A Perspective on Radio Resource Management in B3G

Transcription

1 A Perspecive on Radio Resource Managemen in B3G Oriol Sallen Universia Poliècnica de Caalunya (UPC) Absrac.- Beyond 3G usually refers o heerogeneous scenarios where differen Radio Access Technologies (RATs) coexis and operae in a coordinaed way. This brings a new challenge o offer services o he users over an efficien and ubiquious radio access. In his way, he user can be served hrough he RAT ha fis beer o he erminal capabiliies and service requiremens, and also a more efficien use of he radio resources can be achieved. This challenge calls for he inroducion of new Radio Resource Managemen () algorihms operaing from a common perspecive ha ake ino accoun he overall amoun of resources offered by he available RATs. In his conex, his paper presens he framework for developing algorihms in he B3G scenarios, including some possible approaches. I. INTRODUCTION In parallel wih he developmen of he 3G cellular sysems, oher wireless access echnologies for Wireless Local Area Neworks (WLAN) and Personal Area Neworks (PAN), like IEEE and Blueooh, have been developed, sandardised and have experienced a significan growh, arriving o he mass marke. In urn, in he field of cellular sysems, he exension of GSM (Global Sysem for Mobile communicaions) o GPRS (General Packe Radio Service) including packe ransmission capabiliies in he radio inerface has been a firs milesone in he evoluion pah of 2G cellular sysems owards UMTS (Universal Mobile Telecommunicaions Sysem). The co-exisence and ineracions beween UMTS and GSM/GPRS echnologies consiue one of he key poins for he success of 3G echnologies. As a maer of fac, GSM/GPRS has also followed is independen pah wih he developmen of an improved radio access echnology ha allows higher bi raes hanks o he use of more efficien modulaion schemes. The erm EDGE (Enhanced Daa raes for GSM Evoluion) is used o refer o his improved sysem and he erm GERAN (GSM/EDGE Radio Access Nework) is he name of he evolved radio access nework including hese new capabiliies. As a resul of he above, he scenarios where UMTS will be deployed will probably differ from hose for which i was iniially hough, and i will have o co-exis no only wih previous 2G and 2.5G sysems bu also wih WLAN and oher emerging echnologies. These new scenarios where differen Radio Access Technologies (RATs) will coexis and will operae in a coordinaed way are ofen referred as beyond 3G (B3G) sysems. The availabiliy of several access neworks mus indeed be regarded as a new challenge o offer services o he users over an efficien and ubiquious radio access hanks o coordinaing he available RATs. In his way, no only he user can be served hrough he RAT ha fis beer o he erminal capabiliies and service requiremens, bu also a more efficien use of he available radio resources can be achieved [1]. This calls for he inroducion of new Radio Resource Managemen () algorihms operaing from a common perspecive ha ake ino accoun he overall amoun of resources available in he exising RATs, and herefore are referred o as C (Common Radio Resource Managemen) algorihms [2][3]. Furhermore, for a proper suppor of such algorihms, suiable nework archiecures and procedures mus ensure he desired inerworking capabiliies beween he differen echnologies [4]-[6]. The inerworking archiecure enhanced wih C funcionaliy will pave he way for he exension of hese heerogeneous neworks o include also new 4G radio access echnologies. This paper provides he framework for coping wih he radio resource managemen problem in beyond 3G scenarios, describing firs he differences in how he problem is deal up o 3G sysems and hen how he C funcionaliies are inroduced in beyond 3G sysems. Also some specific approaches o deal wih he problem are analysed. The res of he paper is organised as follows. Secion II and Secion III discuss he problem for 3G and beyond sysems, respecively. Secion IV presens he C funcional model devised in 3GPP (Third Generaion Parnership Projec) and he funcionaliies are discussed in Secion V. Finally, Secion VI presens some sample C soluions and he conclusions are summarised in Secion VII. II. RADIO RESOURCE MANAGEMENT IN 2G AND 3G SYSTEMS A Radio Resource Uni (RRU) can be defined by he se of basic physical ransmission parameers necessary o suppor a signal waveform ransporing end user informaion corresponding o a reference service. These physical ransmission parameers depend on he muliple access echnique being used. In paricular, in FDMA (Frequency Division Muliple Access), a radio resource uni is equivalen o a cerain bandwidh wihin a given carrier frequency. In TDMA (Time Division Muliple Access), a radio resource uni is equivalen o a pair of a carrier frequency and a ime slo. In CDMA (Code Division Muliple Access), a radio resource uni /06/$ IEEE

