HAL Id: pastel-00000671
https://pastel.archives-ouvertes.fr/pastel-00000671
Submitted on 26 Nov 2010
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Améliorations de l’accès paquet en sens montant du
WCDMA
Konstantinos Dimou
To cite this version:
Konstantinos Dimou. Améliorations de l’accès paquet en sens montant du WCDMA. domain_other.
Télécom ParisTech, 2003. Français. �pastel-00000671�
du WCDMA
Konstantinos Dimou
This Ph.D. thesis is the result of a four-year resear h proje t arried out in two
in-stitutions: the rst half of it has been ondu ted in the Departement Informatique et
Reseaux of
E ole Nationale Superieure des Tele ommuni ations (ENST) Paris, Fran e
and the se ond half has been done in the Department of Communi ation Te hnology of
AalborgUniversity(AAU), Denmark. The work inENST hasbeenperformedunder the
dire tionofProfessorPhilippeGodlewski. InAAUmyworkhasbeen ondu tedunderthe
dire tionof my mainsupervisorProfessorPreben E.Mogensen and of my o-supervisors
Dr. Troels B. Srensenand Dr. Jeroen Wigard (NokiaNetworks).
I would like to thankprofessors XavierLagrangeand Jens Zanderfora epting to be
the reviewers of my PhDthesis. Their omments and suggestion helped to improve this
thesis. A spe ialthanksisalso given to themembers ofthejury Prof. Bernard Fino,Dr.
ClaudiuMihailes uand to thepresident ofthe juryProf. DjamalZeghla he.
A signi ant number of people has ontributed to this work. First, I express my
gratitudeto mymainsupervisorsProfessorsP.GodlewskiandP.E.Mogensenfor
a om-modating me in their groups. The work with Prof. P. Godlewski is going to be a very
useful experien e formyprofessional life. Spe ialthanks goes to Prof. P.Mogensen for
hisen ouragement, supportandpatien ehe hasshownto methroughoutourdis ussions.
For dis ussions and suggestions I would like to thank my olleagues in the Cellular
Systems (CSys)group at Aalborg University and at NokiaNetworksin Aalborg. Spe ial
esteemgoestoT.B.Srensenforhispatien einlisteningmeandforhiswiseadvi es. For
assistan e with the implementation of simulators and for simulation ampaigns, Iwould
like to thank ClaudioRosaand Jose OutesCarnero. Forthesame reason, also thanksto
master studentsMalek Boussif and Alberto Ramos. Ifeel thankful to LisbethLarsenfor
fa ilitatingmyeverydaylife: oneveryrequestofmineregardingadministrativeissues,she
was therealways witha smile.
Thisworkhasbeensponsoredbythefren hnationalnetworkofresear hin
tele ommu-ni ations(Reseau Nationalede Re her he enTele ommuni ations,RNRT) andbyNokia;
a thanksgoesto them fortheirnan ialsupport.
I deeplya knowledgethehelpof myfriendsthroughouttheseyears. Iamparti ularly
thankfultomyfriendArturHe kerforhissupportanden ouragementhegave meduring
our numerous and long dis ussions. I fullyre ognize thesigni an e of the support that
MarisaCangelosi and Evdoxia Baniotopoulougave me foralong period.
Finally, I express my profound gratitude to my parents Dimitrios and Morfoula and
to my sister Irini for their un onditional love and support. Without their ontinuous
en ouragement, the han es of this work being a omplished would be very few. This
Les systemes de troisieme generation (3G) ommen ent a appara^tre en Europe et ailleurs. Lemoded'a esutilise surl'interfa eradiode essystemesest leCDMA (Code
DivisionMultipleA ess). EnEurope,lesystemede3Gendeploiementestl'UMTS
(Uni-versal Mobile Tele ommuni ations System), quiemploie leWCDMA (Wideband CDMA)
poursapartied'a es radio.
Une evolution majeure des systemes de 3G en omparaison ave eux de deuxieme
generation (2G) est la possibilite des servi es de haut debit. Ce i permet l'introdu tion
des nouveaux servi es et notamment des servi es de donnees. Il est envisage que les
servi es de donnees formeront la plusgrande partie du tra e oule dans les reseaux de 3G. Ilest prevu que esservi esvont oexisterave laparole,dejapresenteauxsystemes de 2G. Par onsequent, des s enarios de tra mixtedoivent^etre onsideres. L'existen e desdierentstypes detra augmente la omplexite de lagestiondes ressour esradios.
Dans ette these on etudie l'a es par paquet en UMTS. L'etude se on entre sur le
sens montant de la transmission, 'esta dire dumobile (UserEquipment ou UE) vers le
reseau. Ensensmontant,laressour eradiorareestlapuissan ere uealastationdebase
ou Node B(selonlaterminologieUMTS).Le butdenotre travail estd'ameliorer ertains
me anismesde l'a esparpaquetan d'optimiser l'utilisationde lapuissan e.
Lepremierme anismedel'a esparpaquetpourlequeluneortd'ameliorationestfait, estl'adaptationdulienradio. Enparti ulier,l'a entestmissurle asdelatransmission multiservi e. L'UE doit partager le debit global qui lui est alloue entre les dierents servi es a tives. Cesdernierssont vehi ulesdansdes radio bearers (tuyauxsupports).
A
haqueintervalleelementairedetransmission(Transmission TimeInterval ouTTI),l'UE
sele tionneunsous-debitpour haquebearer ; e isefaitparlasele tiond'un\formatde transport" a appliquer pendant la duree TTI. Cette pro edure s'ee tue dans la ou he MAC(MediumA essControl) ;leresultatdelasele tionestune ombinaisondeformats
de transport (Transport Format Combination, TFC) utilisee par la ou he physique. La
pro edure, nommee sele tion de TFC, permet d'adapter la transmission des dierents
servi es aux onditionsvariablesde la propagation radio: elledetermine notablement la
performan ede latransmission.
L'algorithme de sele tiondeTFCest tra e danssesgrandeslignesdanslanorme. Un de sesprin ipesestde favoriserletra tempsreel au detrimentdes servi esdesdonnees parpaquet. Cependant, letra temps reel peut^etre perturbe parle tra donnees sous
ertaines onditions, en parti ulier pourles mobileseloignes du Node B. On propose un
algorithme de sele tion de TFC qui limite es perturbations et qui ore une plus large
zonede ouvertureaux servi estemps reels.
Un autre type de me anisme etudie on erne l'ordonnan ement de paquets entre les
dierentsutilisateurs. C'est unepro edure quiest ontr^olee parla partie xedu reseau. Son impa t sur les performan es du systeme est etudie en detail et evalue a l'aide de simulations. Ensuite,lepotentieldetroisme anismesvisantaameliorerl'ordonnan ement de paquets est evalue. Un premierme anisme est nomme fast Variable Spreading Fa tor
(fast VSF) :les UEsdistants hangent rapidement leurfa teur d'etalement (SF) an de
inter-Un troisieme me anisme nomme \fast s heduling" (ordonnan ement rapide) ra our itle
y le d'ordonnan ement. Lesresultatsont montreque sous ertaines onditions,lemode
de entralise reduitle delaipar paquet jusqu'a 25%. L'ordonnan ement rapideaugmente la apa ite dusysteme jusqu'a10%. Deplus,l'ordonnan ementrapideameliorelaqualite de servi e per ue par par les utilisateurs en terme de debit par utilisateur et delai par
paquettransmis.
Cette dissertation est omposee de six hapitres. Le premier hapitre sert omme
l'introdu tion generale de la these. Le adre du travail est presente : au debut une des riptiongeneraledel'UMTSestdonnee. Elleestsuivieparunedes riptiondesservi es
etdes lasses deservi es en UMTS.La notionde Qualitede Servi e(QoS)en UMTS est
aussi presentee. La modelisation du tra et en parti ulier du tra des donnees par
paquet joue un r^ole important a notre travail. Pour ette raison, la partie suivante du
hapitretraite esujet. Ensuite,lebutdutravailestmanifeste: optimiserl'utilisationdes ressour es radios an d'augmenter la apa ite du systeme et d' ameliorer la QoS oerte auxutilisateurs. Deplus, ettepartiedu hapitrepresentelestravauxanterieursainsique la litterature asso iee au sujet. Le hapitre termine ave une des riptiongenerale de la stru turedu rapport dethese.
