Performance comparisons between cellular-only and cellularfWLAN integrated systems based on analytical models

本资料包含pdf文件1个，下载需要1积分

Front.Comput.Sci.，2013，7(4)：486-495DOI l0.1007/sl 1704-013-2198-zPerformance comparisons between cellular-only andcelular/WLAN integrated systems based on analytical modelsGuozhi SONG ，Liying YANG ，Jigang WU ，John SCHORMANS1 School of Computer Science and Software，Tianjin Polytechnic University，Tianjin 300387，China2 School of Electronic Engineering and Computer Science，Queen Mary，University of London，London E1 4NS，UK⑥ Higher Education Press and Springer-Verlag Berlin Heidelberg 2013Abstract To solve the trafic load imbalance issue in cellu-lar networks，which is often in the form of hot-spots causedby the diferent user mobility 1evels.one of the good solutionsat present is to consuct heterogeneous integrated wirelessnetworks that combine cellular networks and wireless localarea networks(WLANs)together.In general，the trafic vol-ume is significantly heavier in the hot--spots of cellular net·-works and a higher data transferring rate can be provided byintroducing a WLAN so as to raise the utilization of the chan-nel and achieve a good balance between user satisfaction andthe eficiency of the network.In this paper，we provide a com-prehensive performance comparison of the systems both be-fore and after the integration，based on an existing mathemat-ical model，focusing on both the qualitative and the quantita-tive analysis of changes in the performance of the system tovalidate the efficiency and superiority of the celular/WLANintegrated systems over cellular-only systems。

Keywords celular network，WLAN，heterogeneous wire-less network，performance evaluation，trafic load balance，horizontal handover.veical handover1 IntroductionThe next generation wireless networks with the goal of ofer-ing users the capability of achieving anywhere and anytimenetwork access are widely recognized to have to integrate aReceived May 3l，2012；accepted September 25，2012E-mail：guozhi.song###gmm1.tomvariety of heterogeneous wireless access technologies.Cel-lular and WLAN are two of the mainstream modern wirelessaccess networks.Integrating them has become a promisingsolution，as they have complementary characteristics.It isalso popularly thought to be a good strategy to solve a majorproblem of cellular networks，i.e.，trafic load imbalance be-tween hot-spots and other cels.In practice，special kinds ofareas are usually distributed in cellular networks.These areasare called hotspot-areas(i.e.，hot-spots)，in which both callblocking probability and call dropping probabilities tend tobe higher，due to the heavy trafic density[1.Since WLANsare capable of providing high data bandwidth with limitedcoverage for users with low-mobility，the integration of thesetwo heterogeneous wireless networks can take full advantageof each of them and be beneficial to each other by achievingload balancing。

Since the integrated system enables ubiquitous services，intensive studies have been focused on the interworkingbetween celular network and WLAN.Several efectiveschemes have been proposed，and most of them can be cate-gorized into two groups from the perspective of performanceevaluation：one group uses test beds2-5]and the othersimulation[6-8].For example，a test-bed is implementedin2to construct a real network architecture composed ofseveral physical communication devices for the study of in-terworking of WiMax，3G，and Wi-Fi. In6]a simula-tion based on Network Simulator 2(NS2)is built to evaluate the performance of the celular/WLAN integrated sys-terns and in7]the simulation models constructed using Mat-Lab is adopted to compare the performance diference be-，Guozhi SONG et a1. Performance comparisons between celular-only and celular[WLAN integrated systems 487tween a WCDMA/WLAN interworking system and a TDSCDMA/WLAN interworking system.However，a highlydetailed simulation model implies a large number of parameters and the meticulous implementation of every detail of aspecific network increases the complexity of the model andthe mechanism and performance of the network system .Inthis section，we will describe the analytical model used incellular-only scenario and cellular/WLAN scenario respec-tively and give the qualitative performance comparison resuits。

hence such a model can usualy be applied only to very lim-ited situatiOns.The test-bed meth0ds a1sO have similar prob. 2·1 Cellular Only scenrio1ems and the construction of a good test-bed willincur evenmore cost than simulation.For a heterogeneous wireless in-tegrated system，especially the ones with a hierarchical structure，an analytical model like we used in[9]is preferable。

