An Efficient Handoff Scheme Using a Minimum Residual Time First Scheme

Transcription

1 An Efficient Handoff Sceme Using a Minimum Residual Time First Sceme Bilal Owaidat Rola Kassem and Hamza Issa Abstract Wen a mobile station (MS) wit an ongoing call is about to leave a cell te base station (BS) of tat cell ands off te call to te BS of te new cell. Since a BS typically serves a large number of MSs it is likely tat many andoff requests arrive at a sort period of time at te BS were tey are queued for andling according to some queueing discipline. Current andoff scemes use a FCFS discipline wose problem is tat it does not pay attention to te remaining time in te cell of te leaving MS. In tis paper we propose a sceme to remedy te above problem. Te Sortest Job First-like Handoff Sceme (SJFLHS) still queues te MSs requiring andoff but serves always te MS wit te least remaining time in te cell. To tis end te BS keeps monitoring te remaining times of MSs requiring andoff. Te cost of tis overead is insignificant compared to te resulting performance improvement wic is sown ere troug extensive simulations. Keywords cellular network andoff priority sortest job first queue I. INTRODUCTION In te fourt generation wireless networks multiple base stations cooperate to serve mobile stations requests. Wen a MS wit an ongoing call arrives at a BS it will connect only if tere exist a free cannel oterwise its call will be blocked. Once connected tis BS is called te serving base station (SBS) of te MS. A BS as a circular coverage area wit a radius ranging from a few undred meters to a few kilometers in radius. Wen a MS reaces te boundaries of te SBS coverage area it sould disconnect from tis SBS and connect to a new BS wic is called te target base station (TBS). Terefore te corresponding cannel in te SBS is released and on te oter and a cannel is reserved in te TBS. Tis process is known as andoff. Te cannel frequencies could be reused in cells tat are separated in distance so tat tere is no mutual interference between tem. In a practical environment tere is some andoff time delay before tat a MS connects to a TBS. Tis delay is composed of andoff preparation and andoff execution. Te andoff time delay is constrained by a timeout period tat wen exceeded te andoff fails and te call is forced-terminated. Tere are two types of andoff scemes: ard andoffs and soft andoffs [1][2]. In ard andoff te cannel in te SBS is released and ten a cannel in te TBS is reserved. Tus te connection to te SBS is B Owaidat PHD student at te Beirut Arab University Computer Engineering Department Lebanon Debbiye ( b.owaydat@bau.edu.lb) R Kassem Assistant Professor in Beirut Arab University Computer Engineering Department Lebanon Debbiye ( r.kassem@bau.edu.lb) H Issa Associate Professor in Beirut Arab University Computer Engineering Department Lebanon Debbiye ( .issa@bau.edu.lb) ended before or as te connection to te TBS is being created. Determining an accurate trigger time for ard andoffs is a difficult task. Soft andoffs solve tis problem by connecting te MS to te TBS before breaking wit te SBS. Te MS olds te radio resources of te SBS and TBS simultaneously during a soft andoff period. A call is eiter complete or incomplete. A complete call is a one tat is served witout any andoff failure. An incomplete call is eiter a blocked call (a new call tat is blocked) or a forced-terminated call (an ongoing call tat failed to andoff). Note tat a call could succeed multiple andoffs before it is forced terminated. Tere are two major metrics tat define te performance of cellular systems: te probability tat an ongoing call is dropped (because of its forced-termination) and te probability tat a new call is blocked (because of unavailability of a free cannel). Wen a MS is moving wit ig velocity te number of andoff attempts dramatically increases. From a user s point of view it is more annoying to drop an ongoing call tan to block a new call. Terefore andoff calls are given a iger priority to get a cannel over new calls. Tis prioritization rule ave a significant impact on tose two major metrics; it decreases te andoff failure probability at te expense of an increase in te call blocking probability. Te most popular strategies for prioritizing andoff calls are te guarded cannel strategy and