2005 Springer Science + Business Media, Inc. Manufactured in The Netherlands.
Repacking on Demand for Hierarchical Cellular Networks
HSIEN-MING TSAISenior Specialist, Research Institute, Quanta Computer Inc., Taoyuan, Taiwan, ROC
AI-CHUN PANG
Asst. Professor, Graduate Institute of Networking and Multimedia, Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan, ROC
YUNG-CHUN LIN
Ph.D. Candidate, Department of Computer Science and Information Engineering, National Chiao Tung University, Hsinchu, Taiwan, ROC
YI-BING LIN∗
Chair Professor, Department of Computer Science and Information Engineering, National Chiao Tung University, Hsinchu, Taiwan, ROC
Abstract. In mobile telecommunications operation, radio channels are scarce resources and should be carefully assigned. One possibility is to deploy the hierarchical cellular network (HCN). This paper studies a HCN channel assignment scheme called repacking on demand (RoD). RoD was originally proposed for wireless local loop networks. We expend this work to accommodate mobile HCN. A simulation model is proposed to study the performance of HCN with RoD and some previously proposed schemes. Our study quantitatively indicates that RoD may significantly outperform the previous proposed schemes.
Keywords: channel assignment, channel repacking, hierarchical cellular network, repacking on demand
1. Introduction
One of the most important issues in cellular network operation is capacity planning. Especially when the number of cellular subscribers grows rapidly, it is required that the cellular ser-vice provider increases its network capacity effectively. One possible solution is to deploy the hierarchical cellular network (HCN) [5,11,13]. As shown in figure 1, the HCN consists of two types of base stations (BSs): micro BSs and macro BSs. A micro BS with low power transceivers provides small radio coverage (referred to as microcell), and a macro BS with high power transceivers provides large radio coverage (referred to as macrocell). The microcells cover mobile subscribers (MSs) in heavy teletraffic areas. A macrocell is overlaid with several microcells to cover all MSs in these microcells.
In a cellular network, radio channels must be carefully as-signed to reduce the numbers of new call blockings as well as handoff call force-terminations. Several channel planning and assignment approaches have been proposed for HCN [1– 4,7,14–18,20]. Some studies focused on channel assignment according to the received radio signal strength [3,15]. Other studies [2,4,7,20] investigated radio channel packing issues for channel reuse. A basic scheme called no repacking (NR) was described in [16]. In this scheme, when a call attempt (either a new call or a handoff call) for an MS arrives, the HCN first tries to allocate a channel in the microcell of the MS. If no idle channel is available in this microcell, the call attempt overflows
∗Corresponding author.
E-mail: liny@csie.nctu.edu.tw
to the corresponding macrocell. If the macrocell has no idle channel, the call attempt is rejected. Call blockings and force-terminations of NR can be reduced by repacking techniques described as follows [21]. Consider a call attempt for a mi-crocell BSi that has no idle channel. In NR, this call attempt
is served by the corresponding macrocell. If radio channels are available in BSi later, this call can be transferred from the
macrocell to BSi again. The process of switching a call from
the macrocell to the microcell is called repacking. Repacking increases the number of idle channels in a macrocell so that more macrocell channels can be shared by the call attempts where no channels are available in the microcells. Depending on the time when repacking is exercised, several schemes have been proposed. In always repacking (AR) [1,14,17], the HCN always moves a call from the macrocell to the corresponding microcell as soon as a channel is released at that microcell. Some schemes [8,19] perform repacking based on the moving speeds of MSs. In [8], the calls of slow-speed MSs are always moved from the macrocells to the corresponding microcells when these MSs move across the borders of microcells.
In [6,19], repacking on demand (RoD) schemes were pro-posed. Unlike AR, RoD does not immediately perform ing when a channel in a microcell is released. Instead, repack-ing is exercised only when it is necessary. Based on the speeds of MSs, the study in [19] investigated RoD for Exponential cell residence times and call holding times. The study in our previous work [6] investigated RoD for wireless local loop. This paper extends the work in [6] to accommodate mobile networks. We consider general distributions for both the cell residence times and call holding times.
720 TSAI ET AL.
Figure 1. Hierarchical cellular network architecture.