Le deuxieme hapitre omporte omme une presentation generale du reseau d'a es
radiode l'UMTS (UMTS Terrestrial Radio A ess Network, UTRAN). Le hapitre
om-men e ave unedes riptionde l'ar hite tureproto olaire del'UTRAN.Lesproto olesqui
omposent la pileproto olaire de l'UTRANsont brievement de rits ainsique leurs
fon -tionsprin ipales.Lesdierentstypesde anauxenUMTSsontaussipresentes: les anaux logiques,les anauxdetransportetles anauxphysiques. Lades riptionsefo alisesurles
anaux detransportqui onsistuentunenouveaute de l'UMTS parrapportauxsystemes
de2G. Deplus, e on eptde analde transportestlargement utilisedansletravail. Les
denitionsduformatdetransport(TransportFormat,TF),dela ombinaisondesformats
de transportsont donnees. Le me anismede sele tiondu TFCestausside rit. La n du hapitre est dediee a la des riptiond'une autre ara teristique de la ou he physiquede l'UTRANquiestlargementutiliseedans etravail: le hangementdefa teurd'etalement. Dans le sens montant, les UEs sont autorises a hanger leur fa teur d'etalement et par onsequent leur debit a haque TTI an d'adapter leur lien radio. Ce trait est nomme
dansles normes3GPP Variable Spreading Fa tor (VSF).
Le hapitre 3 traite le sujet de l'adaptation du lien radio a la ou he MAC. Ce i
s'ee tue par la sele tiondu TFC, omme il est ite auparavant. Le hapitre ommen e
ave unedes riptionde la ou he MAC. Ensuitela ongurationet lefon tionnementde
la ou he MACdansle asde latransmissionmultiservi een sensmontant sontexposes.
Lasele tionduTFCpresenteunplusgrandinter^etau asdelatransmissionmultiservi e.
Ensuite, l'algorithme de sele tion de TFC, tel qu'il est tra e dans les normes 3GPP est
de rit ; dans le as de transmission de plusieurs servi es, la sele tion du TFC est basee prin ipalementsurlaprioritedesservi es. Lesperforman esdel'algorithmesontevaluees
a l'aidede simulations. Le modele desimulationestde riten detail danslesparagraphes
suivants du hapitre. Le s enario simule ontient trois types de tra : i) de la voix,
le debit total de l'UE et la puissan e transmise par l'UE. Les mesures sont prises en
fon tion de la distan e entre l'UE et le Node B. Les resultats des simulations montrent
que le taux d'erreur augmente tres rapidement apres une ertaine distan e entre l'UE
et le Node B. Ce qui a omme resultat la perte du servi e de la parole. Ce i s'ee tue
a une distan e onsiderablement plus ourte que elle au as ou seulement la parole et
la signalisationsont transmises. Ce i estd^u au me anisme du ontr^ole de puissan e. En parti ulier,pendantlatransmissionsimultaneedelaparoleetdesdonnees,oulapuissan e de transmissionestassez elevee, les ommandesde ontr^olede puissan eexigent souvent
que l'UEtransmette ave lapuissan emaximale. L'UE se rendalorsin apablede suivre
les ommandesposterieures de ontr^ole de puissan e etde maintenirson lienradio. Une
variantede l' algorithmede sele tion de TFCest proposee. Le nouvel algorithmeest sur sesgrandesaxes similaireave l'algorithmede ritdanslesnormes. L'algorithmeintroduit unenouvellemargealapuissan edetransmission. Cettemargeest onsidereeuniquement au asoulaparoleetlesdonneessonttransmisessimultanement. Sonbutprin ipalestde protegerlatransmissiondelaparole. Lessimulationsontmontrequelenouvelalgorithme
presentedesmeilleuresperforman esen omparaisondel'algorithmedelanormeenterme
de taux d'erreur pour tous les servi es, de delai par paquet, de debit total et en terme
de puissan e de transmission. L'algorithme est teste pour un ertain s enario de tra ;
neanmoins, desresultatssimilaires sontattendus pourd'autress enarios de tra .
Le hapitre 4 examine l'a es par paquet en sens montant de l'UMTS. Le hapitre
ommen e ave la des ription generale de la pro edure de l'a es par paquet. Ensuite les me anismes prin ipaux de la pro edure sont de rits : i) l'envoi de requ^etes des UEs
vers le reseau sous la forme de reports de mesures de volume de tra (TraÆ Volume
Measurement Reports,TVMR), ii)l'ordonnan ement depaquets,sur le ^ote reseau et iii) lasele tiondeTFC, surle ^ote UE.L'a entestmisalades riptiondel'ordonnan ement de paquets. Plusparti ulierement,lades riptionsuitdeux axes : d'unpart,l'algorithme d'ordonnan ementparpaquetestde ritendetail. L'algorithmeessayedefaireune allo a-tionequitable desressour es disponibles. Del'autrepart,les problemeslies al'allo ation
depuissan ependantl'ordonnan ement depaquetssont traites.
Alasuitedu hapitre,le modeledesimulationestdetaille. Lemodele onsisteenunreseau ellulairede24 ellules. Les plusimportantsme anismes des ou hes bassesy sont onsideres. Le servi e qui est simuleestun servi edejeu surinternet. Lesresultatsdessimulationsmontrentque pour lemodeleetpourleservi esimulela apa ite dusystemeest1.27Mbps;lesysteme peut orir unea eptable qualite deservi e a20 UEs environ par ellule.
Le hapitre 5 presente les ameliorations proposeesde l'a espar paqueten sens mon-tant. Ces ameliorations visent a : i) augmenter la apa ite du systeme et ii) ameliorer
ertains attributs de la QoS oerte aux utilisateurs, par exemple, le delai par paquet
transmis. Comme il est ite auparavant, les trois ameliorations proposees sont : i) le hangement rapide du fa teur d'etalement (fast VSF), ii) l'a es paquet de entralise et
iii)l'ordonnan ement de paquetsrapide.
Les hemafastVSF proposequelesUE's hangentrapidementleurfa teurd'etalement ande onserverleurpuissan edetransmissionstable. Ce iresulteenuneredu tiondela varian ede l'interferen e aux ellules voisines(interferen e inter- ellulaire). Enrevan he,
globale du systeme est etudie. Une redu tion de la varian e de l'interferen e globale permettra l'augmentationde l'interferen e ibleet parequivalen e de la puissan ere ue
ibleau NodeB. Ce isetraduit enuneaugmentationde la apa ite dusysteme. Pourle
modeledesimulationutilise,lesresultatsontmontrequelefastVSFn'apportepasdegain signi atif. Unedesraisonsprin ipalespour elaest queles dierentes restri tionssurla
sele tiondu TFCne permettent pasauxUEsde stabiliserleurpuissan edetransmission
etpar onsequent de diminuerla varian e de l'interferen e aux ellules adja entes. En mode d'a es de entralise, les UE ont plus de liberte a la sele tion de leur TFC etparequivalen ede leurdebit. LesUE gardent leurbearers ouverts pendant toute leur session. LesUEtransmettentdesdonneesenfon tiondelavaleurdel'interferen eglobale ; ettevaleurestperiodiquementdiuseedansla ellule. Lesresultatsdessimulationsont montre qu' en asde harge de tra faible oumoyennedans la ellule,une redu tionde 25%du delaiparpaquetdes donnees estatteint.
L'ordonnan ement de paquets rapide onsiste a allouer et desallouer des debits plus frequemment auxutilisateurs,par rapportau asde referen e. L'ordonnan ement rapide estatteintendiminuantlesvaleursdetemporisateursquile ontr^olent. Alors,ladete tion
de non-usagede ressour es des ertainsutilisateursest ee tuee plusrapidemement ainsi
quel'allo ationde esressour es auxautresutilisateurs. Lesresultatsdessimulationsont montre qu'ungainde 10% en apa ite est atteint.
Le hapitre 6 ontient les on lusions du travail et ilpropose letravail a venirsur e sujet.
Third generation (3G) ellular systems implement a new multia ess te hnique, the
Code Division Multiple A ess (CDMA). In Europe, the 3rd generation system under
deployment is the Universal Mobile Tele ommuni ations System (UMTS). It utilizes, in
its radioa esspart,thewidebandCDMA standard.
Themajorevolutionof3Gsystemsistheprovisionofhighdatarateservi es. Moreover,
3Gnetworksoerthepossibilitytotheiruserstohavemorethanonesimultaneousservi e.
A pa ket-swit hed mode of data transport has been in luded in 3G systems from the
beginning of their on eption. It is anti ipated that internet-related appli ations along
with voi e, are going to be the main sour es of traÆ in these systems. Consequently,
thepa ket-swit hed mode oexists withthe ir uit-swit hedone,the latterbeingpresent
already inse ond generation ellularnetworks.
Considering that the most restri ting part of a wireless network is the radio a ess
part, the provision of the requested Quality of Servi e(QoS) to a high number of users
in reasestherequirementforaneÆ ientutilizationofradioresour esbothbythenetwork
and theusers.
In this thesis, we mainly investigate the performan e of pa ket data servi es in the
radioa essnetworkofUMTS.Wetrytoenhan eanumberofradioresour emanagement
(RRM)me hanismsthathaveanimpa tontheperforman eofpa ketdataservi es. The
proposed enhan ements aim at optimizingthe utilization of resour es, whi h is expe ted
to in reasesystem apa ity and onsequentlyto improve the oered QoS to pa ket data
users.
In UMTS however, pa ket data servi es are expe ted to oexist with voi e, as
men-tionedabove. Therefore, apartofthethesisdealswiththeperforman eofbothrealtime
traÆ and pa ket datain ase ofmixed traÆ s enarios.