2 is defined by a carrier frequency, a code sequence and a power level. The main difference arising in CDMA is ha he required power level necessary o suppor a user connecion is no fixed bu depends on he inerference level. Thus, he amoun of ransmied power resources will vary along ime according o muliple elemens of he scenario, such as propagaion condiions, inerference, cell load level, ec. In addiion o he main physical dimensions (frequency, ime slo, code sequence and power level), here are oher physical ransmission elemens such as he modulaion scheme, channel coding scheme, ec. Clearly, depending on he exploiaion of he basic dimensions in erms of he former elemens differen specral efficiencies may follow. The objecive of a nework operaor is he deploymen of a nework able o suppor is cusomers wih he required QoS under he arge coverage area. Focusing on he radio par, he oupu of he radio nework planning will be he provision of RRUs along he service area by means of a cerain radio nework opology and a given configuraion of he cell sies. As long as e.g. service peneraion and service usage vary along ime and space, he amoun of radio resource unis o be provisioned also varies and, consequenly, he radio nework planning is an evolving process. Afer he RRUs have been provisioned in he service area during he planning phase, funcions are in charge of allocaing and managing he provisioned RRUs. Cellular mobile communicaions are dynamic in naure. Dynamism arises from muliple dimensions: propagaion, raffic generaion, inerference, ec. Thus, he dynamic nework evoluion calls for a dynamic managemen of he available RRUs, which is carried ou by means of mechanisms wih an associaed number of parameers ha need o be chosen, measured, analyzed and opimized. Besides, mechanisms may overcome a some exen he long erm reaciviy in radio nework planning and deploymen, which oherwise would preven he nework operaor o accommodae sudden and ransien raffic increases [2][7]. funcions should hen ake ino accoun he consrains imposed by he radio inerface in order o make decisions regarding he seing of he differen elemens and parameers influencing he air inerface behaviour. Some of hese are he number of acive users, he number of simulaneous users ransmiing, he corresponding ransmission raes for each user, he ransmied power levels corresponding o every simulaneous user, ec. Clearly, he number of parameers o be conrolled as well as heir differen naure claims for a se of several funcions, whose join behaviour should lead o an overall radio access nework opimisaion. The need of a conrolled air inerface usually requires a srong cenralised componen in he operaion. Neverheless, he higher signalling load associaed o a cenralised operaion usually requires having some decenralised elemens locaed in he erminal side. algorihms are fed by measuremens, some of hem provided from he erminal (e.g. serving and neighbouring cells received power levels) and some ohers from he Radio Access Nework (RAN) side (e.g. serving and neighbouring cells load level). Decisions aken from cenralised algorihms are ransferred o he erminal hrough layer 3 proocol procedures. funcions need o be consisen for boh uplink and downlink, alhough he differen naure of hese links inroduces some differences in he followed approach. Since he differen funcions will arge o rack differen radio inerface elemens and effecs, funcions can be classified according o he ime scales hey use o be acivaed and execued. Then, he se of funcions wih he corresponding ypical ime scales beween consecuive acivaions of he involved algorihm would be [2]: - Order of 1ms: Inner loop power conrol in CDMA - Order of 10ms: Packe scheduling, MAC (Medium Access Conrol) algorihms - Order of 100ms - 1s: Admission conrol, Handover, Congesion conrol, Ouer loop power conrol in CDMA funcions can be implemened in many differen algorihms, his impacing on he overall sysem efficiency and on he operaor infrasrucure cos. Addiionally, sraegies are no subjec of sandardisaion, so ha hey can be a differeniaion