Context-aware Load Balancing for Heterogeneous Smart-Grid Communication Networks

Transcription

1 Inernaional Conference on Inelligen Sysems Research and Mecharonics Engineering (ISRME 05) Conex-aware Load Balancing for Heerogeneos Smar-Grid Commnicaion Neworks Yinao Li,a, Weigo Yan,b, Hongfei X,c, Wei Song,d, Ri Yang,e, Zhibin Zang,f, Dapeng Lin,g and Yeshen He,h Sae Grid Jibei Informaion & Telecommnicaion Company NARI Grop, China Gridcom CO,.LTD,Shenzhen,Gangdong,China a @qq.com, b wgy83866@sina.com, c 4734@qq.com, d @63.com, e lai_en@foxmail.com, f zangzhibin@sgepri.sgcc.com.cn, g lindapeng@sgepri.sgcc.com.cn, h heyeshen@sgepri.sgcc.com.cn Keywords:load balancing; conex-aware; heerogeneos wireless neworks; radio resorce managemen; smar grid. Absrac. The nework of smar grid commnicaions (SGC) is expeced o be heerogeneos for providing diverse access inerfaces. Load balancing is one of he key echnologies in sch heerogeneos neworks, which can increase he sysem hroghp as well as qaliy of service (QoS). In his paper, a conex-aware load-balancing scheme is proposed for SGC neworks by vire of an adapive rigger hreshold. We invesigae he characerisics of differen kinds of services in he wireless local access nework (WLAN) and Long Term Evolion (LTE). Simlaion resls show ha or proposed load balance algorihm achieves beer fairness and hs improves he overall hroghp. Inrodcion The pblic-iliy indsry envisions he radiional power grid o incorporae he advanced commnicaion sysems for improving he conrol, efficiency, reliabiliy and safey of power disribion and resorce managemen. The resling nex-generaion elecric grid, known as smar grid, is expeced o sppor a myriad of applicaions ha have heerogeneos raffic reqiremens. Coninosly increasing he demand of bandwidh, he inensive applicaions and he sers variable appeie for services of smar grid lead o a remendos pressre o he smar grid commnicaion (SGC) sysems. To address he wide variey of applicaions, scenarios, regions, environmens and opologies in smar grid, mliple radio access echnologies (RATs) sch as wireless local access nework (WLAN), Long Term Evolion (LTE) and Worldwide Ineroperabiliy for Microwave Access (WiMax) are ineviable o coexis. Therefore, wih he increasing amon of sers and services, an effecive load balancing approach in heerogeneos SGC neworks is becoming exremely imperaive. Some previos lierares have addressed he heerogeneos SGC neworks. The se of mliple RATs in a smar grid was firs proposed in [] o sppor he varying raffic reqiremens for overhead ransmission line monioring. A he nework layer, archiecres and frameworks o sppor sch heerogeneos neworks have been sggesed in []-[4]. A srvey of hese archiecres is provided in []. Recen proposals are based on he Media Independen Handover fncion defined by he IEEE 80. sandard, which provides a framework o sppor seamless ransiion across neworks based on differen RATs [3]. Anoher relevan sandard, IEEE P900.4, defines bilding blocks for enabling coordinaed nework-device disribed decision making ha conribes o he opimizaion of radio resorce sage, inclding he specrm access conrol, in heerogeneos wireless access neworks [4]. Hierarchical resorce managers have been proposed by he Common Radio Resorce Managemen, Join Radio Resorce Managemen, and Mli-access Radio Resorce Managemen schemes sdied by he 3GPP grop. And he emerging sandards from he IETF and 3GPP commniies o sppor flow mobiliy, when a device swiches is daa flows over mliple 05. The ahors - Pblished by Alanis Press 907

2 wireless access sysems are proposed by [5]-[6]. However, none of he aforemenioned lierare ps forward an effecive load balancing mehod for mliple RATs heerogeneos neworks in smar grid. In his paper, a conex-aware load balancing scheme is proposed for heerogeneos SGC neworks. Boh he fairness and he overall hroghp are invesigaed for demonsrae he enhancemen of he proposed approach. The opimal radio resorce allocaion is deermined for exploring he resorces from eiher celllar base saion (BS) or access poin (AP). User WLAN AP User 5 User WLAN AP 3 LTE BS User 3 WLAN AP M User N User 6 User 4 WLAN AP Sysem Model Figre. Heerogeneos Nework Model in SGC A. Nework Model The heerogeneos SGC nework model nder or consideraion is depiced in Figre, which consiss of one LTE celllar BS and several WLAN APs. And all he sers are locaed in he region covered by LTE BS and WLAN Aps, which can only connec o one RAT simlaneosly. The sers of differen kinds of service in he overlapping area can access o boh RATs, which wold make a big difference on he load condiion of he nework if sers make differen selecion. To acqire he services effecively, each ser shold obain he corresponding physical resorce blocks (PRB) from BS or APs. Moreover, in order o ensre he qaliy of service (QoS) of each ser, a minimm daa rae is se for each service. In addiion, we assme ha differen PRBs lie in differen freqency bands, hs he inerference beween any of wo PRBs cold be ignored. B. Conex Informaion In order o manage PRB in heerogeneos neworks across smar grid efficienly, we need he relevan conex informaion from boh LTE BS and WLAN APs. We consider he following conex informaion in or model. ) LTE BS Load Condiion: We assme ha he average received signal o inerference pls noise raio (SINR) for ser from cell c a ime SINR is deermined by he average channel gain, which is wrien as pc L( p) SINR = () N + ρ p L ( p ) c X( ) c c where L( p) is he average pahloss wih shadow fading from he crrenly serving cell and oher inerference cell, p c denoes he ransmi power for cell c, ρ c represens he load of he cell c and N is he power of addiive whie gassian noise, respecively. Given he SINR, we find he daa rae per PRB by sing he Shannon formlar. Assming ha he adapive coding and modlaion is sed o chieve he Shannon rae limi, he qaniy of resorces allocaed o ser by cell j can be wrien by D N = () R( SINR) B where R( SINR) = log ( + SINR ), D represens he bi rae of ser and B is he bandwidh of per PRB. Therefore, we can acqire he oal load of cell c by D ρc = ρ = (3) ( ) = ( ) = N R( SINR ) B X c X c oal 908