te andoff queueing strategy [3]. In te guarded cannel strategy a fixed number of cannels is reserved exclusively for andoff calls. New calls are not allowed to use tese guarded cannels for service owever a andoff call can use any cannel [4]. If only te dropping probability of andoff calls is considered te guarded cannels algoritm gives good results owever it will increase te blocking probability of new calls dramatically. Te guarded cannel strategy as some drawbacks wen all te cannels (reserved and non reserved) are busy ten te andoff will fail and te call is forced-terminated. On te oter and in te andoff queueing strategy if tere is no available cannel for te andoff request tis request is queued until a cannel gets free. Queueing is possible due to te overlap region between neigbor cells were te MS can reac bot te SBS and te TBS. Te MSs remaining time in te overlapping area specifies te maximum allowed queueing time. In [5] te Predictive Received Signal Strengt (PRSS) studied weter to start a andoff by comparing some quantitative decision values to select a target network. However tis algoritm could not be applied because te MS sould know ow strong is te PRSS of its neigbors in order to decide an early andoff. Two andoff algoritms were developed in [6] based on te PRSS and te current RSS but tey didn t avoid te pingpong effect. Oter vertical andoff decision algoritms were ISSN: All Rigts Reserved c 2015 IJARCET 3559

2 proposed based on fuzzy logic to give te best solution for andoff decision [7][8]. Te problem wit tese algoritms is tat tey all need to establis some proper rules tat require a large memory to store rules in databases. Te following studies [9][10][11] improved andoff decisions by analyzing te signal power received by stations but witout taking te MSs velocities into considerations. Oter studies considered te mobility of MSs to decrease andoff failure probabilities [12][13]. Many oter feasible solutions were developed to enance andoffs by monitoring te MSs positions [14]. Using te global positioning system (GPS) is a way to get te locations of MSs; owever wen used in extreme weater conditions it could suffer from serious interfering problems [15]. A mecanism to determine te location of devices witout using te GPS was developed in [16]. In [17] te andoff mecanism ASAP considered bot te SNR value and te mobility of MSs using geometric analysis. A fast andoff sceme for voip was considered in [18] it is based on selective scanning and cacing to predict te next cell. None of te previously proposed scemes is efficiently able to serve andoff requests wen MSs are moving wit ig velocity. Te present sceme treats eac MS individually it gives a iger priority for cannel reservation to te MS wit te minimum remaining time in te cell τ. It sorts te MSs andoff requests according to teir remaining times and andoffs te one tat as te lowest τ first. In oter words tis sceme implements a sortest job first (SJF)-like queue. Te rest of tis paper is organized as follows: in section 2 we provide some background about andoff strategies ten te system model is introduced in section 3 te simulation model is introduced in section 4 in section 5 te simulation results are given and finally conclusions are presented in Section 6. II. BACKGROUND A cell is represented by a circular region wit center coordinates (x y) and radius R. Te center of te first cell is located at te point (0 0). Eac cell intersects wit its neigbors wit a small area called te overlap region tat is te area were a MS can receive signal from bot cells. Wen a MS is moving across te boundaries of a serving cell to a neigboring cell it will pass troug te overlap region were it can start te andoff procedure to transfer its call from te SBS to te TBS. We assume tat te new calls and andoff calls arrive according to a Poisson distribution wit parameters λ n and λ respectively. Eac station as C cannels tat can serve C calls in parallel. Te time tat te MS spends in te system is referred to as te call olding time T and is assumed to be exponentially distributed wit parameter µ. Terefore µ is te ongoing calls termination rate and 1 µ is te average time a call spends in te system. We sould also note tat wen an ongoing call is ended or anded off to a TBS it is discarded from te records of