Our study develops a simulation model to investigate the performance (i.e., the call blocking, force-termination, and in-completion) for NR, AR, and RoD. In Sections 2 and 3, we describe RoD. In Section 5.1, we propose a discrete event sim-ulation model for HCN channel assignment. In Section 4, we use simulation experiments to compare NR, AR, and RoD. Our study quantitatively shows that RoD outperforms NR and AR.
2. Repacking on demand for hierarchical cellular network (HCN)
This section describes the RoD channel assignment and repacking procedures for HCN. As shown in figure 2, when a call attempt is newly generated from or handed off to the i th microcell, the HCN first tries to assign a channel in the i th microcell to the call attempt (see figure 2(1) and (2)). If no idle channel is available in the i th microcell, the call attempt overflows to the j th macrocell that overlays with the i th mi-crocell. If the macrocell has idle channels, the HCN assigns one to the call attempt (figure 2(3) and (4)). Otherwise, RoD is exercised to identify r epacki ng candi dates (figure 2(5)). Every repacking candidate is an ongoing call that satisfies the following criteria:
Criterion 1. The call occupies a channel in the j th macrocell. Criterion 2. The microcell of this call (i.e., the microcell where
the call party resides) has an idle channel.
RoD selects one repacking candidate to hand off from the macrocell to the microcell where the call resides (see figure 2(6) and (7)). Then the reclaimed macrocell channel is used to serve the call attempt (figure 2(8)). If RoD cannot find any repacking candidate, the call attempt is rejected; i.e., the new call is blocked or the handoff call is force-terminated (figure 2(9)).
We propose two policies to handle the repacking candi-dates in RoD at Step 6 in figure 2. Random RoD (RoD-R) randomly selects a repacking candidate with the same prob-ability. Load Balancing RoD (RoD-L) selects the repacking candidate whose microcell has the least traffic load. Both RoD-R and RoD-RoD-L can be adopted by an HCN that utilizes radio systems such as GSM/PCS1900 [13] or WCDMA [5], where the handoff decision is made by the network.
Figure 2. Channel assignment with RoD.
3. System model for HCN
This section describes the input parameters and output mea-sures for the HCN system model. Our model can accommo-date any cell configurations. For the demonstration purpose, we consider a wrapped mesh cell configuration as shown in figure 3. This configuration consists of four macrocells. Each macrocell covers 4×4 microcells. The wrapped topology sim-ulates unbounded HCN so that the boundary cell effects can be ignored [12]. Without loss of generality, the MS moves to
Figure 3. Hierarchical cellular network with wrapped mesh configuration.
one of the four neighbor microcells with the same probability (i.e., 0.25). Three types of input parameters are considered in our model.
System Parameters: Each macrocell has C radio channels, and each microcell has c radio channels.
Traffic Parameters: The call arrivals to a microcell (for both incoming and outgoing calls) form a Poisson stream with rateλ. The expected call holding time is 1/µ.
Mobility Parameters: The expected microcell residence time of an MS is 1/η.
Several output measures are defined in this study: Pb: the blocking probability that a new call is blocked
Pf: the force-termination probability that a successfully
con-nected call is force-terminated because of handoff failure Pnc: the incomplete probability that a new call is blocked or a
connected call is force-terminated
H : the expected number of handoffs occurred during a call
Based on the source and the target cells of handoff, figure 4 shows five types of handoffs, and five handoff measures are defined.
Hmm: the expected number of handoffs from a microcell to
another microcell during a call (figure 4(a))
Hm M: the expected number of handoffs from a microcell to a
macrocell during a call (figure 4(b))
722 TSAI ET AL.
HMm: the expected number of handoffs from a macrocell to a
microcell during a call (figure 4(c))
HM M: the expected number of handoffs from a macrocell to
another macrocell during a call (figure 4(d))
HR: the expected number of repackings during a call
(figure 4(e))
From the above description, H can be expressed as H = Hmm+ Hm M+ HMm+ HM M+ HR. (1)
Table 1 lists the notation described in this section, which will be used in the remainder of the paper.