The workfo useson theuplinkdire tionoftransmission,hen ethetransmissionfrom
the User Equipment (UE) to the network. A me hanism for whi h an enhan ement
at-tempt is performed is the radio link adaptation. In parti ular, the ase of multi-servi e
transmission is studied. The UE s hedules its total allo ated data rate to its a tivated
servi es, that are ir ulating in its radio bearers. At ea h transmission time interval
(TTI),theUEsele tsapart ofits totaldatarate forea h radiobearer: thisisperformed
throughsele tion of a \transport format" to be applied duringthe TTI. This pro edure
is performed in the MAC (Medium A ess Control) layer; its output is a ombination
of transport formats (Transport Format Combination, TFC) used by the physi al layer.
The pro edure, named TFC sele tion, permits to adapt the transmission of the various
servi es to the hanging radiopropagation onditions: it determinesheavily the servi es
performan e.
TheguidelinesoftheTFCsele tionalgorithminuplinkarepresentedin3Gstandards.
The prin iple of the algorithm is that it favors the transmission of real time traÆ over
servi es.
A se ond me hanismunderstudy isthepa kets heduling among users. It isa
pro e-durethatis ontroledbythenetwork. Itsimpa tonthesystemperforman eisinvestigated
and quantied through system level simulations. We study the potential of ertain
fea-tures to in rease the eÆ ien y of the pa ket s heduling. A rst feature is named fast
VariableSpreading Fa tor (VSF):UEs farfromtheNode B, hangerapidlytheir
spread-ing fa tor in order to maintain their transmission power onstant, whi h stabilizes the
other ellinterferen e. Ase ondfeatureisade entralizeduplinkpa ketdataa essusing
feedba k information on the total interferen e in the ell. A third feature, alled \fast
pa kets heduling",redu esthepa kets heduling y le. Resultshaveshownthatin ases
ofmediumandlowload,thede entralizeda es moderedu esthedelayperpa ketupto
25%. Fasts hedulingin reasesthesystem apa ityupto 10%. Moreover, itimprovesthe
QoSexperien ed by theusersinterms of throughputperuser anddelaypertransmitted
1 Introdu tion: Pa ket A ess and Servi es in Third Generation Cellular
Systems 1
1.1 S ope . . . 1
1.2 UMTS Context . . . 2
1.3 Servi es andQoS Con eptin UMTS . . . 5
1.3.1 UMTS Servi eBearers . . . 5
1.3.2 TraÆ Classes andQoS Attributes inUMTS . . . 6
1.3.3 TargetedQoS metri s . . . 8
1.4 Pa ketData TraÆ Modeling . . . 8
1.5 Aimof thePhDThesis. . . 10
1.6 Overviewof thePhDDissertation . . . 13
2 UMTS Radio A ess Network (UTRAN): Proto ol Ar hite ture and Radio A ess Bearers 15 2.1 UTRANProto ol Ar hite ture . . . 15
2.1.1 Proto ol Ar hite ture . . . 16
2.2 Physi alLayer . . . 18
2.2.1 Physi alChannels . . . 19
2.2.2 Main Fun tions . . . 19
2.3 Radio A essBearer Ar hite ture . . . 21
2.4 Transport Channels . . . 23
2.4.1 TransportFormat (TF) . . . 23
2.4.2 TransportFormat Combination(TFC) . . . 24
2.4.3 TransportFormat CombinationSet (TFCS) . . . 25
2.4.4 TransportFormat CombinationSet Forming . . . 25
2.4.5 TFCS Noti ation to theTransmitterSide . . . 27
2.5 Physi alLayerPro essing . . . 27
2.6 Variable SpreadingFa tor . . . 28
2.7 Con ludingRemarks . . . 29
3 Radio Link Adaptation in Medium A ess Control 30 3.1 MACProto olCon ept inUTRAN. . . 30
3.2 MACProto olBasi Features . . . 31
3.3 UplinkMulti-traÆ Transmission . . . 36
3.3.1 MAC layerfun tions inuplinktransmission . . . 37
3.3.2 Uplinktransmissionpro edure inMAClayer . . . 38
3.4 Transport FormatCombination Sele tion . . . 40
3.4.1 3GPP Compliant Algorithm. . . 41
3.4.2 Alternative algorithms . . . 41
3.5 Simulation Model . . . 42
3.5.1 TraÆ Sour esChara teristi s . . . 43
3.5.2 Simulation Parameters . . . 45
3.6 Simulation Results . . . 46
3.7 Con lusionsand Dis ussion . . . 53
4 Uplink Pa ket Data A ess inWCDMA 55 4.1 Introdu tionand Aim . . . 55
4.2 GeneralIssues. . . 57
4.3 Capa ityRequests . . . 58
4.4 Pa ketS heduling . . . 60
4.4.1 Pa ket S heduler Fun tioning . . . 62
4.4.2 Ranking ofCapa ityRequests . . . 64
4.4.3 Data Rate Downgrade . . . 64
4.4.4 Estimation oftheAvailablePowerforS heduling . . . 64
4.4.5 S heduling Poli y . . . 65 4.4.6 Load Control . . . 67 4.5 PowerIssues . . . 67 4.6 SimulatorDes ription . . . 70 4.6.1 System Model. . . 70 4.6.2 Capa ity Requests . . . 70 4.6.3 Pa ket S heduling . . . 72
4.6.4 TransportFormat CombinationSele tion . . . 73
4.6.5 TraÆ Modeling . . . 74
4.7 Performan eMetri s Used . . . 75
4.8 Simulation Results . . . 77
4.8.1 Impa t of theTFC EliminationMe hanism . . . 77
4.8.2 UplinkPa ketData A essPerforman e . . . 82
4.9 Con lusionsand Dis ussion . . . 86
5 Enhan ements of the Uplink Pa ket Data A ess in WCDMA 87 5.1 Motivation forEnhan ed UplinkPa ketData A essand Goals . . . 87
5.1.1 On theSystem Capa ityIn rease . . . 88
5.1.2 Delay Redu tion . . . 90
5.2 Fast Variable SpreadingFa tor . . . 90
5.2.1 Con ept . . . 90
5.2.2 Preliminary Study on the Capability of the Fast VSF to Stabilize i-fa tor . . . 91
5.2.4 Simulation Parameters . . . 96
5.2.5 Simulation Results . . . 97
5.3 De entralizedUplinkPa ket Data A ess . . . 100
5.3.1 Motivation For De entralized UplinkPa ket Data A ess . . . 100
5.3.2 De entralized UplinkPa ketData A essCon ept . . . 100
5.3.3 SimulatedCases . . . 103
5.3.4 Simulation Parameters . . . 103
5.3.5 Simulation Results . . . 104
5.4 Fast Pa ket S heduling. . . 105
5.4.1 Con ept . . . 106
5.4.2 Simulation Parameters . . . 106
5.4.3 Simulation Results . . . 107
5.5 Con lusionand Dis ussion . . . 110
6 Con lusions and Further Work 111 6.1 Con lusions . . . 111
6.2 Future Work . . . 112
A Simulators Des ription i A.1 Fast Power Control . . . ii
A.2 OuterLoop PowerControl. . . iii
B On the apability of the fast VSF to redu e the varian e of the UE transmission power vi B.1 ConsiderationsontheTTIsizeandUEspeedinrelationwiththefastfading hara teristi s . . . vi
1.1 GeneralUMTS ar hite ture.. . . 3
1.2 Prin iple oftheuplinkpa keta essin UTRAN. . . 4
1.3 SimpliedOverview ofthe layered QoSar hite ture inUMTS. . . 6
1.4 Generi web browsingmodel. . . 9
1.5 Plan ofDissertation. . . 14
2.1 Radio Interfa eProto ol Ar hite tureinUTRAN. . . 16
2.2 UserPlaneproto olsta kinthe aseofestablishmentofdedi ated hannels betweenthe UEand thenetwork. . . 18
2.3 Control Plane proto ol sta k when dedi ated hannels are established be-tween theUE andthe network. . . 19
2.4 Code generation tree of the orthogonal variable spreading odes used in UTRAN. . . 20
2.5 CongurationofanUEsupportingmorethanone radiobearers(userplane proto olsta k). . . 22
2.6 Physi allayerpro essing model fortheuplink. Modelbased on theone in [67℄. . . 28
3.1 Use of the four types of primitives. The numbers next to the primitive namesindi atetheorder of theirtransmission. . . 32
3.2 MACdataPDU . . . 35
3.3 MAC proto ol during uplink multi-traÆ transmission. Transmission is doneon dedi ated logi alandtransport hannels. . . 37
3.4 UplinktransmissioninMAC layerindedi ated mode(UEside). . . 39
3.5 FER forvarious distan esbetweentheUE and theNode B. . . 48
3.6 Spee h BLERforvariousdistan esbetween theUE and theNode B. . . 48
3.7 Per entage of lostIP pa ketforvarious distan esbetweentheUE and the Node B. . . 49
3.8 MeanUEtransmissionpowerforvariousdistan esbetweentheUEandthe Node B. . . 50
3.9 Probabilityof sele ted TFC foradistan e to theNode Bequalto 1400 m.. 51
3.10 CDF of the UE transmission power for a distan e to the Node B equal to 1400 m. . . 51
3.11 Per entage of orre t-erroneous re eption per sele ted TFC in the ase of the3GPP ompliant algorithm. . . 52
thealternative TFCsele tion algorithm. . . 52
4.1 Division of the total re eived (Rx) power inuplinkbetweenRT and NRT traÆ . . . 56
4.2 Overviewof theUplinkPa ket Data A essinWCDMA Uplink. . . 57
4.3 TimingsintheMeasurement Report sendingpro edure. . . 59
4.4 Pa ketS hedulingWorking Environment (gureinspiredfrom orrespond-inggure in[40 ℄).. . . 61
4.5 Time intervalsin thepa kets heduling pro edure. . . 62
4.6 Fun tioningof thePS at ea h pa kets heduling period. . . 63