Although the above schemes are promising，they havesome inherent drawbacks in common，namely，the shortcom-ing of most of them is that only the qualitative performancean alysis has attracted great atention.Moreover，most of theprevious studies always take it for granted that the perfor-mance of the integrated system is certainly better，yet neglectthe performance comparison between an interworking systemand a celular-only network。

Accordingly，the contribution of this paper is to makea comprehensive evaluation，in which both qualitative andquantitative analyses are fully taken into consideration，of thesystem performance based on the analytical model proposedin1 0].By calculating several measures，such as cal block-ing probability and call dropping probability，we are able toevaluate the degree of improvement precisely compared witha cellular-only network.We demonstrate the superiority ofthe integrated system with the metrics obtained。

A preliminary version of the proposed performance evaluation has been published in[11.In this paper，we presentfurther investigations to show the qualitative an d quantitativediferences between cellular-only and cellular/WLAN inte-grated systems and the efect of qiH upon system.The newresults further iustify the superiority and efectiveness of cel-lular/WLAN integration over cellular-only system。

The remainder of the PaDer is organized as folows.Sec-tion 2 discusses the qualitative comparison between cellular-only and celular/WLAN integrated systems.The details ofthe quantitative performance comparisons and numerical re-suits are presented in Section 3.Section 4 discusses the per-formance comparisons of system with effe ct of diferent qiH。

2 Qualitative performance comparisonsAs the first step to make further research，building a suitable system model is the essential foundation for analysingIn wireless communication，the hot-spots，the typical exam-ples of which are a railway station，an o币ce building，an air-port and so on.wil1 be formed when a large number of mobileterminals connect to one access point or base station.In orderto research this issue more efectively，we establish a cellular-only system model as shown in Fig.1.where there are M 4 cells，ofwhich the trafic loadin Cell 1，Cel1 2 and Cell 3 isheavier owing to the hot-spots contained。

HotFig.1 The deployment of cellular-only scenarioAs the users move within the coverage areas，they areable to switch connections among networks. The processof switching connections is called handover.The handoverhas two forli1s：horizontal and vertica1.The horizontal ban-dover refers to the switching process among cells of the samenetwork，while switching among cells of diferent access net-works is called vertical handover.Normally，new connectionblocking probability(call blocking probability)and handoverdropping probability(cal dropping probability)are both rel-atively high for the hotspot areas。

In the cellular system，each cell has the capacity of Ci unitsof bandwidth an d Ti denotes the reservation bandwidths ofcell i.The new ca1l connection arrival processes to cel i fol-lows a Poisson distribution with rate Ai，and accordingly thechan nel holding time of a connection in Cel1 f is an exponen-tially distribution with mean 1 .In addition，at the end ofa cal1 holding time.a connection in cell f may terminate thecal with probability qiT，or switch to the adjacent cell withprobability qiH1-q汀∑f∈A q 。

The probability qiH reflects the holding time of a call and488 Front. Comput.Sci.，2013，7(4)：486-495the times that one handover from one cel to another. To putit another way，it may afect the cal blocking probability andcall dropping probability.The efect of the probability qiHwil be investigated in Section 4。

2.1.1 Cal admission contro1Cal admission control(CAC)is one of the key technologiesfor wireless networks and plays a signifcant role in provid-ing the desired OoS.Based on this，two admission controlstrategies.namely cut-ofr priority and fractional guard chan-nel(FGC)as used in[12]，are used here.From the userspoint of view，generally，it is more annoying to experiencethe abrupt termination of a connection before a call is finishedthan to be blocked on new connection attempts occasionally。