3 ) WLAN AP Load Condiion: We assme ha he imm capaciies of differne WLAN APs are idenical. In his paper, he load of WLAN AP ρ is defined as he raio of he nmber of crren sers accessing o WLAN AP ( U ) o he imm nmber of sers ha can access o he WLAN AP ( U U m ). The load is presened as m ρ W = (4) U 3) WLAN AP Conneciviy Index: The conneciviy index is o measre wheher he sers in he coverage of he WLAN hospos is appropriae o having access o he WLAN AP. Large conneciviy index indicaes ha sers appropriae o access o WLAN APs. The conneciviy index incldes pars: he mobile index and he raio of received signal power of WLAN APs. The laer is ha sers received signal power compared o he imm inp signal power lever of ser from WLAN hospos. We assme ha he imm inp signal power levels of ser are he same o all sers. The formla of he conneciviy index can be expressed as pm, km, = γ γ m, + γ ( γ + γ =, p m, > p ) (5) p where p represens he received power srengh in ime slo from he WLAN AP m, p m, is he received power hreshold, p denoes he imm inp signal level ha received by sers, which is a consan, and γ, γ represen he weigh facors of mobile index and he received WLAN signal power raio, respecively. The higher conneciviy index of ser indicaes he ser is proper o access o he WLAN AP; oherwise, he ser is proper o access o he LTE BS. Problem Formlaion A. Load Balancing Index Load balancing aims o achieve a fair disribion of sers among cells. I is desirable o have a mehod or an index ha can be sed o describe he degree of fairness. Typically, Jain s fairness index is sed as a meric o assess he level of load balancing in a nework. Since Jain s fairness index is sed as a load balancing sandard in he homogeneos nework, we se i o measre he load fairness among differen WLAN hospos. The Jain s fairness index is indicaed by ( ρ j ) β =. (6) n ( ρ j ) In order o make he load well-disribed in he wo neworks, we se ε which represens he raio of he wo nework load o measre he difference. We aim o make he ε close o, he ε can be wrien as U m ρ WLAN n ε = =. (7) ρc nu B. Adapive Balancing Threshold We se δ as he load balancing hreshold. If ρ LTE ρ WLAN δ, LTE BS load compared o WLAN AP is oo heavy, hs a load balancing process is reqired o conver he LTE BS load o WLAN AP. If ρ WLAN ρ LTE δ, which means ha WLAN AP load compared o LTE BS is oo heavy, we need o balance he WLAN AP. For a dynamical nework, a fixed balance hreshold obviosly canno ensre he opimizaion performance of he nework all he ime. So we need he δ change iself verss he load condiion. In his paper, we propose a herisic algorihm. The load balancing hreshold δ sars a a given vale. When a load balancing process has been compleed and he condiion of ρlte ρwlan < δ has no been me, he load balance hreshold δ becomes aδ, hs anoher load balancing process is reqired. Here α have wo vales. When δ is me, α is he small vale. When δ min is me, α is he larger one. w 909

