NTER-CELL LOAD BALANCNG TECHNQUE FOR MULT-CLASS TRAFFC N MMO-LTE-A NETWORKS 1 T.PADMAPRYA, 2 V. SAMNADAN 1 Research Scholar, Podicherry Egieerig College 2 Professor, Departmet of ECE, Podicherry Egieerig College E-mail: 1 padmapriyat185@gmail.com Abstract LTE etwork, load imbalace is the crucial issue which eeds to be hadled i order to exploit most of the beefit from LTE without degradig the throughput. this paper, we have proposed a ter-cell Load Balacig Techique for Multi-class Traffic i MMO-LTE-A Networks. First, we have proposed a robust load balacig framework to efficietly hadle the traffic ad also to keep the throughput as high as possible. To detect the overloaded cell, Call Blockig Ratio (CBR) is used as triggerig mechaism. Moreover, to efficietly balace the cogested cell detected by CBR Heaviest-First-Load-Balacig algorithm is implemeted to avoid cogestio i the traffic.. NTRODUCTON A. MMO-LTE Networks order to cope with the icreased traffic demad, the 3rd geeratio partership project (3GPP) has developed the Log Term Evolutio (LTE) stadard for 4G cellular etworks. This is based o Orthogoal Frequecy Domai Multiplexig (OFDM) waveform for dowlik (DL) ad Sigle Carrier Frequecy Domai Multiplexig (SC-FDM) waveform for uplik (UL) commuicatios. The key objectives of LTE etworks are user high data rates, reduced latecy, improved system capacity ad coverage, low complexity, reduced cost of operatio ad seamless itegratio with existig systems [1] [2]. 3GPP LTE etworks ca achieve high spectrum efficiecy due to the usage of multi-iput ad multioutput (MMO) atea ad orthogoal frequecy divisio multiple (OFDM) techology. There are may MMO schemes stadardized i 3GPP systems, ad the base statio scheduler has the capability to optimally select the MMO scheme that suits the chael coditios of the mobile. A fudametal MMO scheme is that of precoded spatial multiplexig (SM) where multiple iformatio streams are trasmitted simultaeously from the base statio to the mobile. These techiques are appropriate i high SNR areas with rich scatterig eviromets, i combiatio with suitable atea cofiguratios. However, the etwork performace is still iflueced by several factors, amog which itercell iterferece (C) ad load imbalace are two major oes [3] [5]. B. Load balacig i LTE Networks Load Balacig is defied as a automatic way to resolve the overloadig by shiftig traffic towards the light-loaded cells, by makig use of the radio resources more efficietly across the whole etwork. Oe possible way to balace the etwork load is to adjust the etwork cotrol parameters i such a way that overloaded cells ca offload the excess traffic to low-loaded adjacet cells, wheever available. However, this actio might itroduce additioal hadovers, which might cause bad hadover performace, leadig to the result that system would adjust hadover parameters to reorgaize the situatio, which might be i cotradictio to the aim of load balacig [4]. Load balacig scheme is required to miimize the demaded radio resources of the maximum loaded cell to avoid the traffic cogestio i log term evolutio (LTE) etworks. Load imbalace i LTE etworks deteriorates the system performace iflueced by ubalaced load distributio amog earby cells. Hece real-time iter-cell optimizatio is adaptable to eviromet especially whe ubalaced ad time varyig, is eeded [6] [7] [1]. ssues Hadover failure Load distributio Sigalig overhead Cogestio Radio resources [8] [1] [11]. LTERATURE REVEW Weyu L et al [4] have proposed a dyamic hysteresis-adjustig. With this proposed method, the two SON aspects load balacig ad hadover parameter optimizatio ca achieve a better coordiatio. The ew method tues the hysteresis accordig to a key idicator radio lik failure ratio, with realistic cosideratio, thus avoidig the possibility that load balacig has a bad ifluece o the etwork performace, for example, causig a higher radio lik failure ratio ad risk of jeopardizig the ormal fuctio of a etwork. With the proposed method, which is simple ad easy to realize, the etwork hadover performace ad load balacig 22