When the MS residence times are exponentially distributed, RoD may be model by multi-dimensional Markov chain. Since the details are tedious and the assumption is quite artificial, we will not work on this approach. Instead, we use simulation approach to model more realistic scenarios with general MS residence time distributions. In Appendix A, we describe a
Figure 5. Effects of C on Pb, Pf and Pnc(c= 10, λ = 7µ, Vµ= 1/µ2,η = 0.1µ and Vη= 1/η2).
discrete simulation model for RoD. This simulation model is partially validated by several analytic models [6,13], and the details are omitted.
4. Results and discussions
Based on the simulation model developed in Appendix A, we compare NR, AR, RoD-R and RoD-L in terms of the output measures defined in Section 3. In our numerical examples, the radio channel number is c= 10 for every microcell. We assume that the call holding times have a Gamma distribu-tion with mean 1/µ and variance Vµ (the typical value for 1/µ is 1 minute). We also assume that the microcell residence times have a Gamma distribution with mean 1/η and variance Vη. The Gamma distribution is considered because it can ap-proximate many distributions as well as measured data (see Lemma 3.9 in [9]). Note that when Vµ= 1/µ2(Vη= 1/η2),
Figure 6. Effects of C on Hmm, Hm M, HRand H (c= 10, λ = 7µ, Vµ= 1/µ2,η = 0.1µ and Vη= 1/η2).
the Gamma distribution becomes an Exponential distribution. The call arrivals to every microcell form a Poisson process with the arrival rateλ. In each simulation run, 105− 106call
arrival events per microcell are executed to ensure that simu-lation results are stable. Furthermore, we run replications such that the confidence intervals of the 99% confidence levels are within 3% of the mean values in most cases. The effects of the input parameters are described as follows.
Effect of the macrocell channel number C. Figures 5(a), (b) and (c) plot Pb, Pf and Pnc as functions of C, where λ = 7µ, Vµ= 1/µ2,η = 0.1µ and Vη = 1/η2. These
fig-ures show the intuitive results that for all approaches un-der investigation, Pb, Pf and Pnc decrease as C increases.
When C is small, macrocell channels are the bottleneck resources. Increasing C significantly improves the Pb, Pf
and Pnc performance. When C is larger than a threshold
C∗, the macrocell channels are no longer the bottleneck, and adding extra macrocell channels only insignificantly reduces Pb, Pf and Pnc. The importance of our study is that
for any specific input parameters, we can find the thresh-old C∗. For example, C∗= 12 for RoD and AR in figure 5. Figures 6 (a)–(d) plot Hmm, Hm M, HR and H as functions
of C, whereλ = 7µ, Vµ = 1/µ2,η = 0.1µ and V
η= 1/η2.
For all approaches, figure 6(a) shows that Hmmis a
decreas-ing function of C. On the other hand, Hm M (figure 6(b))
is an increasing function of C. These phenomena are due
Figure 7. Effects ofη on Pnc(C= 8, c = 10, λ = 7µ, Vµ= 1/µ2and Vη= 1/η2).
724 TSAI ET AL.
Figure 8. Effects ofλ on Pnc(C= 8, c = 10, Vµ= 1/µ2,η = 0.1µ and
Vη= 1/η2).
Figure 9. Effects of variance (Vµ) of call holding time on Pnc(C= 8, c = 10,
λ = 7µ, η = 0.1µ and Vη= 1/η2).
to the fact that when C increases, all handoff activities in-volving macrocell channels will increase, and the number of pure microcell to microcell handoffs will decrease. The performance figures for HMm and HM M are similar to that
for Hm M, and the details are omitted. Figure 6(c) shows
that when AR is exercised, HRincreases as C increases for
the following reason. Increasing C results in more repack-ing candidates, and more repackrepack-ings are exercised. On the other hand, for RoD-R and RoD-L, HR increases and then
decreases as C increases. This non-trivial phenomenon is
Figure 10. Effects of variance (Vη) of microcell residence time on Pncand H (C= 8, c = 10, λ = 7µ, Vµ= 1/µ2andη = 0.1µ).
Table 1 Notation.