Therefore，a handover connection is designed to have a higherpriority。

Now，we wil1 describe the two admission contro1 strategiesin detail.Cut-0Ir priority is to reserve a portion of channelsfor handover calls，that is to say，the system only accepts therequests from handover users once the current channel occu-pancy reached a given threshold Ti.Whenever a channel isreleased，it is returned to the common pool of unused chan-nels.On the other hand，fractional guard channel schemefirst proposed by Ramjee et a1.131，has been shown to bemore general than the cutof priority scheme.We define nias the number of customers in the system curently.In FGCscheme，when/'i> ft- new calls are accepted with acertain probability o)i that depends of the current channel oc-cupancy(i.e.，the number of busy channels).Of course，whenHi≤ Ci- ，the new and handover connections can be ac-cepted with the same probability under both admission poli-cies2.1.2 Birth-death processesBirth-death processes，where transitions are allowed only be-tween neighboring process states，are special cases of Markovchain.Here，birth-death processes are used to simplify theanalysis of systems with random arrivals and departures ofcustomers(requests or cals etc.)over time。

In the mode1.the arrival of cals is regarded as one birth-death process and as mentioned above，we can learn that birth rate in state i，and，z f is the death rate in state i.P )denotes the probability of state fti and the sum of probabilitiesfor all states is equalto 1。

∑P(ni)1 (1)Assuming steady state，we obtain0≤ni≤Ci～Ti。

C ～Ti.Pi(ni)(1-o)i)Pi(Ci). (3) -mFci-The handover dropping probability is：Bh。 ：Pi(Ci)In the birth-death process coresponding to cell i，rate is defined as follows：Hi ， n i

Vji is the horizontal handover rate of cel j ofered to cell iand it can be calculated by：vji [Aj(1-Bj)Hj(1～Bh，)gJ2.1.3 Erlangs fixed-point approximationIn this section.the iterative algorithm based of the Erlangsfixed-point approximation[14is used to compute the block-ing/dropping probabilities in the system.The detailed pro-cess of the algorithm，the convergence of which is obtainedafter a few iterations，is illustrated in Fig.2。

As shown in Fig.2.the initia1 values of several parame-ters are given in the first step，namely Bi Bh. 0；after-wards，the new connection blocking probability and the han-dover dropping probability could be calculated from Eq.(5)。

Moreover，the function I·I is defined as∑fIBj(new)-日 ands： 10-8。

2.1.4 Numerical resultsIn order to reflect the perform ance of the celular.only sys。

tern.the numerical results based on the network deploymentshown in Fig.1 are ilustrated in Fig.3(a)and(b).Here ( is the arrival rate of new cals per minute in each celular cel1. As shown in the figures，we can see clearly that the newconnection blocking probability and the handover dropping， 帆， Z PP 嘶、 、， 1 1 - -，L /L P P，It，、L Guozhi SONG et a1. Performance comparisons between cellular-only and cellular/WLAN integrated systems 489Fig.2 Flow chart of the iterative algorithm in cellular-only systemArrival rate (。 )/connetions per minuteArrival rate A( )/connetions per minuteFig.3 Celular-only system (a)new connection blocking probability；(b)handover dropping probabilityprobability of Cell 1，Cell 2，Cel 3 are much higher than thatin Cel 4.The main reason for this is the existence of hot-spots and as a consequence the traffiC load imbalance prob-lem is particularly prominent.As a whole，the overall per-formance of the cellular-only system is SO poor that itS quitedifhcult to provide a satisfying service and therefore the traftic load imbalan ce problem has to be addressed in order tomaintain the OoS leve1。

2.2 Cellular-W LAN scenarioGeneraly，most of the mobile terminals in hot.spots arelow speed and consequently we could bring one or moreWLAN into each cell in ceUular/WLAN system for thesake of addressing the trafic load imbalance problem。

Based on the former model of cellular-only network，webuilt the system model for cellular/WLAN interworkingnetwork using the strategy adopted by Enrique Stevens-Navaro in101. As is shown is Fig.4，two adjacentW LANs are introduced into Cel 1 and Cell 2 separatelyan d there is one WLAN in the coverage area of Ce11 3。