4 These seings are made becase hey cold make a limi oδ, δ ( δ, δ ) min, so as o ensre he raionaliy of load balancing. C. Mahemaical Model In order o improve he load condiion of heerogeneos neworks in smar grid, we need o enhance he load raio and fairness, which will help improve he hroghp of he heerogeneos neworks. Therefore, we can formlae or problem in he following mahemaical model. ( ρ j ) ρwlan β = ; ε = n ( ρj ) ρlte s.. C: ρlte ρwlan δ (8) C: N < Noal X( ) = c C3: Um < U m [,, 3, M] In his mahemaical model, consrain C means he difference beween he LTE BS load and WLAN AP load is smaller hanδ, so he load can be disribed niformly in he heerogeneos neworks. Frher, consrain C denoes he nmber of PRB allocaed o sers is smaller han N oal, which is he oal nmber of PRB. Moreover, U is he nmber of sers accessing WLAN AP and m U is he imm nmber of sers ha an AP can serve. Consrain C3 ensres he nmber of he sers served by each AP is smaller han U. Conex-aware Load Balancing Algorihm According o he discssion above, we can propose a conex-aware algorihm o solve he load balancing problem in heerogeneos neworks across smar grid. Throgh or proposed algorihm, we can effecively achieve he load balancing in heerogeneos neworks. The oline of he proposed algorihm is depiced in Figre. sar Yes r r d c WLAN No r r d c WLAN No Yes Find he ser having minimm SINR Find he AP wih imm ilizaion efficiency Find he AP wih he imm conneciviy index for he ser Find he ser in his AP wih he minimm conneciviy index Handover his ser o he WLAN AP Handover his ser o he LTE BS Recalclae he and he r WLAN r c d = ad Figre. Conex-aware Load Balancing Algorihm Simlaion Resls In order o validae or proposed algorihm, we need o simlae a pracical scenario. In or scenario, here is one LTE BS and for WLAN APs locaed in he covered area. Meanwhile, we 90

5 assme ha he signal loss beween he LTE BS and sers inclding pah fading and freqency selecive fading can be calclaed by lg ( d ) + lg (.3)( db), and he pah loss model beween he sers and he WLAN AP is lg ( d ) + lg (.3)( db). Moreover, he simlaion parameers of he sysem are shown in TABLE I. TABLE I. SIMULATION PARAMETERS parameer vale parameer vale The coverage radis of LTE BS 500m Power Specrm Densiy of Noise 80 dbm Transmi power of LTE BS 40W WLAN bandwidh MHZ Nmber of Resorce blocks 00 Nmber of WLAN APs in a cell 4 LTE Resorce block Bandwidh 80KHZ The coverage radis of WLAN 40m Nmber of cells in a sysem 7 Transmi power of WLAN AP 0. W Shadow sandard deviaion 8 db Maximm ser nmber of each WLAN AP 0.8 Original Raio Algorihm Raio.6.4. Load Raio Nmber of Users Figre 3. Load Raio in heerogeneos neworks..8 Original Fairness Index Algorihm Fairness Index.6.4 Fairness Index Nmber of Users Figre 4. Fairness Index in heerogeneos neworks. x Original LTE Throghp Algorihm LTE Throghp Original WLAN Throghp Algorihm WLAN Throghp Original Toal Throghp Algorihm Toal Throghp. Throghp Nmber of sers Figre 5. Throghp in heerogeneos neworks. 9

6 Figre 3 shows he raio of he average WLAN AP load o LTE BS load in heerogeneos SGC neworks. Before he load balancing algorihm is applied, he fairness β in heerogeneos neworks is abo 0.4. Throgh or proposed load balancing algorihm, β can enhance o 0.9 approximaely, which indicaes ha he load beween LTE BS and WLAN APs is balanced since some sers have been ransferred o he WLAN APs. Figre 4 indicaes he fairness index in neworks. Wih he increasing of sers, he original index is arond 0.6. By or proposed algorihm, he fairness index in he neworks can be improved p o 0.9. Therefore, his implies ha he proposed algorihm can balance he load condiion among LTE BS and differen WLAN APs. The hroghp of he heerogeneos neworks is shown in Figre 5. As he nmber of sers in he LTE BS increases, he nework oal hroghp gradally increases. When he nmber of sers is arond 80, he oal hroghp is imized, where he proposed algorihm has improved he performance by 30%. Afer he load balancing algorihm, he WLAN hroghp increases while he LTE hroghp decreases, which is de o he fac ha some sers have been ransferred o he WLAN APs from LTE BS. Conclsions In his paper, a conex-aware load balancing scheme was proposed in heerogeneos SGC neworks. By he proposed algorihm, we can improve he load condiion of heerogeneos SGC neworks effecively wih sfficien conex informaion. Throgh he nework opimizaion realized by or proposed algorihm, we can acqire a higher oal nework hroghp, which garanees a beer environmen for smar grid commnicaion. References [] K. Hng e. al, On Wireless Sensor Commnicaion for Overhead Transmission Line Monioring in Power Delivery Sysems, Proceedings of IEEE SmarGridComm, Oc. 00. [] L. Gavrilovska, V. Aanasovski, Resorce Managemen in Wireless Heerogeneos Neworks (WHNs), Telsiks Ocober 009. [3] IEEE Sd , IEEE Sandard for Local and Meropolian Area Neworks, Par : Media Independen Handover Services, IEEE, Jan [4] S. Bljore e. al., Archiecre and Enablers for Opimized Radio Resorce sage in Heerogeneos Wireless Access Neworks: The IEEE Working Grop, IEEE Commnicaions Magazine, Jan [5] A. Oliva, C. Bernardos, M. Calderon, T. Melia, and J. Zniga, IP Flow Mobiliy: Smar Traffic Offload for Fre Wireless Neworks, IEEE Commnicaions Magazine, vol. 49, no. 0, Oc 0. [6]3rd Generaion Parnership Projec, Archiecre enhancemens for non-3gpp accesses, 3GPP TS 3.40, 0. 9