4.7 Pa ketS hedulingAlgorithm. . . 66
4.8 Estimation of the new total re eived power dueto the hange of the load of auser. . . 69
4.9 Network layout onsistedof three- ellse tor ongurationwith dire tional antennas. . . 72
4.10 Transport FormatCombination Sele tionPro edure. . . 74
4.11 A simplemodelingapproa h to in lude losed looptransmissionmodel. . . 74
4.12 Probabilityof sele ted TFC forall of theusersin thesystem(TFC Elimi-nation\Onand \O",8 usersper ell, ellradius 933 m). . . 78
4.13 CDF of the UE transmission power (TFC Elimination \On and \O", 8 UEs per ell, ellradius933 m).. . . 78
4.14 PDFofthenoiserise(TFCElimination\Onand\O",8UEsper ell, ell radius933 m). . . 79
4.15 Probabilityofsele ted TFCforalloftheusers(TFCElimination \Onand \O",8 usersper ell, ellradius1866 m). . . 79
4.16 PDF of the noise rise (TFC Elimination \On" and \O", 8 UEs per ell, ellradius 1866 m).. . . 80
4.17 FER as a fun tionto the distan e from the servingBS (TFC Elimination \On" and \O", 8UEs per ell, ellradius1866 m). . . 80
4.18 CDF of theEb/No target (TFC Elimination Me hanism \On" and \O", 8 UEsper ell, ellradius 1866 m). . . 81
4.19 PDF of the noise rise (TFC Elimination \On" and \O", 8 UEs per ell, ellradius 1866 m).. . . 82
4.20 PDF of the noise rise (32 UE per ell, weighting fa tor for ina tive users (WFIU) is set0,0,5 and1 respe tively). . . 83
4.21 Mean Cell ThroughputfordierentnumberofUEs inthe ell. . . 84
4.22 Probabilityof sele ted TFC forthe ases of 8,20 and 32users per ell. . . . 84
4.23 PDF of thenoiseriseforthe ases of 8,20 and 32UEs per ell. . . 85
5.1 Basi on ept ofthe fastVSF s heme. . . 91
5.2 UE at the ellborder. . . 92
5.3 Implementationof thefast VSFs heme. . . 95
5.4 CDF of i-fa tor (fast VSF \On" and \O", 12 UEs per ell, traÆ model s enario B). . . 97
model s enario B). . . 98
5.6 CDF of the experien ed delay per pa ket all (fast VSF \On" and \O", 12 UEsper ell, traÆ model s enarioB). . . 99
5.7 Average UE throughputduringa tivityperiod(fastVSF \On"and \O", 12 UEsper ell, traÆ model s enarioB). . . 99
5.8 Motivation fortheproposalof thede entralized s heme. . . 101
5.9 Pa ketdatatransmissionme hanisminthe de entralizeds heme. . . 102
5.10 CDFoftheexperien eddelayperIPpa ket,forthe asesofthe entralized and de entralized s hemes(10UEs per ell). . . 104
5.11 CDF of the UE transmission power, for the ases of the entralized and de entralizeds hemes. . . 105
5.12 PDF of the noise rise, for the ases of the entralized and de entralized s hemes. . . 105
5.13 Probabilityofsele ted TFCforall ofthe usersinthesystem,forthe ases the pa ket s heduling periodis 500 and 100 ms, ( olumns 1 and 2 respe -tively,32 UEs per ell, traÆ models enario A). . . 107
5.14 PDF ofthenoiserise,forthe ases thepa kets hedulingperiodis500 and 100 ms (12UEsper ell, traÆ models enario B). . . 108
5.15 CDFofthedelayperpa ket all,forthe ases thepa kets hedulingperiod is 500 and100 ms (12UEs per ell, traÆ model s enario B). . . 109
5.16 Probability of sele ted TFC for all of the users, for the ases the pa ket s hedulingperiodis500 and100ms,( olumns1and2respe tively,12UEs per ell, traÆ models enario B). . . 109
A.1 Modelof errorsof thepower ontrol ommand . . . ii
A.2 Modelof theouter loop power ontrol inuplink. . . iii
A.3 GeneralModelof theouter loop power ontrol. . . iv
A.4 Pseudo- ode ofthesimulatedmodelof theouter loop power ontrol. . . v
B.1 Fast fadingin relationwiththeTTIsize. . . vii
B.2 Envelopeauto orrelationoffast fadinginClarke'ss attering model([55 ℄). . viii
B.3 Standarddeviation of theUE transmissionpowerversus UE speedforthe ases thefast VSFis dea tivated and a tivated. . . x
1.1 UMTStraÆ lassesandtheirrequirementsintermsofdelayanderrorfree
delivery oftheinformationelements. . . 7
2.1 Exampleof aTransportFormatSet . . . 24
2.2 Transport hannelattributes. . . 25
2.3 Dynami attributesof aTransportFormatCombinationSet(TFCS).. . . . 26
3.1 Parametersettingsof the\ModiedGaming"traÆ model. . . 44
3.2 Simulation mainparameters . . . 45
3.3 Attributes of the transport hannles used in the simulation. Parameter valuesareinlinewith theonesin[82 ℄. . . 46
3.4 The TFCSused inthesimulation. . . 47
3.5 Total UE throughput for the distan es of 1200m to 1400 m for in ase of the3GPP ompliant and thealternative TFCsele tion algorithms. . . 49
3.6 BLER forspee h, per entageof lostIP Pa kets and mean UE throughput forthe aseofthe3GPP ompliantalgorithmandwhenaTFCElimination Oset of 4dBis applied.. . . 53
4.1 Most important simulation parameters. . . 71
4.2 Capa ityrequests relatedparameters. . . 73
4.3 Parametersof theimplementedpa kets hedulingfun tion. . . 73
4.4 Parametersettingsof the\ModiedGaming"traÆ model. . . 75
4.5 Cell Throughput, UE throughput FER for the ases TFC Elimination is \On" and \O". The ellradiusis 933 mand 1866 m. . . 81
4.6 Cell Throughput, UE throughput and FER (32 UEs per ell, WFIUis set to 0,0,5 and 1).. . . 83
4.7 Mean andstandarddeviationof theNR,CellThroughput, UEthroughput and delayperpa ket allforthe ases of 8,20 and32 UEs per ell. . . 85
5.1 Mean and standard deviation ofthe re eived power intheserving Node B 1 and inanadja ent Node B,dueto an UE at thebordersofthe ell. . . . 94
5.2 Simulation parametersforthefast VSFs heme.. . . 97
5.3 CellThroughput,UEthroughputanddelayperpa ket all(fastVSF\On" and \O",12 UEsper ell,traÆ models enario B).. . . 99
5.4 Timing related parameters for the referen e ase and for the fast pa ket s heduling ase. . . 106
pa kets hedulingperiodequalto500msand100ms,forbothofthetraÆ model s enarios. . . 110
A.1 Parametersrelated to thefast power ontrol fun tion. . . iii
B.1 Time requiredinorderto have anauto- orrelationequalto 0.7 and0.5 for
variousUE speeds. . . viii
A
AC AdmissionControl
AMC AdaptiveModulationand Coding
AMR Adaptive Multi-Rate
AVI A tualValue Interfa e
B
BCH Broad ast Channel
BLER Blo kErrorRate
BMC Broad ast/Multi ast Control
C
CBR Constant BitRate
CCTrCh CodedCompositeTRansportCHannel
CDF Cumulative DensityFun tion
CDMA Code DivisionMultipleA ess
CN CoreNetwork
CPCH CommonPa ketChannel
CRC Cy li Redundan yChe k
CS Cir uitSwit hed
D
DCCH Dedi ated ControlCHannel
DCH Dedi ated CHannels
DSCH Downlink SharedChannel
DU DataUnit
F
FACH Forward A essChannel
FDD Frequen yDivisionDuplex
FER Frame Erasure Rate
H
HARQ Hybrid Automati RepeatRequest
HSDPA HighSpeedDownlink Pa ketA ess
IP Internet Proto ol
L
LC LoadControl
M
MAC MediumA ess Control
MMS MultimediaMessage Servi e
N
NR NoiseRise
NRT NonRealTime
O
OLPC OuterLoop PowerControl
OVSF OrthogonalVariable SpreadingFa tor
P
PCH PagingChannel
PC PowerControl
PDCP Pa ketData Convergen e Proto ol
PDF ProbabilityDensity Fun tion
PDU Proto ol Data Unit
PS Pa ketS heduler
Pw Power
Q
QoS QualityofServi e
QPSK QuadraturePhase Shift Keying
R
RAB Radio A essBearer
RACH Random A essChannel
RAN RadioA ess Network
RNC Radio Network Controller
RNTI RadioNetwork TemporaryIdentity
RNS Radio Network Subsystem
RRC Radio Resour eControl
RRM Radio Resour eManagement
RT RealTime
S
SAP Servi eA ess Point
SCH Syn hronization Channel
SF Spreading Fa tor
SHO SoftHandover
SMS Short MessageServi e
StD StandardDeviation
T
TBS TransportBlo kSize
TBSS TransportBlo kSetSize
TCP TransmissionControlProto ol
TCTF Target Channel TypeField
TD-CDMA Time DivisionCDMA
TDD Time DivisionDuplex
TF Transport Format
TFC TransportFormat Combination
TFCI TFCIdentity
TFCS TransportFormatCombination Set
TFS TransportFormat Set
TTI TransmissionTimeInterval
Tx Transmission
U
UE User Equipment
UMTS UniversalMobileTele ommuni ations System
UTRAN UMTSTerrestrialRadio A essNetwork
V
VSF Variable SpreadingFa tor
WCDMA Wideband Code DivisionMultipleA ess
WFIU Weighting Fa tor forIna tiveUsers
Introdu tion: Pa ket A ess and
Servi es in Third Generation
Cellular Systems
1.1 S ope
A lassi alproblemin ellularnetworksisthe ontrolofmultiplea essofuserstothe
physi al medium. Among both dire tions of transmission, the uplink, i.e. the dire tion
from theuser tothe network isthe one thatexhibits thehighests ienti interest.