Notation Description
System parameters C the number of radio channels in a macrocell c the number of radio channels in a microcell
Traffic parameters
λ the call arrival rate to a microcell 1/µ the expected call holding time
Mobility parameters 1/η the expected microcell residence time
Output measures Pb the blocking probability
Pf the force-termination probability
Pnc the incomplete probability
Hmm the expected number of microcell-to-microcell handoffs per call
Hm M the expected number of microcell-to-macrocell handoffs per call
HMm the expected number of macrocell-to-microcell handoffs per call
HM M the expected number of macrocell-to-macrocell handoffs per call
HR the expected number of repackings per call
H the expected number of handoffs per call
explained as follows. When C is small, macrocell chan-nels are the bottleneck resources. Increasing C results in more repacking candidates, and more on-demand repack-ings are exercised (just like AR). When C is large (C> 5 in figure 6 (c)), macrocell channels are no longer the bot-tleneck. Increasing C results in fewer blockings as well as force-terminations, and less on-demand repackings are needed. Therefore HRdecreases as C increases. Figure 6(d)
shows the net effects of repackings and all types of hand-offs. In this figure, H is an increasing function for NR and AR. For RoD-R and RoD-L, as C increases, H increases and then decreases.
Comparison of NR, AR, RoD-R and RoD-L. Figures 5(a), (b) and (c) indicate that repacking approaches (AR, RoD-R and RoD-L) have the same Pb, Pf and Pnc values, which
are much less than those for NR. If the HCN is engineered at Pnc= 2% (see the horizontal dashed line in figure 5(c)),
C= 5.5 for repacking approaches and C = 10.5 for NR, and repacking approaches can save 5 macrocell channels over NR.
Figure 6(c) shows that HR,AR HR,RoD−R > HR,RoD−L > HR,N R= 0. This result is the same as that for WLL,
and the details will not be given. The reader is referred to our previous work for the details [6]. Figure 6(d) indicates that HA R HRoD−R > HRoD−L > HN R. Specifically,
Figure 11. Simulation flow chart for RoD-R.
when the HCN is engineered at C= 5.5 (i.e., Pnc = 2%)
RoD-R and RoD-L reduce 62% and 66% of handoffs over AR, respectively.
Effect of the Mean Microcell Residence Time 1/η. Figure 7 plots Pnc as a function ofη, where C = 8, λ = 7µ, Vµ=
1/µ2, and Vη = 1/η2. This figure shows an intuitive result that for all approaches, Pnc increases asη increases. The
non-trivial result is that to keep the same Pncperformance
(e.g., Pnc= 3.5%), the repacking approaches can support
726 TSAI ET AL.
Effect of the Arrival Rateλ. Figure 8 plots Pncas an
increas-ing function ofλ, where C = 8, Vµ= 1/µ2,η = 0.1µ and
Vη = 1/η2. It shows that to keep the same Pncperformance
(e.g., Pnc = 2%), repacking approaches can support more
call arrivals (more than 11%) than NR.
Effect of the Variance Vµ for the Call Holding Time. Figure 9 plots Pncas a decreasing function of variance Vµ,
where C = 8, λ = 7µ, η = 0.1µ and Vη= 1/η2. Note that,
for the call holding time distributions with the same mean value 1/µ, the standard deviation σ =VµBy the Cheby-shev’s Inequality, the probability that the call holding times are out of range [1/µ −5
√ Vµ 3 , 1/µ +
5√Vµ
3 ] is smaller than
36 percent for all Vµvalues. For example, if Vµ= 100/µ2,
then5
√ Vµ
3 = 50/3µ and the probability that the call holding
time exceeds 53/3µ is smaller than 36 percent. We note that as Vµincreases, more large and small call holding times are observed. More short call holding times implies that more calls are completed before next new call attempts arrive or next handoff attempts are exercised. Thus the numbers of blocked calls and force-terminated calls decrease.
Effect of the Variance Vηfor the Microcell Residence Time. Figures 10(a) and (b) plot Pncand H as functions of variance
Vη, where C = 8, λ = 7µ, Vµ= 1/µ2andη = 0.1µ. These
figures show that for all approaches, Pncand H decrease as
Vηincreases. From the residual life theorem [10], the mean value of the first microcell residence time increases as Vη
increases, which implies that more calls will complete in the first microcell before they are handed off to the next cells. Therefore, both Pncand H drop as Vηincreases.