Similarly，each cell has the capacity of C units of band-width and，z denotes the current number of the calls con-nected in the Cel f. The area with only celular accessiS referred to as cellular-only area，while the area coveredby both a cel and a W LAN iS referred as double-coveragearea.In our analysis.statistical equilibrium iS assumed forthe whole network。

DHa1Fig.4 The deployment of cellular-W LAN scenarioMoreover，the new call connection arrival processes alsofollow Poisson distributions，with rates and in cell andWLAN respectively.and the channel holding time of a con-nection in cellular Cell i and WLAN k is an exponential dis-tribution with mean l/ix and 1/.t，respectively.The super-scriptsC”，W”，r”，andh”denote cel1.WLAN，new andhandover tra币C.respectively。

2.2.1 PolicyCompared with the strategies adopted in a cellular networksystem，a new resource sharing scheme is considered while皇苦 2q I Q0- -盘 每 Z 童--号 0q∞ulddoJp Ja苦口蠹墨490 Front. Comput.Sci.，2013，7(4)：486-495the strategy of CAC is still the same.Resource sharing is re-alised based on the agreement among the diferent radio ac-cess network(RAN1 providers with all the mobile terminalsequipped with interfaces to access al these RANs.Its opera-tions could be described as follows：if a handover request isnot accepted in one network due to the lack of any additionalresource available.then the request is transferred to the othernetwork.In particular.only when the user is inside the dual-coverage area can the connection be transferred to the othernetwork in the cellular system。

2.2.2 Birth-death processesHere，the connection process of the call is stil seen as a birth-death process.Similarly，the formula is defined as followswhen arriving at the steady stage in the Cell i。

- l pC(ni)nipCi， 0≤ni≤C -rc，(ni-1)or Pc ，z ， c ， C -rc

Besides，the horizontal handover rate of cellular Cell J of-feredtoCeliisdefined as：u c ；(1-8)p7∑ Cxj(1-B cx∈Aw(1-B )p ∑rz(1-BC)pC；，(18)ye zyzw t， Cical handover rate of WLAN k ofered to over-lay cellular Cell i is defined as：w c (1- w w c∑ (1-B WWC ∑u -B WWC∑qck(1-曰 )p ，(20) 十Similarly，the occupancy of W LAN k evolves with birthrates and w based on the state mk，and death rate 。

The birth rate in the WLAN is as follows：p ∑u披w∑ c北w∑f∈A jeD[ g∈《 ， mk≤ Tw ∑ w∑fl cw∑《 ，， g∈Dmk≤c - . (23)Moreover，the horizontal handover rate of W LAN Z ofered toWLAN k is as follows：w敞w (1-曰Wf JqW撬∑uw (1-BhW)ql'∈A∑u； (1-BhW)ql'∑《f(1-BhW)qtV.(24)And the vertical handover rate of celular Cell J ofered to(14) WLAN k is as foilows：canbefound usingEq u； ；(1-BCj)Rj q；∑ (1- )Rjkq；xeA∑UyT(1- )Rj 口 ∑ (1-B： 5)y∈w z∈。

c。r。 ding y，m。de 。。 m。binh rate in thecel 2.2.3 Erlang's fixed-p。int approximati。nlularCel1 i is as folows： u ∑/c口 ∑ ∑%wc∑ ，jeA he f w≤ C - rc，Hi>C -Tc.(17)As mentioned above，the iterative algorithm，of which the in-tegral notion is the same as the scheme in cellular-only sys-tern，is used to compute the blocking/dropping probabilitieshere.The major distinction between the two algorithms is theexistence of vertical handover for the introduction of WLAN。

CP-P∑ Guozhi SONG et a1. Performance comparisons between cellular-only and cellular/WLAN integrated systems 491Consequently，the detailed implementing process of the algo- 2.2.4 Numerical resultsrithm is illustrated in Fig.5。