te SBS. A. Guarded cannels strategy In te guarded cannel strategy a MS requests a andoff once it enters te coverage area of te new cell tat is wen Fig. 1. State transition diagram for te guarded cannel strategy. Λ n Λ Λ n Λ Λ n Λ Λ n Λ Λ Λ C-C H C Μ 2Μ 3Μ ΜCC H ΜCC H 1 CΜ it enters te overlap region were it can reac bot te SBS and TBS. Te number of cannels reserved for andoffs is C H. Let E n be te state tat tere are n ongoing calls in a cell and P n be te steady state probability for a cell to be in state E n. Tese probabilities are determined by using te Markovian birt-deat process sown in te state diagram in Figure 1 and are found to be: (λ n+λ ) n n!µ P P n = n 0 n < C C H. (λ n+λ ) C C H λ n ( C C H ) n!µ P n 0 C C H n C. Were P 0 is found from te equation P 0 + P P n = 1 ( C CH (λ n + λ ) n P 0 = n!µ n n=0 ) C (λ n + λ ) C C H λ n (C C 1 H) + n!µ n n=c C H +1 Let P f represent te andoff failure probability and P B represent te call blocking probability because no cannels are available. A call is blocked if all te non-guarded cannels were busy wen tis call was initiated terefore: P B = C n=c C H P n A andoff fails wen all te cannels (guarded and nonguarded) are busy terefore: P f = P C B. Handoff queuing strategy In te andoff queueing strategy tere are no guarded cannels. Here if te number of busy cannels is less tan C ten te TBS accepts every incoming call. But wen te C cannels are all busy ten te TBS will only accept andoff calls. Tese andoff calls wait for teir turn in a queue until a cannel gets free and ten tey are treated equally on a FCFS basis. Let te time tat a mobile station spends in te overlap area to be exponentially distributed wit rate η tus η is te departure rate for calls tat leave te overlap region. Te mobile station s association time wit a cell is exponentially distributed wit rate v also v is te departure rate for calls ISSN: All Rigts Reserved c 2015 IJARCET 3560

3 Λ n Λ Λ n Λ Λ n Λ Λ Λ Λ X 1 Y 1 Θ 0 1 C C+1 C+2 X 2 Y 2 SBS {XY} TBS ΜΝ 2ΜΝ CΜΝ CΜΝ +(Μ+Η) CΜΝ +2(Μ+Η) CΜΝ +3(Μ+Η) Fig. 2. State transition diagram for te queuing strategy. Fig. 3. A MS leaving its SBS and moving toward a TBS. tat leave te cell. Figure 2 sows te state diagram for te queuing strategy. Here te steady state probability is: (λ P n +λ ) n 0 n!(µ+v) n < C. P n = n (λ P n +λ ) c λ n c 0 [c(µ+v)+j(µ+η)] n C. Were + P 0 = n=c ( c 1 n=0 (c 1)!(µ+v) c 1 n c j=0 (λ n + λ ) n n!(µ + v) n (λ n + λ ) c λ n c (c 1)!(µ + v) c 1 n c j=0 [c(µ + v) + j(µ + η)] Te call blocking probability is P B = n=c and according to [19] te andoff failure probability is η (j + 1) P c+j P f = c (µ + v) + (j + 1) (µ + η) j=0 P n ) 1 III. SYSTEM MODEL In te present study we use a model tat combines bot strategies (guarded cannels and andoff queuing). In tis model a MS starts a new call anywere in a cell and ten it travels towards a neigbor cell. We assume tat te velocity (speed and direction) of eac MS remains constant wile in a cell. But two different MSs may travel wit two different velocities. Wen a MS leaves its SBS and connects to a TBS it will travel wit a new velocity towards te neigbor cell of te TBS. Te velocity magnitude of te MS canges from cell to cell and is uniformly distributed on te interval [V min V max ]. Te time spent in a cell and te time spent in te overlap region is not te same for all MSs. We specify a tresold time τ ; a MS can reserve a cannel wen its τ becomes less tan tis tresold (τ < τ ). Note tat tis cannel will start serving te call wen te MS enters te overlap region. If no cannel is available at te TBS te MS keeps trying to succeed te andoff as long as it is still in te overlap region. If te MS leaves te overlap region before succeeding to andoff its call ten te call is forced-terminated. Terefore τ and τ are two major factors tat affect te call forced termination probability. If two MSs (MS1 and MS2) ave τ MS1 < τ MS2 < τ ten MS1 is more likely to leave te cell before MS2. In previous studies te MSs are treated equally and are served in a FCFS discipline. Tis discipline could serve MS2 before MS1 if it requested te andoff first and tis could lead to te forced-termination of te call of MS1 if tere are no more free cannels at te TBS. To solve tis problem te MSs sould not be treated equally. In SJFLHS we make sure tat te MSs are