effect are both guarateed compared with covetioal solutios. However there occurs hadover failure. Hao Wag et al [6] have proposed etwork structure costraits ad a practical suboptimal algorithm, called Heaviest-First Load Balacig (). Usig the algorithm the etwork ca get sigificatly better load balacig while maitaiig the same etwork throughput at the price of a bit more hadovers compared with the traditioal sigal stregth-based hadover algorithm. However the load balace idex i mobile sceario is lower. Ad the radio resource cosumptio icreases. WANG Mi et al [7] have proposed a mi-max load balacig (LB) scheme to miimize the demaded radio resources of the maximum loaded cell. For the mixed multicast ad uicast services, multicast services are trasmitted by sigle frequecy etwork (SFN) mode ad uicast services are delivered with poit-to-poit (PTP) mode. The mi-max LB takes ito accout poit-to-multipoit (PTM) mode for multicast services ad selects the proper trasmissio mode betwee SFN ad PTM for each multicast service to miimize the demaded radio resources of the maximum loaded cell. Based o the solutio of this miimizatio problem, if the maximum loaded cell does ot overload, the mi-max LB will chage PTM mode ito SFN mode for multicast services to achieve high quality of service (QoS). The proposed mi-max LB scheme requires less radio resources from the maximum loaded cell tha SFN mode for all multicast services. L Bo et al [8] have proposed a iter-domai cooperative traffic balacig scheme focusig o reducig the effective resource cost ad mitigatig the co-chael iterferece i multi-domai Het-Net. The detailed implemetatio for the proposed traffic balacig scheme is desiged. the umerical evaluatio, the geetic algorithm (GA) as a optimizatio method is used to demostrate that the total effective resource cost is sigificatly reduced through their proposed iter-domai traffic balacig scheme comparig with the itra-domai traffic balacig scheme. The 43% of the resource cost is saved. The proposed scheme has great advatages i iterferece maagemet i Het-Net. However the cell-edge throughput ad the average cell throughput is ot icreased effectively. WANG Hao et al [9] have proposed a mobility load balacig (MLB) as a importat use case i 3GPP self-orgaizig etwork (SON), i which the servig cell of a user ca be selected to achieve load balacig rather tha act as the cell with the maximum received power. this paper, a uified algorithm is proposed for MLB i the LTE etwork. The proposed algorithm is evaluated for users with differet kids of QoS requiremets, i.e., guarateed bit rate (GBR) users with the objective fuctio of load balace idex ad o-gbr (GBR) users with the objective fuctio of total utility, respectively. The proposed algorithm leads to sigificatly balaced load distributio for GBR users to decrease the ew call blockig rate, ad for GBR users to improve the cell-edge throughput at the cost of oly slight deterioratio of total throughput. However with the larger arrival rates the more will be ubalaced loadig. Mig Li et al [1] have proposed a LTE virtualizatio framework (that eables spectrum sharig) ad a dyamic load balacig scheme for multi-enb ad multi-vo (Virtual Operator) systems. They compare the performace gai of both schemes for differet applicatios, e.g. VoP, video, HTTP ad FTP. They also ivestigate the parameterizatio of both schemes, e.g. sharig itervals, LB itervals ad safety margis, i order to fid the optimal parameter settigs. The LTE etworks ca beefit from both NV ad LB techiques. Omar Altrad et al [11] have proposed a geeral loadbalacig algorithm to help cogested cells hadle traffic dyamically. The algorithm is based o clusterig methods ad ca be applied to ay wireless techology such as LTE, WiMAX ad GSM. The algorithm ca be automatically cotrolled ad triggered whe eeded for ay cell o the system. t ca be implemeted i a distributed or semidistributed fashio. The triggerig cycle for this algorithm is left for the operator to decide o; the uderlyig variatios are slow so there is o eed for fast self-optimizig etwork (SON) algorithms. The distributio of the load of the cogested cell to its eighbor is oe step oly, which sigificatly reduces the sigalig overhead ad wastig of resources i the lightly-loaded cells compared to covetioal methods.. PROPOSED SOLUTON A. Overview From the existig works, we ca observe certai drawbacks, the hadover failures i DHA [4], the load balace idex i case of mobile scearios decreases [6], the waste radio resources is saved for the uecessary multicast services [7], due to periodic data collectio [8] leads to large-scale sigal chages ad sigalig overhead, the hadover [1] might cause packet loss problem, the reductio of cogestio [11] is almost less. this paper, we propose to desig a load balacig framework for MMO-LTE etworks. A load balacig framework is developed [6] which balaces the etire etwork load while keepig the etwork throughput as high as possible. Here by aalyzig the complexity of the optimizatio problem, etwork structure costraits are preseted, ad a practical suboptimal algorithm called Heaviest-First Load Balacig () is proposed. The mai objective of this framework is to make use of eforced hadover to balace the load betwee differet cells ad keep the etwork throughput as high as possible at the 23