For ir uit-swit hed servi es and in parti ular for voi e, the multiple a ess issue in
uplink has been resolved in the rst and se ond generation of ellular systems ([36 ℄).
For pa ket-swit hed servi es, whi h have been onsidered from the on eption of third
generation(3G) ellular systems,themultiplepa ket a essindu esspe i problems; in
parti ular,inCodeDivisionMultipleA ess(CDMA)systems,wherethemanagement of
power is a very important issue ([58 ℄). In order to meet the needs in Quality of Servi e
(QoS)oftheirdierentservi es,thirdgenerationwirelessnetworksarerequiredtoperform
an eÆ ient management of power andof other s ar eradioresour es([22℄).
This thesis dealswith the pa ket data a ess inthe uplink(UL) in Universal Mobile
Tele ommuni ationSystem(UMTS).The latteroneistheproposed3GsysteminEurope
and in Japan and it utilizes the CDMA mode in its radio a ess part ([43℄). S enarios
with (i) pa ket data traÆ only or (ii) with a mix of pa ket data traÆ and voi e are
investigated.
Among theelementswitha keyrole inthepa ketdataa ess aretheMediumA ess
Control(MAC) proto ol and the pa ket s heduler(PS). The MAC proto ol ontrols the
a ess to the physi al medium. In uplink, it exhibits a higher interest, sin e it is
semi-autonomous (it ispartly oordinated by thenetwork as itis goingto be dis ussedlater).
The PS is the entity in harge of allo ating resour es to pa ket data users. One of the
resour esto be s heduledisthepower.
This introdu tory hapter isorganizedasfollows: inthenextse tion (se tion 1.2)the
isgiven. Thedis ussionthenfo usesonthear hite turallayoutoftheuplinkpa ketdata
a ess pro edure. The approa h we used in this thesis is \top-down" and therefore in
the next se tion (se tion 1.3) we outline the servi es and the quality of servi e on ept
inUMTS.Among thisrange ofservi es, pa ketdataservi es are expe ted to bebetween
the ones that are going to prevail in UMTS. This evolution has reated a onsiderable
resear h eort on the modeling of traÆ that originates from pa ket data appli ations.
Moreover, traÆ modelingis animportantpartofourstudyhere. Therefore,thisissueis
treatedinse tion1.4. Inse tion 1.5,the aimof thisthesis ispresentedand inse tion 1.6 theoverview ofthedissertation isgiven.
1.2 UMTS Context
The initialaim of the rst and se ond generation ellular networks was to oer voi e
servi e. In this respe t, these systems are quite su essful. The onstantly in reasing
numberoftheirsubs ribersandthehigherrequestforabetterqualityforthevoi eservi e,
raised the demand for larger apa ity. In addition, the tremendous su ess of internet
duringthelastyears pushed theneed forwirelessinterneta ess. Third generation(3G)
ellularnetworksareintrodu edwith theaimto satisfytheseemerging needs([59℄).
As mentioned inthe previousparagraph, third generation ellular systems utilizethe
CDMA mode in their radio a ess network (RAN) part. In Europe, the proposed 3G
systemistheUMTS.Itsstandardizationisstillanongoingpro ess. Dependingonthetype
of overage desired, UMTS appears in two dierent modes: Frequen y Division Duplex
(FDD) and Time Division Duplex (TDD). The former one is suitable for outdoor and
ma ro ellular areas, whereasthe latter is mostly targeted for indoorand mi ro ellular
environments. TheFDDmodeutilizestheWidebandCDMA(WCDMA)a esste hnique,
whereastheTDDmode appliestheTime DivisionCDMA (TD-CDMA) s heme.
General UMTS Ar hite ture
It is envisaged that UMTS will oer in its maturityservi es with dierent data rates,
ranging from high-speed multimedia ommuni ations, su h as video alls, to low speed
servi essu hastheShortMessageServi e(SMS). Inaddition,ithasto providethevoi e
servi easwell. Itisexpe tedthatUMTSwillallowdataratesupto2Mbpsinmi ro- ell
and indoor environments and 384 kbps inwide areas ([43℄). In order to be able to oer
these servi esto manyusers, UMTSwillheavilyrelyon pa ketswit hing. Pa ket
swit h-ing allows several users to share thesame resour es. Therefore, theUMTS Radio A ess
Network (UTRAN) is on eived ina waythatboth ir uitswit hedand pa ketswit hed
traÆ an be a ommodated ([49℄)as itshown ingure1.1.
It hasto be notedthatgure1.1, asthebiggestpartof thedes riptionthatis related
to UMTS, is based on the UMTS standards. The standardization body of UMTS is
alledthirdGenerationPartnershipProje t(3GPP).Sixreleasesof3GPPstandardshave
appeared so far. In this dissertation, the dis ussion is based on the release 5 of 3GPP
standards,forthereason thatthisisthe latest ompleted3GPP release. The majorpart
UMTS Radio Access Network (UTRAN)
Circuit Switched
Domain
Packet Switched
Domain
Core Network
Node B
Node B
Node B
Iub
UE
UE
UE
UE
UE
Iub
Iub
RNC
RNC
Iur
Figure1.1: GeneralUMTS ar hite ture.
is the radio link adaptation me hanism in MAC layer, that is going to be dis ussed in
hapter3. Inrelease5,thedes riptionofthisme hanismismore ompletedin omparison
toprevious3GPPreleasesandtheguidelinesforthealgorithmoftheradiolinkadaptation
me hanismareoutlined.
Figure 1.1 illustratesthe generi UMTS ar hite ture, with the ommon radio a ess
network (RAN) for both ir uit swit hed and pa ket swit hed servi es. It an be seen
that the ore network (CN) is split into two domains: the ir uit swit hed domain and
the pa ket swit hedone, with the former one supporting ir uitswit hed traÆ and the
latter pa ketswit hedtraÆ respe tively.
UTRAN Overview
TheUMTSradioa essnetworkisthepartoftheUMTSthatmanagesthetransmission
over the air interfa e, the latter named Uu. It is onstituted by two types of fun tional
omponentsasit an beseen ingure1.1: theRadioNetworkController(RNC)and the
NodeB. The RNCisthegatewayto the ore network (CN)andit ontainsthefun tions
that ontrol the radio resour es and the mobility within UTRAN. The Node B is the
Station. A ertainNodeBis onne tedtoonlyoneRNC,whileadesignatedRNC an be
onne tedtomorethanoneNodeBs;usuallyone RNCis onne tedtosome tensofNode
Bs. The area that is onsisted of an RNC and of the Node Bs onne ted to it is alled
radionetwork subsystem (RNS). It is implementation dependent if RNCs are onne ted
betweenthem; ifitisthe ase, RNCsare onne tedthroughthe Iurinterfa e[71 ℄.