5. Conclusions
This paper studied repacking on Demand (RoD) for chan-nel assignment in a hierarchical cellular network (HCN). We developed simulation models to investigate the RoD perfor-mance on the blocking probability Pb, the force-termination
probability Pf, the incomplete probability Pnc and the
ex-pected number of handoffs H during a call. We compared RoD with other HCN channel assignment schemes including no repacking (NR) and always repacking (AR). We showed that RoD and AR have the same Pb, Pf and Pncperformance.
Compared with NR, both RoD and AR reduce the Pb, Pf, and
Pncperformance at cost of increasing the number of handoffs.
With the same Pncperformance, the repacking approaches can
support much faster MSs or more call arrivals than NR does. Moreover, our study indicated that RoD results in much less handoffs than AR does.
5.1. Simulation model
This appendix describes the discrete event simulation model for RoD-R. Other strategies (such as NR, AR and RoD-L) can be studied by similar simulation models, and the details are omitted. In this simulation model, three types of events are defined to represent call arrival, call completion, and MS
movement. These events are inserted into the event list, and are deleted/processed from the event list in the non-decreasing timestamp order. A simulation clock is maintained to indicate the simulation progress, which is the timestamp of the event being processed. The following attributes are defined for an event e:
r
The type attribute indicates the event type. An Arrival event represents a new call arrival. A Move event represents an MS movement from one cell to another. A Complete event represents a call completion.r
The ts attribute indicates the time when the event occurs.r
The tc attribute indicates the time when the call correspond-ing to e will complete. Note that tc≥ ts.r
The mc attribute indicates the microcell where the MS (cor-responding to this event) resides.r
The ma attribute indicates the macrocell where the MS (corresponding to this event) resides.r
The is macro attribute indicates whether the call corre-sponding to this event occupies a macrocell channel. If a macrocell channel is occupied, is macro= 1. Otherwise, is macro= 0.An array mc ch[i ] is used to represent the number of the idle channels of microcell i . Another array ma ch[ j ] is used to represent the number of the idle channels of macrocell j . When a handoff occurs, two variables new mc and new ma are used to indicate the next micro and macro cells where the MS moves into. The outputs measured in the simulation are the number N of total call arrivals during the simulation, the number Nb
of blocked calls, the number Nf of force-terminated calls, the
number Nhof handoffs. From the above output measures, we
can compute Pb = Nb N , Pf = Nf N− Nb , Pnc= Nb+ Nf N , HR = NR N− Nb and H= Nh N− Nb . (2)
Figure 11 shows the simulation flow chart for RoD-R, which is described as follows. Step 1 initializes the input parameters. Step 2 generates the first Arrival event for each microcell and inserts these events into the event list. In Steps 3 and 4, the next event e in the event list is processed based on its type. There are three cases:
Case I. e.type= Arrival: At Step 5, if N − Nb> N∗(in our
simulation, N∗ = 6 × 105× 64), then Step 6 computes the
output measures using (2), and the simulation terminates. Otherwise, Step 7 generates the next Arrival event e1 for
the same microcell (i.e., e1.mc = e.mc, e1.ma = e.ma and
e1.is macro=0). The timestamp of e1 equals e.ts plus the
inter call arrival time generated by a random number genera-tor. Then e1is inserted into the event list. At Step 8, if
micro-cell e.mc has idle channels (i.e., mc ch[e.mc] > 0), then a channel is assigned to the incoming call. Then mc ch[e.mc] is decremented by 1 at Step 9. Step 10 computes the call
completion time e.tc as e.ts plus the call holding time. Step 10 also determines the MS move time Th when the MS
moves out of the microcell. Thequals e.ts plus the cell
res-idence time. Then Step 11 determines the next event (i.e., a Move event or Complete event) for the call corresponding to e. If the MS will move to another cell after call comple-tion (i.e., e.tc ≤ Th), then Step 12 is executed to generate
a Complete event e2 where e2.ts = e.tc, e2.mc = e.mc,
e2.ma = e.ma and e2.is macro = e.is macro. Event e2
is inserted into the event list. On the other hand, if the MS moves to another cell before the call completion (i.e., e.tc > Th) at Step 11, then Step 13 is executed to generate
the next Move event e2 with the timestamp Th for this
call (i.e., e2.ts = Th, e2.tc = e.tc, e2.mc = e.mc, e2.ma =
e.ma and e2.is macro = e.is macro). Event e2 is inserted
into the event list.