same time.for the detectio of overloaded cells, call blockig ratio (CBR) [11] is used as the triggerig method. The the algorithm based o load balacig is ivoked ito the cogested cells [6]. B. Detectio of Overloaded Cell order to detect the overloaded cells, Call Blockig Ratio (CBR) is the real parameter which idicates the degradatio of the system whe ay overload occurs as show i Fig.2. Here, m is detected as the overloaded cell. our proposed solutio, CBR is used as a triggerig mechaism to ehace load balacig i LTE: CBR Where CBR= blocked calls/total accepted calls Also, is predefied threshold reserved for operator use which is decided by the quality of service (QoS) achieved by trasmissio mode of multicast services. Sice, P(t) ad are both determied by the allocatio betwee users ad cells, the problem is to V (t ) fid the optimal allocatio that maximizes for the curret timeslot t. Let load at each cell i at time slot t is give by L i (t) = bu(t) / b(t) (2) where b(t) ad bu(t) deotes the umber of PRB ad umber of used PRB at cell i. The average load of the etwork at time t is give by Li L1 i (t) = i N / N (3) The load balace idex ca be give by 2 = ( Li( t) L1 i( t)) i N (4) Defie a allocatio idicator variable m,, which is equal to 1 whe enodeb m allocates a physical resource block (PRB) to user at timeslot t or to otherwise. Hece the load defiitio of cell m ca be formulated as: m ( t ) N m, / b (5) Here b is the total umber of PRB. Represetig the allocatio by the ( ( ) :,, ) matrix m, t m E N t, hece the problem is equivalet to the followig maximizatio problem with (t): Figure 1. Detectio of Overloaded Cell C. ter Cell Load Balacig Usig Heaviest First Load Balacig This sectio first describes about the detectio of overloaded cell ad the sub optimal algorithm that efficietly balace the load without affectig the throughput. a) Defiitio of Utility Fuctio Give E enodebs ad N mobile users, we first fid a optimal assigmet betwee mobile users ad cells. For this, first we defie a utility fuctio i the multicell etwork which is give as below: V (, )( t ) P ( t ) ( t ) (1) Where ad are weightig coefficiets o etwork throughput ad load balace idex respectively. Differet values of ad i solvig the joit optimizatio problem i equatio (1) ca suitably be selected betwee the tradeoff betwee load balacig ad etwork throughput. P ( m p N m, m, is the etwork throughput at the timeslot t. P m, (t) is the available Shao rate at time slot t give by P m, (t) = W i,k log 2 + SNR i,k (t) (1) Where ) N Also, represets miimal throughput of each user. The costraits i equatio (7) represets that all cells have almost the same capacity limitatio, ad also the umber of user operated by oe enodeb ca t exceed the umber of its total PRBs. Costrait i equatio (8) represets that oe user ca oly be operated by oe enodeb at some specific timeslot t. Costrait i equatio (9) represets that user ca be operated by the enodeb which ca afford it a throughput value larger tha the Assumig that cell i hadovers a user k to a target cell j for load balacig, the followig coditio should be satisfied: u(j) + u(i) > u(j)+u(i) (11) 24
where u(j) ad u(i) are the updated values of idividual utility fuctios after hadover for cell j ad i, respectively. b) Heaviest First Load Balacig Algorithm each load balacig choose the heaviest loaded oe whose load exceeds the threshold to perform load balacig accordig to Heaviest-First Load Balacig which is described as below: //At the mth load balacig cycle// 1. Each ad every enodebs receive load status from its eighborig cells with CBR. 2. Cell m is the heaviest oe. 3. if load of cell m exceeds threshold, go to ext step 4. else stop. 5. cell m, fid user ad target cell c with the largest y,. m c 6. f it satisfies iequality (11), the switch user to cell c 7. Update other users gai i cell m, the go to ext step 8. Else stop. 9. f load status of cell m still exceeds threshold, 1. the go to step 5. 