Uplink Pa ket A ess in UTRAN
It is anti ipated that among the range of servi es provided by UMTS, internet
re-lated appli ations and servi es su h as web browsing and gaming, are going to be the
dominant ones in some years. Hen e, a signi ant part of the UMTS traÆ is going to
be pa ket data, originatingfrom internet-related appli ations and other pa ket-swit hed
servi es su h astheMultimediaMessageServi e(MMS).
(1) Capacity Request
(2) Packet Scheduling
(3) Capacity Allocation
(4) Selection of
transmission mode
(5) Packet Data
Transmission
Node B
RNC
UE
Iub
PS
Iub
Associated Signalling
Figure 1.2: Prin ipleof theuplinkpa ket a essinUTRAN.
The prin iple of the pa ket data a ess in the uplink dire tion of transmission in
UTRAN is illustrated in gure 1.2. In the s enario presented in the gure, as in the
whole thesis, we onsider that a pa ket session is already open. We also onsider that
the initial a ess pro edure has been performed and a onne tion has been established
betweentheUEandthenetwork. Afterthe onne tionestablishment,signalingmessages
an beex hangedbetween theUE andthe network.
Even though a onne tion is established between the network and ea h UE having
pa ket data, this onne tion is not the lassi al \ ir uit- onne tion". The prin ipal dif-feren e of thestudiedhere pa keta ess to the\ ir uit- onne tion"isthat thedatarate
of the onne tion isnot ne essarily onstant duringits whole period. Due to the bursty
is performed by an entity alled Pa ket S heduler (PS). With the PS being lo ated in
the RNC, as it an be seen in gure 1.2, the pa ket s heduling annot be performed on
pa ket-by-pa ketbasis. Theprin ipalreasonforthisisthedelayimposedbytheinterfa es
betweentheUE and theRNC (informationon thedelaysimposedbyUTRANinterfa es
and by its omponents an be found in [22 ℄). Therefore, the uplink pa ket s heduling
onsistsinallo atinga listof dataratesto the UEsfora ertainperiod.
Theuplinkpa ketdataa essisbasedontherequest-responsemodel,asit anbeseen
in gure 1.2. Upon data burstarrival, UEs issue apa ity requests to the PS. Following
thepa kets heduling,thenetwork notiestheUEsthatare grantedresour eswitha list
ontainingthepossibleformatsand onsequentlythedataratestheyareallowedtoapply
duringtheiruplinktransmission. TheUE an thenstartthetransmissionofpa ket data.
Atea h transmissiontimeinterval,theUEsele tsone oftheformats intheallo atedlist.
1.3 Servi es and QoS Con ept in UMTS
One of the major developments of third generation ellular systems, like UMTS, in
omparison to the se ond generationones, is theprovision of new servi es. This be ame
feasibleinthesesystemsmainlyduetothehigherdataratesthey anoer. Moreover,the
UMTSar hite tureis on eived inawaythatthemulti-servi etransmissiontoorfrom a
single UEis fa ilitated.
InUMTS,asinmostofthenetworks,the lassi alissueofprovidingtherequestedQoS
tousersoverhasbeen onsidered. Forthisreasonservi ebearersareintrodu edinUMTS.
An importantfeatureof UMTSservi ebearersis thata negotiationof theirproperties is
allowed. Hen e,UMTS an supportawide rangeof servi es thathave dierent QoSand
thisisanotheradvantage over GSM.
1.3.1 UMTS Servi e Bearers
From a user standpoint, a servi e is onsidered end to end, hen e from one user to
theother. Consequently, therequestedQoS appliesfortheendto end onne tion. Ifthe
onne tion lies a ross dierent networks, ea h network should provide the QoS over its
part. InUMTS,theQoSofa onne tion isguaranteedbytheUMTSBearerServi e. The
latter one sets the me hanisms within UMTS that attempt to provide the agreed QoS;
e.g. thetransportofdataintheuserplaneoftheUMTSproto olsta k,the ontrolofthe
asso iated-to-datasignallingandother QoSmanagement fun tions. TheUMTS bearer is
onsisted of two parts: the Radio A ess Bearer (RAB) and the Core Network Bearer.
Theformeroersthebearer servi einUTRANandthelattersupportsthebearerservi e
intheCoreNetwork (CN).Figure 1.3representsasimpliedoverview ofthelayered QoS
ar hite tureinUMTS.(FigureisnotanelaborateplanofthelayeredQoSar hite turein
UMTS, asit an be found inthe UMTS standard [61 ℄. Forillustrative purposes and for
the larityofthepresentation,onlytheelementsusedinthisworkarepresentedtherein).
In aseof a onne tion withmore thanone servi es,a RABperservi eis established. A
detaileddes riptionof theUMTS bearer servi e and ofother related issues an befound
UTRAN
CN
External
Network
Terminal
Equipment
UE
End-to-End Service
UMTS
UMTS Bearer Service
External Bearer Service
Radio Access
Bearer Service
CN Bearer
Service
Figure 1.3: SimpliedOverviewof thelayered QoSar hite tureinUMTS.
2 the wayUTRAN proto ols set me hanisms and ommuni ate in order to oer a single
RABservi e oramix ofRAB servi esto an user isdisplayed.
1.3.2 TraÆ Classes and QoS Attributes in UMTS
For a ertain servi e, the properties of its servi e bearer are dened on the basis of
the traÆ of the servi e. Hen e, dierent servi ebearer typesare asso iatedto dierent
traÆ types. Inorder to ensurethatUMTS is apableof providingQoS inana eptable
resolution, fourdierent servi e bearersare denedin UMTS. Consequently, thevarious
typesof traÆ are ategorizedinto fourtraÆ lasses.
Numerousways to ategorizetraÆ maybedened,e.g. unidire tional,bidire tional,
symmetri ,assymetri ,et . InUMTS,theprin ipaldistin tive hara teristi amongthese
traÆ lassesis theirdelaysensitivity. TheUMTS traÆ lassesare thefollowing:
onversational
streaming
intera tive
ba kground
Conversational Streaming Intera tive Ba kground
Delay Intolerant Tolerant Tolerant Tolerant
DelayJitter Intolerant Intolerant Tolerant Tolerant
ErrorData Tolerant Tolerant Intolerant Intolerant
Delivery
Typi al Spee h, Internetstreaming Webbrowsing, email,
Servi es Fa e-to-fa e (radio, dataretrieval MMS
video onferen e web ameras) fromdatabases
Table 1.1: UMTS traÆ lasses and their requirements in terms of delay and error free
deliveryof theinformationelements.
thetoleran e ofthevarioustraÆ lassesindelay,delayjitterand residualbiterrorrate. It also itessome typi al servi espertraÆ lass.
Conversational Class
The mostknown servi e of this lass is thespee h. Traditionally,it ispro essedas a
ir uit-swit hedappli ation. ThetraÆ isbidire tionaland usuallysymmetri al.
Due to the onversational nature of the servi es of this lass, the requirement in
ab-solute values of delay are rigorous. In addition, almost no varian e in the experien ed
delayof theservi e is tolerated, sin e the qualityof these servi es is di tated by human
per eption. Anotherdistinguishingpropertyofthis lassisthefa tthatthetimeinterval
betweeninformationentities(blo ksorpa kets)shouldbepreservedalmost onstant,sin e
asmalldelayinthetimingsof onse utiveinformationentitieswoulddegradesigni antly
thequalityof servi eper eived byhumans. Hen e,therequirementfordelayjitterisalso
stringent. Otherappli ationsin lude,video telephony( onferen e)and videogames.
Streaming Class
Typi alservi es ofthis lassaretherealtimeaudioorreal timevideoon theinternet.
In their majority, \streaming servi es are unidire tional and onsequently highly
asym-metri . Usually, the dire tion of transmission is from a network server to a user. The
quality of servi e is also determined by human per eption, sin e usually the re eiver is
a human. Delay requirements are thus stri t, butsin e the streaming data ow is time
aligned in the re eiving end, the requirementsin absolute delay valuesare not as severe
asfor\ onversational servi es. Thistimealignmentisa heivedthroughbuering. It isof
higher importan e though that the inter-arrival time of onse utive informationentities
exhibitslowvarian e. Hen e,thedelayjittershouldbelow. However, thetimealignment
inthere eiver an easily \ orre t"the delayjitter throughappropriatebuering.
Intera tive Class
Servi es where an user or a ma hine is on lineand is requesting data from a remote
lass, asof anykind of servi e that is basedon therequest-response pattern, is that the
user, who is requesting data from a remote server, expe ts the responsewithin a ertain
period[54℄. TraÆ of this lass isheavilyasymmetri , buta little bitmore traÆn inUL
thaninthestreaming appli ationsis expe ted.
The time elapsed from the moment the user issued a request and the moment the
user has re eived the information requested, i.e. the so alled round trip time (RTT),
is one of the key features of this traÆ lass. It determines the satisfa tion and hen e
theQoS experien ed by theuser. Dependingon the servi e, thevalue of theRTTvaries
signi antly. Another hara teristi of this traÆ lass is that data has to be re eived
errorfreeat therequesting endpoint. The mainreasonforthisisthatthese appli ations
tolerate delay. Therefore, error orre tionte hniquessu hasretransmissionsarefeasible.