At Step 8, if microcell e.mc has no idle channel (i.e., mc ch[e.mc] = 0), the call attempt overflows to macrocell e.ma. Step 14 checks if macrocell e.ma has idle channels (i.e., ma ch[e.ma] > 0). If so, a macrocell channel is assigned to the incoming call at Step 15. In this case, ma ch[e.ma] is decremented by 1, and e.is macro = 1. The simulation proceeds to execute Steps 10–12 (or 13).
If macrocell e.ma has no idle channel (i.e., ma ch[e.ma] = 0) at Step 14, then Step 16 checks if there are repacking candidates in macrocell e.ma. If so, Step 17 randomly selects a repacking candidate (i.e., a call corresponding to a Complete or Move event e3) to
be handed off from macrocell e.ma to microcell e3.mc
(i.e., e3.is macro is set to 0), and the reclaimed macrocell
channel is assigned to the incoming call (i.e., e.is macro is set to 1). Step 17 also decrements mc ch[e3.mc] by 1,
and increments both NR and Nhby 1. Then the simulation
proceeds to execute Steps 10–12 (or 13). If no repacking candidate is found at Step 16, the incoming call is blocked, and Nbis incremented by 1 at Step 18.
Case II. e.type= Move: Step 19 selects the next microcell new mc and its macrocell new ma for the MS correspond-ing to event e. At Step 20, if the call of the MS occupies a microcell channel (i.e., e.is macro = 0), the occupied channel of microcell e.mc is released (mc ch[e.mc] is incremented by 1 at Step 21). At Step 22, if microcell new mc has idle channels (i.e., mc ch[new mc] > 0), a channel is assigned to the handoff call. Then at Step 23, the call is handed off to the microcell (i.e., e.is macro = 0), and mc ch[new mc] is decremented by 1. Step 24 incre-ments Nhby 1. Step 25 updates the current cell for the call
(i.e., e.mc = new mc and e.ma = new ma) and computes the next MS move time Thfor the call. Then the simulation
proceeds to execute Steps 11 and 12 (or 13).
At Step 22, if all channels in the new microcell are busy (i.e., mc ch[new mc] = 0), then Step 26 checks whether macrocell new ma has an idle channel. If so (i.e., ma ch[new ma] > 0), a macrocell channel is assigned to the handoff call. Then at Step 27, the call is handed off to the macrocell (i.e., e.is macro is set to 1), and ma ch[new ma] is decremented by 1. The simulation
proceeds to execute Steps 24, 25, 11 and 12 (or 13). On the other hand, if no channel is available in macrocell new ma at Step 26, then Step 28 checks if there are repacking candidates in new ma. If so, Step 29 is executed to repack the calls (same as Step 17), and the simulation proceeds to execute Steps 24, 25, 11 and 12 (or 13). At Step 28, if no repacking candidate is found, the call is force-terminated and Step 30 is executed to increment Nf by 1.
At Step 20, if the call occupies a macrocell channel, then Step 31 checks if the MS is moving out of its macrocell e.ma. If so (i.e., new ma = e.ma), the call is handed off to the new cell. Step 32 increments ma ch[e.ma] by 1; i.e., the occupied channel of e.ma is released. Then the simulation proceeds to execute Steps 22–30 or to execute Steps 22–25, 11 and 12 (or 13). At Step 31, if the MS does not move out of its macrocell (i.e., e.ma = new ma), the simulation proceeds to execute Step 25 and then Steps 11 and 12 (or 13). Case III. e.type= Complete: At Step 33, if the call occupies a macrocell channel (i.e., e.is macro = 1), then Step 34 is executed to increment ma ch[e.ma] by 1. Otherwise, Step 35 is executed to increment mc ch[e.mc] by 1.