issue among manufacurers and operaors. The specific sraegies are leading o an increase in compeiiveness among he manufacurers. sraegies of legacy neworks (e.g. GSM/GPRS) are of raher low dimensionaliy, i.e. only a few parameers are needed o une heir opimaliy. On he conrary, in he case of UTRAN (UMTS Terresrial Radio Access Nework), he required flexibiliy o accommodae differen service requiremens ogeher wih he more sophisicaed naure of he WCDMA echnique urned ino more dimensions in he opimisaion problem hus increasing he complexiy of he problem. Finally, WLANs are also expeced o play an imporan role in he provision of high daa rae services, so ha should be argeed if QoS needs o be provided. III. RADIO RESOURCE MANAGEMENT IN B3G SYSTEMS In a B3G where several RATs coexis he managemen of he provisioned RRUs can be seen as a problem wih muliple dimensions. Every RAT is based on specific muliple access mechanism exploiing in urn differen orhogonal dimensions, such as frequency, ime and code. Then, local mechanisms are needed for every considered RAT: GERAN (GSM/EDGE Radio Access Nework), UTRAN and WLAN, as shown in Fig. 1 (a). In addiion o ha, a proper coordinaion among he differen RATs can be achieved wih developmen of Common (C) algorihms. C is based on he picure of a pool of radio resources, belonging o differen RATs bu commonly managed, as shown in Fig. 1 (b). Then, he addiional dimensions inroduced by he mulipliciy of RATs available provide furher flexibiliy in he

3 way how radio resources can be managed and, consequenly, overall improvemens may follow. Noice ha he C vision allows also considering differen amouns of radio resources spaially available, because in erms of curren nework deploymen deploymen, differen spaial availabiliies are found for he exising RATs. For example, GERAN ends o be he mos widespread RAT, while UTRAN is no ye everywhere deployed and in urn WLAN hospos wih reduced coverage areas are also widespread around ciies. GERAN c UTRAN WLAN -GERAN -UTRAN -WLAN GERAN c (a) C UTRAN WLAN (b) Fig. 1 a single RAT level managing orhogonal muliple access dimensions C managing a pool of orhogonal muliple access dimensions IV. C FUNCTIONAL MODEL The funcional model assumed in 3GPP for C operaion considers he oal amoun of resources available for an operaor divided ino radio resource pools. Each radio resource pool consiss of he resource unis available in a se of cells, ypically under he conrol of a RNC (Radio Nework Conroller) in UTRAN or a BSC (Base Saion Conroller) in GERAN. Two ypes of eniies are considered for he managemen of hese radio resource pools [4][5], as shown in Fig The eniy, which carries ou he managemen of he resources in one radio resource pool of a cerain radio access nework. This funcional eniy involves differen physical eniies in he RNS (Radio Nework Subsysem) or BSS (Base Saion Subsysem) depending on he specific considered funcions, alhough for represenaion purposes i is usual o assume he eniy residing in he RNC or he BSC. Noice ha differen eniies do no necessarily belong o differen radio access echnologies. - The C eniy, which is involved in he coordinaed managemen of he resource pools under differen eniies. In his way, decisions on radio resources usage may ake ino accoun he resource availabiliy in several eniies. Each C eniy conrols a number of eniies and may communicae wih oher C eniies as well, hus f f collecing informaion from oher eniies ha are no under is direc conrol. The ineracions beween and C eniies involve mainly wo ypes of funcions: a) Informaion reporing funcion The informaion reporing funcion allows he eniy o repor relevan informaion o is conrolling C. The reporing can be performed periodical or even-riggered, or even a a given insan, and i is oally up o C eniy s reques. The exchange of informaion is also possible beween differen C eniies in order o know he saus of heir corresponding eniies. There are mainly wo ypes of informaion o be repored o he C eniy: - Dynamic common measuremens on cells conrolled by a given RNC or BSC eniy. These measuremens include he curren cell loads, ransmied carrier power, he received oal wideband power, inerference measuremens, ec. - Saic informaion on cells conrolled by a given RNC or BSC eniy. This includes he knowledge abou he cell relaions (e.g. if hey are overlapped or if hey belong o differen HCS layers), he cell capabiliies (e.g. wheher a cell suppors GPRS, EDGE,...) he cell