Ba kground Class
This ategory is onsisted of servi es where the re eiving end is not expe ting data.
Hen e, servi es of this lass are tolerant in delay. Examples of this kindof servi es are
email and the Multimedia Message Servi e (MMS) ([28 ℄). As in the intera tive lass
servi es,alsoforservi esofthis lass,there isarequirementoferror-freedeliveryofdata.
QoS Class Attributes
The traÆ lassofa servi e onstitutesone of itsQoS attributesandassu hit
deter-minesthepropertiesofitsallo atedservi ebearer. AnumberofotherQoSattributeshas
beendened in UMTS [61℄. Most remarkable among them (other than thetraÆ lass)
arethemaximumandguaranteedbitrate,theresidualerrorrate,thedelay,thehandling
of prioritiesand thein-order-deliveryof informationentities.
1.3.3 Targeted QoS metri s
Inthisreport, thework fo usesontheperforman eofthepa ketdataa ess. Forthis
reason, we hosean intera tive servi e inorder to performour investigation. The QoS is
measuredinterms ofguaranteeddata rate,residualerrorrate and thetransfer delay per
informationentity,hen epertransmittedpa ket.
Asa real timeappli ation,thespee his hosen. Hen e,a onversational lassservi e.
The QoSattributeweare interested inforspee h isthe residualerrorrate.
The denitionof this QoS metri s in UMTS an be found in[61 ℄. The way they are
measuredin oursimulationsis going to be des ribedin the orresponding hapters. The
optimizationwork donehereaimsat these QoS attributes.
1.4 Pa ket Data TraÆ Modeling
TraÆ modelingis very important when studyingRRM algorithmsand onsequently
inourstudy. Considering,thattheaimhereistoin reasesystem apa ityandtoimprove
theQoS ofthe users,itis importantto modela urately thetraÆ from servi es.
periods, with the a tivity period emulating the time a user speaks and the ina tivity
period emulating the silen e. The data rate during a tivity periods is onstant. This
model is introdu edby Brady in[8 ℄ and it is known as the`on-o' model. Variations of
thistraÆ modelhave beenproposedforUMTS in[? ℄and in[78 ℄ .
packets
packet call
reading time
inter-arrival time
packet session
time
Figure1.4: Generi web browsingmodel.
Newservi esand onsequentlynewtypesoftraÆ in3Gsystemshaveurgedaresear h
eort on traÆ modeling. The fo us has beenturned on the the modeling of the traÆ
originatingfromWorldWideWeb(WWW)relatedappli ations. Theinitialapproa hhas
been to model WWW traÆ in UTRAN, based on traÆ hara terizations arried out
in wired networks. Although this approa h seems heavily questionable, onsidering the
limitationsimposedbythe airinterfa eand theintermediateme hanismsset inorder to
mitigatethem(e.g. theWirelessAppli ationProto ol,WAP),itisaratherfairone,sin e
3G radionetworks will also be used for internet a ess from portable omputers, where
`wired' WWW modelsareappli able. Thus, onthe basisofthe work donein[2 ℄and [41 ℄
the"webbrowsing"modelof[83 ℄hasemerged. Therein,WWWtraÆ ismodeledbyusing \pa ket sessions", \pa ket alls" and \reading time" periods, as it an be seen in gure
1.4. A pa ket session starts when a user starts WWW browsing session and ends when
hestopsdownloadinginformation. Ea h pa ketsessionis onsistedofanumberofpa ket
alls. A pa ket allemulates thedataburstoriginatingfrom a web pageand thereading
timethe ina tivityperiodbetweentwopa ket alls. This time orrespondsto theperiod
during whi h the user is reading the downloaded web page before sending a request for
thenext page. Withina pa ket all, datagrams, orpa kets, arriveat ertaininter-arrival
time. The above parameters, i.e. pa ket allduration and size, reading time, datagram
size and inter-arrival time, follow ertain distributions with ertain statisti s. In ea h
model, these statisti s and distributions are tailored in order to emulate hara teristi s
of the servi e to be modeled, e.g. data rate during a tivity, burstiness, round trip time
delays,et . Therefore,thestatisti softheweb browsingmodelsareheavilyin uen edby
traÆ is hara terized by\heavy-tail"distributions, whi h explainstheimportant traÆ
varian e (burstiness) even in large time s ales ([38℄). Therefore, the pa ket all size in
the web-browsing model is usually hara terized by heavy-tailed distributions su h as
Pareto. Forsimulation reasons however usuallytrun atedheavy-tail distributionsor
fat-taildistributionssu hasthelognormal areusedinorder to des ribethepa ket allsizes.
Moreover, these parameters an be modied inorder to modeltraÆ emanatingfrom
otherthanwebbrowsingservi es. Asanexample,in[6 ℄,thedes ribedabovetraÆ model
istailored inawaythat emulates thetraÆ from web gaming.
A familyoftraÆ modelsthatarebasedon theweb browsingmodel,hasbeen reated
the last years, as an example in [25℄ and in [42 ℄. They aim at in reasing the modeling
a ura y oftheexistingwebbrowsing modelbyessentially modifyingthedistributionsof
its hara teristi s. These distributions result from the measured web traÆ . A generi
web browsing model that onsiders the nature of TCP traÆ and its pro edures (slow
start, ongestionalgorithm,et ) hasbeenproposedin[1℄.
The webbrowsingmodelhoweverappliesforthedownlink. Forwebbrowsing
appli a-tions, the traÆ inuplinkis mainly onsistedof TCPlayera knowledgements, resulting
thus in a low data rate. In order to study the performan eof the pa ket data a ess in
uplinka servi e resulting in higher data rates is required. Measurements and studies of
uplink traÆ have shown that the servi e that generates the highest data rates in
up-link is the network gaming [60 ℄. On the basis of the statisti s of network gaming ([6 ℄),
a generi traÆ modelforintera tive servi es inuplinkhasbeenproposedwithin3GPP,
theso- alled \modied gamingmodel" ([78℄). This modelwithsome slight modi ations
hasbeenusedinourwork. Fortheeasinessofthereader,its hara teristi sarepresented
inthenext haptersalong withtheother simulation assumptions.
1.5 Aim of the PhD Thesis
Aim
In this thesis our obje tive is to propose methods and algorithms that optimize the
utilization fa tor of resour es allo ated to pa ket dataservi es inthe uplinkdire tion of
transmissioninWCDMA. This willimprove theperforman eof pa ketdataservi es, i.e.
itwillin reasesystem apa ityanditwillimprovetheexperien edQoSofusersinuplink.
Ba kground
Current internet appli ations su h as web browsing are hara terized by the
lient-server model;usersissuerequests to remote servers forlesorfor any othertypeof data
that is lo ated in them ([56℄). Hen e, data is downloaded from the network to users in
mostof the ases. Asa result,in pa ketswit hed networksthe largest part of thetraÆ
is observed in the network-to-users dire tion. Consequently, the downlink dire tion of
transmission in UMTS is expe ted to exhibit higherload than the uplink, with the last
this reason, in the last years the resear h eort was primarily dire ted to the downlink
dire tion of transmission with an outstanding example the implementation of the High
SpeedDownlinkPa ketA ess (HSDPA).
The expe ted in rease of load in the downlink and the introdu tion of new servi es,
su h as the MMS, however, raise the demand for a orresponding improvement in the
uplinkdire tionof transmissionofpa ketdata. In thatrespe t,feasibilitystudiesforthe
enhan ement of the uplinkpa ket data transmissionhave started in 3GPP [78 ℄.
Candi-datefeatureshavebeensuggestedinalmostallofthelayersandforalloftheme hanisms
involved in the UL pa ket data a ess for both the UTRAN and the UE. Some of the
proposedfeaturesare: advan edUEtransmittersandre eivers,substitutionoffastpower
ontrolfun tionfromadaptivemodulationand oding(AMC)te hniques(forthe
mainte-nan eoftheradiolink),improved odings hemes, optimizationofthepa kets heduling,
shorter transmissiontimeintervals and fastdedi ated hannel establishment and release,
aswellashybridautomative repeatrequest (HARQ) me hanisms.
Studied Features
Amongtheproposedfeatures,improvementsinthepro edureofformatsele tionforthe
pa ket data transmissionand in pa ket s heduling are dis ussedhere. Asit is presented
above, the rst one is a me hanism of sele tion of the appropriate format, essentially of
thedata rate,among theallo ated onesto theUE at ea h transmissioninterval. Forthe
uplinktransmission, thisme hanism is lo ated in the UE. Given that one of the riteria
of a format sele tion are the radio propagation onditions, the TFC sele tion is thus a
radio link adaptation me hanism. In ase of multi-servi e transmission, the me hanism
performs the s heduling of the allo ated to theUE resour es to its dierent servi es. In
3GPP do umentation, e.g. [73 ℄,it is alledTransportFormat Combination (TFC)
sele -tion.