Fig·5 Flow chart of the iterative algorithm in cellular-W LAN systemFirstly，the initial values of several parameters are given. namely B 。 0；afterwards，the handoverrates could be calculated by Eqs.(18)，(20)，(24)，and(25)；then，we compute total trafic ofered to cell and WLAN fromEqs.(16)，(17)，(22)，and(23)，respectively.The functionis defined as∑，IBj(new)-B嘘10-。

Arrival rate ( )/connetions per minuteIn this section.the simulations are conducted based on thesystem deployment shown in Fig.4 and then the correspond。

ing qualitative analysis is presented for a comprehensive eva1。

The simulation parameters in the numerical experimentsare 40， 80， 2， I， 0.25，∈5，20]， 5，R驰0.75。

Figs.6(a)and(b)shows the new call blocking probabilityand handover dropping probability of each cell and WLANin the integrated celular/WLAN system separately when thenew connection arrival rate is increased from 5 to 20 con-nections per minute in al celular cells。

As expected，the two probabilities in cellular Cell 1 andCell 3 are the highest for the number of neighboring celsis the largest among the four cells，and thus the possibleresource consumption allocated to handover users may in-CreaSe。

In both figures，curves labeled WLAN are all lower thanthose labeled as cellular.The reason for this is mainly thatthe bandwidth capacity of WLAN is larger and its coveragearea js smaller2.3 Qualitative comparison resultsThe comparison of the average of the new cal blocking prob-ability an d handover dropping probability between celular-only system and the integrated cellular/WLAN system aredescribed in Figs.7(a)and(b)separately.As is shown inthe picture，the tWO probabilities of the integrated system areboth at a low level compared to those of a cellular-only sys。

tern，for any value ofthe tota1 arrival rate in the range ofArrival rate (。)/connetions per minuteFig·6 Celular-WLAN system (a)new connection blocking probability；(b)handover dropping probability占-- 碍 o(I uIddo印 J。 0 II矗王---n 0I4 lI- u0- --日Q r0Z492 Front.Comput.Sci.，2013，7(4)：486-495Arrival rate ( )/connetions per minute Arrival rate ( )/connetions per minuteFig.7 (aComparison of the average of new connection blocking probability；(b)comparison of the average of handover dropping probability45-105.Due to the additional capacity added by introducingthe WLAN in dual-coverage areas.the handover droppingprobability in the integrated system is capable of achieving1ower，of which the minimum is less than l 0-j and the peakvalue is not exceed 1 0～.So.the performance of the inte-grated system has been improved signifcantly thanks to thestrategy of admission control and resource sharing。

3 Quantitative performance comparisonsIn the preceding sections，we have made a qualitative analysisbased on extensive numerical analysis.The conclusion thatthe integrated system is superior to the cellular-only networkmight easily be drawn。

However，a new problem has arisen：how much of an im-provement in performance can be obtainedIn order to an-swer the question，several measure indexes are introduced toreflect specifc improvement more accurately。

Before the definition of measure parameters are presented，the histogram，which is at the basis of the data obtainedabove，represented the diference of the two probabilities，new cell blocking probability(NBP)and handover droppingprobability(HDP)，between cellular-only system and the in-tegrated cellular/WLAN system is shown in Fig.8。

We can see clearly that D--value of new call blocking prob-ability has a larger scale of variation with the parametervarying from 5 to 20.Based on the results shown in the his-togram，we introduce the following indexes：Mean of the D-value of new cal blocking probabilityMNBP：-M NBP 1 / [max er( - e (删， (26)0.2O0.180.160.14翌 0.12霉 0.10凸 0 080.O60 04002000Cell 4 /connections per minuteFig.8 The tendency of the diference valuewhere is the arrival rate Af，， 血n and Amax represents thelower limit and the upper bound separately。