1. Else stop. V. SMULATON RESULTS A. Simulatio Model ad Parameters The Network Simulator (NS2) [12], is used to simulate the proposed architecture. the simulatio, 5 mobile odes move i a 12 meter x 12 meter regio for 5 secods of simulatio time. All odes have the same trasmissio rage of 25 meters. The simulated traffic is Costat Bit Rate (CBR). The simulatio topology is shows i Figure 4. The simulatio settigs ad parameters are summarized i Table-1. Figure 2: Simulatio Topology B. Performace Metrics The proposed Distributed Load Balacig for Multiuser Multi-class Traffic (DLBMM) is compared with the techique [6]. The performace is evaluated maily, accordig to the followig metrics. Packet Delivery Ratio: t is the ratio betwee the umber of packets received ad the umber of packets set. Packet Drop: t refers the average umber of packets dropped durig the trasmissio Badwidth Utilizatio: t is the amout of badwidth utilized at the cell Delay: t is the amout of time take by the odes to trasmit the data packets. C. Results Here the load of cell-1 is balaced amog the cells 3 ad 4. The trasmissio rate is varied as 1, 1.5, 2, 2.5 ad 3Mb ad the above performace metrics are evaluated at cell-1 for dowlik CBR traffic. Figure 3 to 5 show the graphical represetatio of the results for Badwidth utilizatio, delay, packet delivery ratio ad packet drop. Rate Vs Badwidth Utilizatio (CBR) Table-1 Simulatio Settigs Mb/s 2 15 1 5 Fig 3: Rate Vs Received Badwidth Rate Vs Delay(CBR) Delay(Sec) 25 2 15 1 5 Fig 4: Rate VS Delay 25
D e la iv e ry R a t io Pk ts.6.4.2 1 8 6 4 2 Rate Vs DeliveryRatio(CBR) Rate(MB) Fig 5: Rate Vs Delivery Ratio Rate Vs PacketLost(CBR) Fig 6: Rate Vs Packet Lost Because of the distributed load balacig ad optimizatio i, the overloaded traffic is evely distributed, there by reducig the packet drop ad improvig the badwidth utilizatio. Hece from the figures, we ca see that outperforms i terms of badwidth utilizatio by 42%, delay by 33%, packet delivery ratio by 39% ad packet drop by 21%. CONCLUSON this paper, we have proposed a Distributed Load Balacig for Multi-user Multi-class Traffic i MMO-LTE Networks. A robust load balacig framework is implemeted that efficietly hadles the overloaded traffic ad also keep the throughput as high as possible. CBR is used to detect the overloaded cell that detects the load as soo as the service degradatio happes. order to efficietly hadle the cogested cell detected by CBR, Heaviest- First-Load-Balacig algorithm is implemeted to avoid ay ay kid of cogestio i the etwork. REFERENCES [1] Madhava Vajapeyam, Aleksadar Damjaovic, Jua Motojo, Tigfag Ji, Yogbi Wei ad Durga Malladi, Dowlik FTP Performace of Heterogeeous Networks for LTE-Advaced, Commuicatios Workshops (CC), 211 EEE teratioal Coferece o EEE 211. [2] Rooh Keedy ad Awoke Megistie, Load Balacig i Heterogeeous LTE-A Networks, (212). [3] Zhihag Li, Hao Wag, Zhiwe Pa, Na Liu ad Xiaohu You, Dyamic Load Balacig i 3GPP LTE Multi-Cell Fractioal Frequecy Reuse Networks, Vehicular Techology Coferece (VTC Fall), 212 EEE 212. [4] Weyu L, Xiaoyu DUAN, Shucog JA, Li ZHANG, Yu LU ad Jiaru LN, A Dyamic Hysteresis-adjustig Algorithm i LTE Self-Orgaizatio Networks, Vehicular Techology Coferece (VTC Sprig), 212 EEE 75th EEE, 212. [5] MMO Trasmissio Schemes for LTE ad HSPA Networks, 29. [6] Hao Wag, Liaghui Dig, Pig Wu, Zhiwe Pa, Na Liu ad Xiaohu You, Dyamic Load Balacig ad Throughput Optimizatio i 3GPP LTE Networks, Proceedigs of the 6th teratioal Wireless Commuicatios ad Mobile Computig Coferece ACM 21. [7] WANG Mi, FENG Chu-ya ad ZHANG Tia-kui, Mi-max load balacig scheme for mixed multicast ad uicast services i LTE etworks, The Joural of Chia Uiversities of Posts ad Telecommuicatios 19.2 (212). [8] L Bo, WANG Xi-yua ad YANG Da-cheg, terdomai traffic balacig with co-chael iterferece maagemet i multi-domai heterogeeous etwork for LTE-A, The Joural of Chia Uiversities of Posts ad Telecommuicatios 211. [9] WANG Hao, LU Na, L Zhihag, WU Pig, PAN Zhiwe1 & YOU Xiaohu, A Uified Algorithm for Mobility Load Balacig i 3GPP LTE Multi-Cell Networks, Sciece Chia formatio Scieces 56.2 (213). [1] Mig Li, Liag Zhao, Xi Li, Xiaoa Li, Yasir Zaki, Adreas Timm-Giel ad Carmelita Görg, vestigatio of Network Virtualizatio ad Load Balacig Techiques i LTE Networks, Vehicular Techology Coferece (VTC Sprig), 212 EEE 75th EEE, 212. [11] Omar Altrad ad Sami Muhaidat, Load Balacig Based o Clusterig Methods for LTE Networks, algorithms 8 (213). [12] Network Simulator: http:///www.isi.edu/sam/s 26