capaciies (e.g. he number of available ime slos) or he available QoS (e.g. maximum bi rae for a given service or average buffer delay) b) decision suppor funcion This funcion describes he way how he C and eniies inerac for aking decisions. For example, i is possible ha he C simply advises he eniy, so ha he remains as he maser of he decisions, and, on he conrary, i is also possible ha he C is he maser so ha is decisions are binding for he eniy. - Informaion reporing - decision suppor eniy C eniy eniy - Informaion reporing eniy Fig. 2 C funcional model C eniy - Informaion reporing - decision suppor eniy V. C FUNCTIONS As explained in secion II, he main funcionaliies arising in he conex of a single RAN are: admission and congesion conrol, horizonal (inra-sysem) handover, packe scheduling and power conrol. When hese funcionaliies are coordinaed beween differen RANs in a heerogeneous scenario, hey can be denoed as common (i.e. hus having common admission conrol, common congesion conrol, ec.) as long as algorihms ake ino accoun informaion abou

4 several RANs o make decisions. In urn, when a heerogeneous scenario is considered, a specific funcionaliy arises, namely RAT selecion (i.e. he funcionaliy devoed o decide o which RAT a given service reques should be allocaed). Afer he iniial RAT selecion decision, aken a session iniiaion, verical (iner-sysem) handover is he procedure ha allows swiching from one RAN o anoher. The successful execuion of a seamless and fas verical handover is essenial for hiding o he user he underlying enabling infrasrucure. Issues relaed o verical handover comprise scanning procedures for he erminal o discover available RANs, measuremen mechanisms o capure he saus of he air inerface in he differen RANs, verical handover riggers (i.e. he evens occurring in he heerogeneous nework scenario ha require he sysem o consider wheher a verical handover is acually required or no), verical handover algorihm (i.e. he crieria used o decide wheher a verical handover is o be performed or no) and proocol and archiecural aspecs o suppor handover execuion. Verical handover procedures from one RAN o anoher may be useful o suppor a variey of objecives, such as avoiding disconnecions due o lack of coverage in he curren RAT, blocking due o overload in he curren RAN, possible improvemen of QoS by changing he RAT, suppor of user s and operaor s preferences in erms of RANs usage or load balancing among RATs. Thus, he verical handover procedure enables anoher dimension ino he C problem and provides an addiional degree of freedom in rearranging raffic [3][8]-[10]. VI. SAMPLE C SOLUTIONS According o he framework presened in he previous secions, some specific base-line soluions are addressed coping wih he RAT selecion problem in heerogeneous scenarios. In paricular, a service-based, load balancing-based and radio nework-based sraegies for UTRAN/GERAN scenarios are presened as illusraive examples. a) Service Based RAT selecion A service-based RAT selecion policy is based on a direc mapping beween services and RATs [11]. As an example, in a scenario including voice and ineracive service assuming all erminals have muli-mode capabiliies (i.e. hey can operae eiher wih UTRAN or GERAN), wo possibiliies would be: - Allocaes voice users ino GERAN and ineracive services ino UTRAN. - Allocaes voice users ino UTRAN and ineracive services ino GERAN. If no capaciy is available in he primary RAT, he oher RAT is seleced insead. If no capaciy is available in he alernaive RAT, he service reques ges blocked (a service se-up) or dropped (during service life-ime). As deailed in [11]for a scenario wih seven omnidirecional cells wih radius 1 km for UTRAN and GERAN and assuming ha he cells of boh sysems are co-sied, i is found ha he policy allocaing voice users o GERAN performs beer in erms of oal aggregae hroughpu. The main reasons for his are wo-fold. Firs, wih respec o www users, a higher hroughpu can be obained in UTRAN as long as DCH channels are used while in GERAN www users are subjec o a scheduling algorihm. In urn, from he voice users poin of view, if he disance beween cell radius was se o 500m, no significan differences would be observed beween boh sraegies, bu when increasing he radius, a higher degradaion is observed in case voice raffic is allocaed o UTRAN because UTRAN users a he cell edge experience some erroneous ransmissions due o power limiaions and he inerference-limied naure