The pa ket s heduling in the uplinkdire tion of UTRAN is of parti ular interest, as
dis ussed above. The pa ket s heduler grants \ apa ity rights" to UEs, by allo atingto
them a set with possible data rates they an transmit, sin e s heduling on a pa ket by
pa ket basis is not feasible, mainlydue to delays in the radiointerfa e. The resour e to
bes heduledin theuplinkis there eived powerinthe NodeB([58 ℄).
Our goal is to improve these two pro edures without applying major hanges in the
overall pro edure ofthe uplinkpa ket dataa ess. Theexpe ted gainfrom the
improve-ment in the pa ket s heduling is the in rease in system apa ity and onsequently the
improvement of the QoS of the pa ket data servi es. The expe ted gain from the
im-proved TFC sele tion me hanism under dis ussion in this thesis is that the addition of
pa ketdata servi esis notdegradingthe performan eof theexisting real timeones.
The guidelinesfora TFC sele tion algorithm forthe uplinkdire tionof transmission
have been des ribed in the 3G standard [73 ℄. The main hara teristi of this algorithm
isthatit favors thetransmissionofthehighestpriorityreal-timetraÆ overpa ketdata.
Inthe ase ofsimultaneous transmissionofareal timeservi eand apa ketdataone,the
probabilityoflossin reasesin omparisonto the aseofsinglereal-timetransmission,due
ee t. For this reason, a simulator emulating the uplinkmulti-servi e transmission of a
single UEhas beenimplemented.
The performan e of the uplink pa ket s heduling has been investigated by means of
system level simulations. Performan e evaluation riteria are the system apa ity and
theoered QoS to users. We proposethree features that aim at improvingthe eÆ ien y
of the pa ket s heduling and hen e at in reasing system apa ity. In the rst feature,
UEs are hanging rapidly the spreading ode and hen e the spreading fa tor they use
for their transmission in order to minimize the varian e of their transmission power. In
thisdissertation,thisfeatureis alledFast Variable SpreadingFa tor (VSF). These ond
featureproposesan uplinkpa ketdata a ess whereUEs parti ipatemore inthe ontrol
of the overall pro edure. Therefore, this a ess mode is alled de entralized. The third
studiedfeatureis thefasterpa ket s heduling.
Asso iated Work
In order to performourinvestigations,westudied and implementedpa ket datatraÆ
models, we analyzed the UTRAN proto ol ar hite ture with an emphasis on the
me h-anisms that realize the uplink multi-servi e transmission. We have also implemented
features related to the propagation in the physi al layer su h as radio hannel models,
power ontrol (PC)me hanisms and re eiver antenna diversity. Radio resour e
manage-ment(RRM)me hanismssu hasload ontrol, softhandover(SHO) ontrolarealso take
into a ount in our studies. Moreover, in order to assess the system performan eand to
analyzeresultswe studied on eptssu has ellinterferen eand theratioofotherto own
interferen e, theso alledi-fa tor.
Related Literature
Radioresour emanagementin3G ellularsystemshasre eivedalotofattentioninthe
resear h ommunityduringthelastyears. Mostoftheeorthoweverhasbeendire tedto
the downlink dire tion of transmission. Therefore theliteraturerelated to issuesstudied
here,su h aspa kets heduling, indownlink iswide, e.g. [39 ℄,[27 ℄, [9℄,[19 ℄,[3 ℄.
The systemmodeling forthe studyof pa ket s heduling algorithmsbothindownlink
anduplinkisdes ribedin[52 ℄andin[35℄. Inaddition,in[52℄apa kets hedulingalgorithm
based on traÆ priorities is presented and its performan e is assessed. In [35 ℄ various
pa kets hedulingalgorithms\aredesignedbasedondierentphilosophies: rate orutility
optimal, fairness-based, and user-oriented". Obtained results show that \the
fairness-based approa hes seem to be a good ompromise and are good andidates for pra ti al
implementations in a real system". Uplink pa ket s heduling and its performan e have
been studied also in [?℄, [48 ℄, [23 ℄, [16 ℄, in [57 ℄ and in [34 ℄. The modeling of the pa ket
s heduling is addressed also in [?℄. Therein, the impa t of pa ket data traÆ on the
performan eof spee h servi esis also dis ussed. A pa ket s hedulingalgorithm basedon
loadandinterferen einthesystemisproposed. Pa kets hedulinginthe aseofasystem
with two lasses of users \delay-tolerant and delay-intolerant users". Two transmission
users an transmit at a ertain time. Results show that this time-s heduling approa h
aords a better ell throughput. In [23℄a pa ket s heduling algorithm for dualtraÆ in
UMTS is presented. Data rate allo ations for the \pa ket traÆ " are performed on the
basis of interferen e in the ell and of user's transmission power in reases the usage of
radio resour es. In [34 ℄ the performan eof a number of pa ket s heduling algorithms is
assessed. The major observation is that s heduling users with good hannel onditions
one-at-a-timeand useswithbad hannel onditionsinlargergroupsoersthehighest ell
throughputamong thestudiesalgorithms.
Radio linkadaptation throughformatsele tion inuplinkhasbeenstudiedin[17 ℄and
[46 ℄. AnumberofTFCsele tionalgorithmsforpa ketdataservi eshasbeendis ussedin
[46 ℄. Algorithms aiming at guaranting a ertain data rare fora servi e orat maximizing
the user transmitted data rate are proposed. The performan e of the dis ussed TFC
sele tionunderdierentadmission ontrolpoli iesisinvestigated. DierentTFCsele tion
algorithmsinuplinkalongwithotherRRMfun tions,su hasload ontrolhavealso been
treated in [51 ℄ and in [53℄. The dieren e being that in [53 ℄ a s enario of mixed real
and non real timetraÆ is studied. The sele tion of TFChas been dis ussedalso in[4 ℄,
witha fo us onthe impli ationsof theTFC sele tion me hanism onto the physi al layer
performan e.
1.6 Overview of the PhD Dissertation
The rest of the dissertation is organized into ve hapters. The main ontributions
of the thesis arepresentedinthe three hapters rangingfrom hapter 3 to hapter 5. In
hapter 2 a briefoverview of the UMTS Terrestrial Radio A ess Network (UTRAN) is
given. Therein, the aim is to introdu e the UTRAN elements that are needed for the
understanding of the next hapters. The overall on lusions of this thesis are dis ussed
in hapter 6. Ea h hapter starts with a small paragraph introdu ing the issues to be
dis ussedinthatdesignated hapter anditendswithaparagraph ontaining on lusions.
Figure 1.5illustratesthegeneraloutline ofthe mainbodyofthisdissertation.
An outline of ea h hapter follows:
In hapter 2, a summary of UTRAN is given and its layered proto ol ar hite ture
is then des ribed. A brief overview of the physi al layer is given as well. Radio bearer
ar hite tureisalsodis ussed. Inaddition,theUEar hite tureinthe aseofmulti-servi e
uplinktransmissionis presented.
Chapter 3 gives anoutline of theMedium A essControl(MAC) proto ol. After the
presentationofthemainfun tionsoftheMACproto ol,we fo usonitsrole intheuplink
transmission;inparti ular,onthelinkadaptationissuethroughappropriateTFCandthus
dataratesele tion. Westudytheperforman eofthestandardizedTFCsele tionalgorithm
and we then dis ussalternative TFCsele tion algorithms. We omparethe standardized
TFC sele tion algorithm to one alternative algorithms with the help of dis rete event
simulations.
An analysisof theuplinkpa ketdataa essfollowsin hapter4. The prin ipal
me h-anismsinvolvedintheuplinkpa ketdataa essareidentied;theirimpa tontheoverall
over-Enhancements of the Uplink Packet Data Access in WCMA
General UTRAN
Description: Protocol
Architecture and
Radio Access
Bearers
Transport Format
Combination
Selection
Uplink Packet
Data Access in
WCDMA
Enhancements of
the Uplink Packet
Data Access
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Protocol
Architecture
MAC Protocol
Packet
Scheduling
Fast VSF
Physical Layer
Overview
Radio Bearer
Architecture
TFC Selection
Decentralized UL
Packet Data Access
Fast Packet
Scheduling
Figure1.5: Planof Dissertation.
all performan e is dis ussed. System performan e is evaluated by means of simulations.
The network level simulator therefore built is then shortly des ribed. Next, we list the
performan emetri sweuse andat theendwe presentand analyze simulation results.
Possible enhan ements in the uplinkpa ket data a ess are proposed and evaluated
in hapter 5. First, the fast variable spreading fa tor feature is dis ussed. Then, an
alternative mode of uplinkpa ket data transmission, where the ontrolof thepro edure
is more distributedthan in the standardized one is presented. Therefrom, this mode is
alledde entralizedinthisreport. Attheend,theimprovementthroughfasters heduling
isassessed.
The main on lusionof thedissertation and thefuture work are presentedin hapter