BP( )and BP( )denotes the mean of new call block-ing in celular-only system and cellular/WLAN system sepa-rately and can be calculated from the following formulae：BP( ) 1∑ (27)2 卅 )' ，where ( )is the new call blocking probability of Cel i cor-responding to certain in celular-only system。

Similarly，Bc(A)and ( )denotes the new call blockingprobability corresponding to certain/l of Cel f and WLAN kin cellular/WLAN system。

Mean of the D-value of handover dropping probabilityM HDP：-MHDp 1 lM )- DP(删. (29)Actualy，the definition of MHDP is the same as that ofMNBP and the meaning of the parameters is similar to a large鲁qB 2q II-dd2 J。 0 II 占鲁 对 0Jq 口显u0- --g 每oZGuozhi SONG et a1. Performance comparisons between cellular-only and cellular/WLAN integrated systems 493extent. DP( )and DP( )denotes the mean of han-dover dropping probability in celular-only system and cel。

lular/WLAN system separately and can be calculated fromthe following formulae：D ( ， NDP( 既( (30)Nc Nwcrease in the percentage of the new connection blocking prob-ability and han dover dropping probability，the enhancementof which are more than 50% and 4O%.Therefore the eel-Iular/WLAN integration can be used as a good solution toalleviate the tramc imbalance problem in a celular networkwith many hot-spot areas。

㈤ ㈤)， e4 Per form～anc ecoHmp sons0 yste- hwhere Bhi( )is the new cal blocking probability of Cell icorresponding to certain in cellular-only system。

Similarly，nmc( )and 8 ( denotes the handover droppingprobability corresponding to certain of Cel i and WLAN kin cellular/WLAN integrated system。

The percentage of reduction in new call blocking probabil-ity PNBP iS defined as：P /lmi )×10%。

The percentage of reduction in handover dropping probability PHDP is defined as：PH。r ( 篓 ! ： ]× 。%c3By incorporating the data into the form ulae above，we cancalculate the four factors and the results are given in Table 1。

Table 1 Qufimtive comparison resultsWeareabletoobservefrom thetablethattheperforman ceboost for the whole network system with a significant in-Arrival rate ( n)/connetions per minuteAnother group of numerical experiments is conducted in or-der to make a further comparative analysis of the system withthe impact of diferent qiH values。

As the probability of a call switching to the adjacent cell，qiH will certainly afect the value of the call blocking prob-ability and cal dropping probability.The experiments areconducted with the sanqe network deployment and configura-tion as in previous sections and the results are shown in thefigures folowed。

The efect on the perform ance of cellular-only and celular--WLAN integrated systems under diferent values of qiH canbe observed in Fig.9 and Fig.10，respectively。

Wle can see clearly that the relationship is positive core-lation between qiH and the two probabilities.That is to say，both the new call blocking probability and the handover drop-ping probability present a significant uprising fiend with theincrease of the qiH。

The main reason is that qiH is the critical parameter afect-ing the average resident time and the total holding time of acall in each cel1.With regard to any call，hence，the largerqiH is，the bigger the total holding time is，and the averageresident time in one cel wil be lower relatively。

As a result.the total number of handover cals will beArrival rate (。)/connetions per minuteFig.9 Cellular.only system(a)new connection blocking probability with diferent qiH；(b)handover dropping probability with diferet qiHIIl 8q∞暑dd0电 矗 0p蠹王兽-号DoJ 普 曩u0- -8 亭QZ494 Front.Comput.Sci.，2013，7(4)：486-495Arrival rate (。)/connetions per minute Arrival rate (。)/connetions per minuteFig.10 Celular-WLAN system (a)new connection blocking probability with diferent qiH；(b)handover dropping probability with diferent qiHgreatly increased.As a consequence of that，a significant rise proach for seamless mobility in next generation wireless networks。