of WCDMA. b) Load Balancing-based RAT selecion Load balancing (LB) is anoher possible guiding principle for resource allocaion in which he RAT selecion policy will disribue he load among all resources as evenly as possible. Specifically, he seleced RAT will be he one having he lowes load. Therefore, an influenial run-ime parameer in a load balancing decision-making procedure is he load meric. For UTRAN, an average of he cell load facor can be used, while in GERAN a useful way o measure he daa load is o measure he average amoun of ime slos uilized by GSM/EDGE services. When comparing he LB RAT selecion algorihm agains he service-based policy in a scenario wih seven omnidirecional cells for UTRAN and GERAN co-sied wih cell radius 500m, service-based policy exhibis higher dropping values han policy LB for high load levels. This is explained bearing in mind he load disribuion in GERAN induced by servicebased and LB policies. In paricular, for service-based he load is a is maximum value mos of he ime which implies a lack of flexibiliy in order o accommodae handover users being redireced o GERAN. Therefore, service-based may incur in more poenial dropping siuaions han in he case of LB policy appliance which presens more flucuaions in he load values and can provide resources o incoming handover users if necessary. c) Radio Nework based RAT selecion This caegory of policies for he RAT selecion procedure akes ino accoun he differen resource consumpion ha a given user may have in one or oher RAT depending on he specific nework characerisics. As an example, le consider in he following he so-called indoor policy. In his case he selecion would be done aking ino accoun wheher a user is locaed indoor or oudoor, under he consideraion ha WCDMA capaciy is highly degraded by indoor raffic users, as saed in [12], where capaciy reducions of up o 80% are observed when half of he users in a scenario are indoor. Consequenly, according o his policy indoor users will be

5 allocaed in GERAN while oudoor users will be allocaed in UTRAN. I is worh menioning ha oher similar policies could also be considered making use of e.g. pah loss measuremens (i.e. allocaing high pah loss users in GERAN and low pah loss users in UTRAN). Resuls shown in [11] reflec ha, when he indoor policy is applied, he BLER is significanly reduced. VI. CONCLUSIONS This paper has focused on he problem in heerogeneous wireless neworks where differen RATs coexis. The framework for developing C sraegies including he funcional model and a descripion of he major funcionaliies has been provided. Also differen approaches for he RAT selecion procedure have been presened as illusraive examples. VII. ACKNOWLEDGEMENTS This work is parially funded by he European IST projec AROMA ( and by he COSMOS gran (ref. TEC , Spanish Minisry of Science and Educaion and European Regional Developmen Fund). REFERENCES [1] G.Fodor, A. Eriksson, A. Tuoriniemi, Providing Qualiy of Service in Always Bes Conneced Neworks, IEEE Communicaions Magazine, July, 2003, pp [2] J. Pérez-Romero, O.Sallen, R.Agusí, M. Díaz-Guerra, Radio Resource Managemen Sraegies in UMTS, John Wiley & Sons, [3] A. Tölli, P. Hakalin, H. Holma, Performance Evaluaion of Common Radio Resource Managemen (C), ICC Conference, Vol. 5, April, 2002, pp [4] 3GPP TR v5.0.0 Improvemen of across RNS and RNS/BSS [5] 3GPP TR v0.3.0 Improvemen of across RNS and RNS/BSS (Pos Rel-5) (Release 6) [6] J. Pérez-Romero e al. Common Radio Resource Managemen: Funcional Models and Implemenaion Requiremens, 16h PIMRC Conference, Berlin, [7] J. Zander, S.-L. Kim, Radio Resource Managemen for Wireless Neworks, Arech House, [8] S. Lincke-Salecker, The Benefis of Load Sharing when Dimensioning Neworks, Proceedings of he 37 h Annual Simulaion Symposium (ANSS 04), April, [9] S. Lincke-Salecker, Performance and Service Issues in Selecing Adapive Placemen as a Load Disribuion Technique, IEEE 59h Vehicular Technology Conference, VTC 2004-Spring, Milan, [10] M. Sieber, M. Schinnenburg, M. Lo, Enhanced Measuremen Procedure for Verical Handover in Heerogeneous Wireless Sysems, 14h PIMRC Conference, Beijing, [11] J. Pérez-Romero, O. Sallen, R. Agusí Policy-based Iniial RAT Selecion algorihms in Heerogeneous Neworks, 7h MWCN Conference, Marrakesh, [12] J. Pérez-Romero, O. Sallen, R. Agusí, On The Capaciy Degradaion in W-CDMA Uplink/Downlink Due o Indoor Traffic, VTC in Fall 04 conference, Los Angeles, USA, 2004.