served in ascending order of teir τ i.e. wen two MSs request a andoff te one wit te lowest τ gets te cannel first. Tis is similar to te Sortest Job First (SJF) queueing discipline [20]. Let N be te number of MSs tat starts teir calls in a cell during a time T. Te position of every MS is updated every t << T. If MS1 is a mobile station located at coordinates (X 1 Y 1 ) te new position of MS1 depends on its velocity vector V MS1 wic is constant bot in magnitude and angle θ (taken wit respect to X-axis) as sown in Figure 3. Note tat te magnitude V MS for a MS is uniformly distributed over te interval [V min V max ]. Te value of θ is found to be: θ = tan 1 ( XT BS X 1 Y T BS Y 1 Were X T BS and Y T BS are te coordinates of te center of te TBS. Te new coordinates (X 2 Y 2 ) of MS1 are : ) X 2 = X 1 + t V cos θ Y 2 = Y 1 + t V sin θ MS1 is moving on te straigt line Y = Y 1 + (X X 1 ) tan θ it will leave te SBS s coverage area wen it reaces te point of intersection (X Y ) of tis straigt line wit te coverage area of te SBS wic is a circle of radius R. Te remaining time τ MS1 for MS1 in te SBS is : (Y Y MS1 ) 2 + (X X MS1 ) 2 τ MS1 = V MS1 Wen a MS ave τ < τ it will ask its SBS to andoff its call. Te SBS sorts te MSs every t seconds in ascending order of teir τ ten it request from te TBS a cannel for te ISSN: All Rigts Reserved c 2015 IJARCET 3561

4 TABLE I. VALUES USED IN SIMULATION MS2 MS1 SBS MS3 TBS Parameter Value Description R 3000 m Radius of a Cell S 50 Number of Cells N 100 Number of MSs in a cell C 32 Number of cannels in a cell GC 2 Number of guarded cannels in a cell V min 1 m/s Minimum speed of a MS V max 30 m/s Maximum speed of a MS λ 1/10 Calls/sec Call Arrival rate µ 1/180 Calls/sec Call Service rate τ 10 sec Tresold time Fig. 4. Handoff requests wit different priorities. MS tat as te lowest τ. For example referring to Figure 4 if τ 1 < τ 2 < τ 3 < τ ten MS3 is moving slowly and sould be anded off last. On te oter and MS1 is moving faster tan MS2 and MS3 and sould reserve a cannel first. IV. SIMULATION Wen a new call arrives at an SBS it will ceck weter tere is a free (non-guarded and not reserved by a MS) cannel to serve tis call if no cannel is free ten te call is blocked. Wen a MS wit an ongoing call ave τ < τ it will attempt to reserve a cannel (guarded or not guarded) at te TBS. If te TBS as an available cannel it will reserve it for tis MS and will allocate it as soon as tis MS enters te overlap region. However if te TBS as no free cannels tis MS sould wait for its turn in a queue tat is periodically sorted in ascending order of τ. If a MS finises its call wile waiting in te queue ten te corresponding reserved cannel is released. If a andoff call leaves te SBS wile waiting in te queue ten te call is forced terminated and te corresponding reserved TBS cannel is also released. Te following algoritm sows ow SJFLHS treats new calls and andoffs. New Call If (tere is a free non guarded cannel) Allocate te cannel Else Block call (New call blocking) Reservation If (tere is a free cannel (guarded or non guarded) at te TBS) Reserve tis cannel Else Ignore te request Allocate te reserved cannel to te MS Else If tere is an available cannel Allocate te available cannel to te MS Else Block te Call (Forced Termination) We ave simulated te above model using a C++ program. Let N be te number of calls tat starts in a cell during a time T. First we specify te number of MSs in eac cell. Ten we specify te call arrival and ending times and also te initial position and te velocity vector for eac MS. Next we update te position and τ for eac MS every t seconds. Finally te SBS sorts te calls in ascending order of teir τ and ten it will andoff te MS wit smallest τ first. Tat is te most critical calls are always anded off first. Tis sould reduce te andoff failure probability at te expense of an increase in call blocking probability. Te parameters used in te simulation are sown in Table I unless oterwise specified. We sow te effect of various parameters on te call blocking probability P B and andoff failure probability P f A. Impact of new calls arrival rate on P B and P f Figure 5 (a) and (b) sow P B and P f for FCFC and SJFLHS against new calls arrival rate λ. We notice tat te increase in λ as lead to more blocked calls. Also P B in SJFLHS