplied,theblocking

probabilityforthealgorithmof[2]withoutTRRPisobtained

(rightpointsofthecurves).Thisistheworstcaseutilization.

Fromthefiguresonecanseethatsettingthethresholdtoa

lowvalueleadstoadecreaseofserviceprobabilityofspeech

callsandtounnecessarydegradationoftransmissionratefor

thedataserviceclass,whereasagainintermofservice

,

the

overestimationofthethresholdleadstoanegativeimpacton

imalthreshold

forabovementionedscenarioisJ=200,correspondingtoa

e2

Blockingprobabilityforspeech(70%)andvideostreaming(3马鸣风萧萧电视剧全集免费 0%)

0.3

---speech

-data

0.25

0_2

M

2

0.15

0,

\')0.1_

0.05____..

515

Threshold[numberofchannels]

DatarateforVideoStream(N=315,p=35Erlang)

125

120

115

110

E

CU105

U)100

90go

-Ma)85C580

C15

F575

70

65

515

Threshold[numberofchannels]

ngprobabilityanddatathroughputforspeechandvideostreamn

versusthethreshold.

wecomparetheperformanceofthealgorithmof[2]withthat

ofouralgorithmTheresultsindicatethesignificantsystem

performanceimprovementwhichcanbeattainedbyTRRP.

Withincorporationofareservationschemeinouralgorithm

ear,

thatadecreasedblockingprobabilityofdatacallsisachievedatthe

expenseofsomedegradationofserviceprobabilityforspeech

trafficduetothereservationdatacalls

getmoreaccesstothe

imumnumberofusersaUMTS

cellcanhandleusingtheabovementionedtraffic

management

strategyisabout1075forspeechand462fordataservice.

Different

Bk

%

systemscenariosSpeechVideoStream

WithoutTRRP

2.2727.13

WithTRRP2.2715.16

WithTRRPand

channelreservation283

11.45

Table2:Systemperformancewith/withoutTRRPandreser-

vation.

SIONSANDFURTHERWORK

Inthispaperwehavedevelopedafastandaccuratere-

cursionforefficient

computationof

GoS

foramulti-service

scenariointheUMTS

system,bytakingintoaccountthe

orithm

presentedinthispaperisnotonlyusefulfortheperformance

predictionoftheUMITS

systems,

butalsoforobtaininga

suitableadmissioncontrolpolicyforoptimalusageofthe

systemresources.

Thefollowingconclusionscanbedrawnfromthenumerical

results:

*

Theblockingprobabilityofanewcalldependsbothon

thestateofthe

system

andtheresourcerequired.

*

Theincreaseofarrivalrateofdataserviceclassesin-

creasestheblockingprobabilityforallserviceclassesin

the

system.

*

Incorrecttrafficdifferentiationin

termofadjustedthresh-

oldresultsinunacceptablecallblockingprobabilityfor

allservice-classesorleadstodissatisfactionofalready

acceptedusers(toolongtransmissiondurationofthecalls

withreducedtransmissionrate).

.In

caseofaproperlyselectedadmissionpolicywe

observedareduceddatacalllossfromabout28%for

theno-thresholdcasetoabout11%.

ThealgorithmallowstostudyperformanceofUMTS

systems

intheirnaturalenvironmentswithappropriateradio

interfacecharacteristicsandrealistictrafficvolumes.

Althoughourresultshavebeenpresentedforatwo-class

system,withouralgorithmwecaneasilyextendourmodel

formoreserviceclasses,aswellasformultipleadmission

thresholds.

Theproposedmethodenablestoanalysesystemperfor-

manceunderconsiderationofotherimportantUTRANfea-

turessuchaschannelreservationgeneralizationoftraffic

arrivalcharacteristics,handover,aswellaspacket-service

delay,whichisundercurrentinvestigation.

REFERENCES

[1]ein,,,QoS-Aware

Pr)oJ

gQSHINE,Oct.2004.

[2]n,s,,tionofMulti-

16,Spec.

Sein.,m,Sept.2004.

1301

{RN

[3]nginaSharedResourceEnvironment.

IEEE

.,29,1981.

[4]lios,etis,tion-dependent

lPerformanceEvaluation,48,2002.

[S]yticapproximationoftheuplink

18,Berlin,Aug./Sept.2003.

[6]k,wski,ngprobability

calculationinthe《明日歌》古诗原文 uplinkdirectionforcellularsystemswithWCDMA

,Dresden,September2004.