Of the new call blocking probability and the handover drop- In：Proceedings of the 33rd IEEE Conference on Local Computerpingprobabi1itywiu betriggered·5.Nseltehw orAk.sM20。0bi8l，。 呻 锄 。。 d i te 。rkj g 。hitect i5 ConclusionsIn this paper，a comprehensive evaluation of the performanceof a wireless cellular system with unevenly distributed traf-tic is presented and al in-depth comparison of the cellularnetwork both before and after the WLAN integration is alsoconducted.A11 the numerical simulations are based on anexisting mathematical mode1.By calculation of several mea-sure factors，the precise degree of improvement of systemperformance can be obtained.Finaly，the eficiency of theintegrated system is proved via qualitative and quantitativeanalysis，and therefore the credible basis can be provided forflture research。

Acknowledgements This work was supported by the Tianjin Higher Ed-ucation Fund for Science and Technology Development(201 10808)，andthe National Natural Science Foundation of China(NSFC)(Grant No。

References24Verta 0、Mastroianni C，Talia D.Evaluation of methods for determin。

ing the mobile traffic distribution in cellular radio networks.IEEETransactions on Vehicular Technology，2001，50(6)：1629-1635Gondi V K.Agoulmine N Secure interworking ＆ roam ing ofWiMAX with 3G and Wi-Fi.In：Proceedings of IFIP/IEEE Interna。

tional Symposium on Integrated Network Management.2009，133-136Park S，Lee D，Kim S，lhm J，Oh S.A performance evaluation ofhandoffmethod between WLAN and cdma2000 lx Ev-Do system.In：Proceedings of the International Conference on Networking，Systems，Mobile Communications and Learning Technologies.2006.134Kim Y S，Kwon D H，Suh Y J.A client-based vertical handof ap-6791l14802.1 l WLAN/CDMA2000 networks.In：Proceedings of the 2ndAnnual Conference on Communication Networks and Services Re-search.2004、75-79Li-der Chou W L，Lin C，Lin Y Huang C M.Seam less handover inW LAN and cellular networks through intelligent agents.Joumal ofInformation Science and Engineering，2007，23(4)：1087-l 101Jin S，Song G，Cuthbert L.Perform ance comparison of verticalhandover between WCDMA/WLAN and TD-SCDMA/WEAN inter-working systems.In：Proceedings of AISPC08- the 2nd AnnualIEEE Student Paper Conference.Aalborg：IEEE Press，2008，1-5Xia Shen L Call admission control with load-balancing capabilityin integrated celular/WLAN networks.The Institute of Electronics，Inform ation and Communication Engineers Transactions on Commu。

nications.2010.93f51：1l9O-1204SongG，W uJ，Schormans J，YangL，CuthbertL.Aperform ance studyof hierarchical heterogeneous wireless integrated networks.Journalof Applied Mathematics&Information Sciences，2012，6(2S)：497-504Kirsal，Y Gemikonakli E，Ever E，Mapp G，Gemikonaklj O An ana-lytical approach for perform ance analysis of handofs in the next gen-eration integrated celular networks and WLANs.In：Proceedings ofthe 19th Intemational COlference on Computer Communications andNetworks.London：IEEE Press.20l0.1-6Yang L.Song G.Wu J.A performance evaluation of celular/WLANintegrated networks.In：Proceedings of the 4th International Sympo-sium on Parallel Architectures.Algorithms and Program ming.Tian-jin：IEEE CPS Press，201 1，131-135Enrique S N，Wong Vincent W S.Resource sharing in an integratedwireless celular/WLAN system.In：Proceedings of Canadian Con-ference on Electrical and Computer Engineering，2007，631-634Ramjee R，Towsley D，Nagarajan R.On optimal call admission con-trol in cellular networks.In：Proceedings of IEEE International Con。

ference on Computer Communications.1996.43-5OSong G，Wu J，Schormans J，Cuthbert L.Erlangs fixed-point ap-proximation for performance an alysis of HetNets.Journal of AppliedMathematics.2012.2012：1-13告I州 0口 口-q00 I 0勺II 雹童- 号 0J口詈 写00- --B。≥0Z