is sligtly iger tan te FCFS strategy and tis is because te former strategy succeeds te andoffs better tan te latter terefore tere are less cannels for new calls. On te oter and P f as decreased wen λ as increased and tat is because an increase in λ will result in more blocked calls and terefore andoffs will ave more cannels for teir service. Also it is clear tat P f in SJFLHS is significantly lower tan tat of te oter strategy. Cancellation If(A call associated wit a reserved cannel as finised) Free tis reserved cannel Handoff If (MS as reserved a cannel) B. Impact of calls service rate on on P B and P f Figure 6 (a) and (b) sows P B and P f against te service rate µ. Bot probabilities ave decreased wen µ as increased. Te ongoing calls will get served muc better wen te service rate is ig terefore te new calls and te andoffs will ave more cannels to compete for. Tis of course will lead to a decrease in blocking probabilities. Again SJFLHS gives lower P f tan FCFS tecnique. ISSN: All Rigts Reserved c 2015 IJARCET 3562

5 PBFCFS PBSJFLHS New Calls Arrival RateCallsSecond a PBFCFS PBSJFLHS New Calls Arrival RateCallsSecond a PfFCFS PfSJFLHS New Calls Arrival RateCallsSecond b PfFCFS PfSJFLHS New Calls Arrival RateCallsSecond b Fig. 5. Rate. Call Blocking And Handoff Failure Probabilities vs. Calls Arrival Fig. 6. Rate. Call Blocking And Handoff Failure Probabilities vs. Calls Service C. Impact of guarded cannels on P B and P f In Figure 7 (a) and (b) we sow te impact of guarded cannels on bot tecniques. Wen te number of guarded cannels as increased P B as increased wile P f as decreased. P f in SJFLHS is lower tan P f in te FCFS strategy of course at te expense of an increase in P B. D. Impact of time tresold on P B and P f Figure 8 (a) and (b) sow P B and P f for different tresold times. Wen tresold τ as increased P f as decreased because te MS will ave more time to reserve a cannel wic means tey ave more advantage to succeed teir andoffs. On te oter and P B as increased because less remaining cannels are available for new calls. Also in SJFLHS P f is lower tan te FCFS strategy wile P B is iger. E. Impact of MS velocity on P B and P f In tis part we assume tat te MS s maximum velocity V max wic is sown in Table I does not cange and te MS s minimum velocity V min will increase from 1 meter/second up to V max. Wen a MS is moving wit a ig velocity it will leave te coverage area of a cell muc faster tan if it is moving wit a low velocity. Terefore te MS will request more andoffs for its call tis will increase te total number of andoffs in te system as sown in Figure 9. Te increase in andoffs will also result in more andoff failures. In Figure 10 we sow te impact of velocities on P B and P f. Wen te velocity of MSs increased P f as also increased wile P B as decreased because more andoff failures mean more cannels for new calls. Also in SJFLHS P f is lower tan te FCFS strategy of course at te expense of a small increase in P B. F. SJFLHS vs ASAP In Figure 11 we compare our sceme to a recently developed sceme ASAP. It is clear tat SJFLHS as significantly decreased te andoff failure probability P f in comparison to ASAP owever it as also increased significantly te call blocking probability P B. V. CONCLUSIONS In tis paper we ave presented a novel cellular andoff sceme particularly suitable for MSs moving wit ig velocity. Te idea of te sceme is not to andle MSs queued for andoff on a FCFS basis but rater to select for service te MS wit remaining time in te cell less tan all oters in te queue a sceme akin to te sortest job first (SJF) discipline. Tis sceme decreases te likeliood tat a MS leaves te cell before its andoff request is andled a common problem wit FCFS scemes. Te simulation program used to validate te proposed sceme is written in te C language wic requires more ISSN: All Rigts Reserved c 2015 IJARCET 3563

6 PBFCFS PBSJFLHS Guarded Cannels a RateHandoffsCustomer FCFS SJFLHS Minimum VelocityMetersSecond Fig. 9. Handoff Rate vs. Velocity PfFCFS PfSJFLHS Guarded Cannels b Fig. 7. Call Blocking And Handoff Failure Probabilities vs. Calls Guarded Cannels PBFCFS PBSJFLHS TresoldSeconds a PBFCFS PBSJFLHS VelocityMetersSecond a PfFCFS PfSJFLHS VelocityMetersSecond b Fig. 10. Call Blocking And Handoff Failure Probabilities vs. Velocity. Fig. 8. PfFCFS PfSJFLHS TresoldSeconds b Call Blocking And Handoff Failure Probabilities vs. Tresold. effort but yet provides more control tan very ig level simulation languages and platforms. Te simulation results sow tat te proposed sceme decreases noticeably te andoff failure probability compared to FCFS based scemes. Te overead due to always placing te request wit te least remaining time at te ead of queue is O(n) a practically desired property. Te simulation also sows tat te decrease in te probability failure rate comes at te expense of an increase in te call blocking probability clearly because a cannel taken for a andoff call is basically taken from a new call. However as is well known in cellular tecnology to block a new call is more preferable tan to terminate an ongoing one. ISSN: All Rigts Reserved c 2015 IJARCET 3564

7 Fig. 11. ASAP. PBSJFLHS PBASAP New Calls Arrival RateCallsMinute a PfSJFLHS PfASAP New Calls Arrival RateCallsMinute b A comparison of call blocking probabilities bewteen SJFLHS and [10] A. Roy J. Sin and N. Saxena Multi-objective andover in lte macro/femto-cell networks IEEE Communications and Networks vol. 14 no. 5 Oct [11] Y. Yu and D. Gu Te cost efficient location management in te lte picocell/macrocell network IEEE Commun Lett vol. 17 no. 5 May [12] T. Guo A. Quddus N. Wang and R. Tafazolli Local mobility management for networked femtocells based on x2 traffic forwarding IEEE Trans Ve Tecnol vol. 62 no. 1 Aug [13] J. Astorga M. Aguado N. Toledo and M. Higuero A ig performance link layer mobility management strategy for professional private broadband networks Journal of Network and Computer Applications vol. 36 no. 4 Jul [14] A. Xafa and O. Tonguz Dynamic priority queueing of andover calls in wireless networks: an analytical framework IEEE J Sel Area Commun 2004 vol. 22 no. 5 pp Jun [15] S. Hong M. Lee H. Cu S. Kwon and J. Speyer Experimental study on te estimation of lever arm in gps/ins IEEE Trans Ve Tecnol 2006 vol. 55 no. 2 pp Mar [16] Y. Cen and S. Tsao A low-latency scanning wit association mecanism for real-time communication in mobile wimax IEEE Trans Wirel Commun 2012 vol. 11 no. 10 Oct [17] C. Jenui Y. Zuxiu and W. Lei An apropos signal report and adaptive period (asap) sceme for fast andover in te fourt-generation wireless networks Journal of Network and Computer Applications vol. 45 no. 1 Oct [18] J. Xiao and F. Liu A pre-scanning fast andoff sceme for voip in wlans International Journal of Future Generation Communication and Networking vol. 8 no. 2 Feb [19] Y. Lin. and A. Pang Comparing soft and ard andoffs IEEE Trans Ve Tecnol vol. 49 no. 3 pp May [20] K. Aida Effect of job size caracteristics on job sceduling performance. Springer Berlin Heidelberg Jan REFERENCES [1] B. Quang R. Prasad and I. Niemegeers A survey on andoffs lessons for 60 gz based wireless systems IEEE Communications Surveys and Tutorials vol. 14 no. 1 pp [2] S. Amit U. Abay and D. Moan An overview of andoff tecniques in cellular networks International Journal of Advanced Engineering Tecnology vol. 3 no. 1 Mar [3] N. Tripati J. Reed and H. VanLandinoam Handoff in cellular systems IEEE Personal Communications vol. 5 no. 6 pp Dec [4] D. Hong and S. Rappaport Traffic model and performance analysis for cellular mobile radio telepone systems wit prioritized and nonprioritized andoff procedures IEEE Trans. on Veicular Tecnology vol. 35 no. 3 pp [5] L. Xia L.Jiang H. Cen and H. Liao An intelligent vertical andoff algoritm ineterogeneous wireless networks. IEEE International conference on Neural Networks and Signal Processing Jun [6] B. Cang and J. Cen Cross-layer-based adaptive vertical andoff wit predictive rss in eterogeneous wireless networks IEEE Transactions on Veicular Tecnology vol. 57 no. 6 pp [7] C. Ceken and H. Arslan An adaptive fuzzy logic based vertical andoff decision algoritm for wireless eterogeneous networks. IEEE 10t Annual Wireless and Microwave Tecnology Conference Apr [8] X. Haibo T. Hui and Z. Ping A novel terminal-controlled andover sceme in eterogeneous wireless networks Computers and Electrical Engineering vol. 36 no. 2 pp [9] A. Rat and S. Panwar Fast andover in cellular networks wit femtocells. IEEE International conference on Communications (ICCC) Jun ISSN: All Rigts Reserved c 2015 IJARCET 3565