To accommodate RoD-L, we only need to modify Steps 17 and 29 of the flowchart in figure 11. The simulation model is partially validated by the analytic model in [6] for the no mobility case. The details are omitted.
Acknowledgment
We would like to thank the anonymous reviewers. Their comments have significantly improved the quality of this paper. This work was sponsored in part by NSC Excellence project NSC93-2752-E-0090005-PAE, NSC 93-2213-E-009-100, NTP VoIP Project under grant number NSC 92-2219-E-009-032, IIS/Academia Sinica, ITRI/NCTU Joint Research Center, Quanta, National Science Council under contract NSC 92-2213-E-002-049, CCL/ITRI under contract T1-92021-6, Microsoft and Intel.
References
[1] R. Beraldi, S. Marano and C. Mastroianni, A reversible heirarchical scheme for microcellular systems with overlaying macrocells, in: Proc. of IEEE Infocom (1996) pp. 51–58.
[2] L.J. Cimini, G.J. Foschini, C.-L.I and Z. Miljanic, Call blocking per-formance of distributed algorithms for dynamic channel allocation in microcells, IEEE Tran. on Comm. 42(8) (1994) 2600–2607. [3] B.O.P. Gudmundson, (Sollentuna, SE), H. Eriksson, (Vallentuna, SE)
and O.E. Grumlund, (Kista, SE), Method of effecting handover in a mobile multilayer cellular radio system. U.S. Patent (1995).
[4] C.-J. Ho, C.-T. Lea and G.L. Stuber, Call admission control in the mi-crocell/macrocell overlaying system. IEEE Tran. on Vehicular Tech. 50(4)(2001) 992–1003.
[5] H. Holma and A. Toskala, WCDMA for UMTS (John Wiley & Sons, 2000).
[6] H.-N. Hung, Y.-B. Lin, N.-F. Peng and H.-M. Tsai, Repacking on de-mand in two-tier WLL. Accepted and to appear in IEEE Transactions on Wireless Communications.
728 TSAI ET AL.
[7] C.-L. I, L.J. Greenstein and R.D. Gitlin, A microcell/macrocell cellular architecture for low- and high-mobility wireless users. IEEE JSAC 11(6) (1993) 885–891.
[8] B. Jabbari and W.F. Fuhrmann, Teletraffic modeling and analysis of flex-ible heirachical cellular networks with speed-sensitive handover strat-egy. IEEE JSAC 15(8) (1997) 1539–1548.
[9] F.P. Kelly, Reversibility and Stochastic Networks (John Wiley & Sons Ltd., 1979).
[10] L. Kleinrock, Queueing Systems; Vol. I: Theory (Wiley, 1975). [11] X. Lagrange, Multitier cell design, IEEE Communications Magazine
35(8) (1997) 60–64.
[12] Y.-B. Lin and V.W. Mak, Eliminating the boundary effect of a large-scale personal communication service network simulation, ACM Tran. on Modeling and Computer Simulation 4(2) (1994) 165–190. [13] Y.-B. Lin, W.-R. Lai and R.J. Chen, Performance analysis for dual band
PCS networks. IEEE Tran. on Computers 49(2) (2000) 148–159. [14] K. Maheshwari and A. Kumar, Performance analysis of microcellization
for supporting two mobility classes in cellular wireless networks. IEEE Tran. on Vehicular Tech. 49(2) (2000) 321–333.
[15] P.A. Ramsdale, (Walden, GB) and P.S. Gaskell, (Shelford, GB), Han-dover techniques. U.S. Patent (1994).
[16] S.S. Rappaport and L.-R. Hu, Microcellular communication systems with hierarchical macrocell overlays: Traffic performance models and analysis, in: Proceedings of the IEEE 82(9) (1994) 1383–1397. [17] R. Steele, M. Nofal and S. Eldolil, Adaptive algorithm for variable
teletraffic demand in highway microcells. Electronics Letters 26(14) (1990) 988–990.
[18] F. Valois and V. Veque, Preemption policy for hierarchical cellular net-work. In: 5th IEEE Workshop on Mobile Multimedia Communication (1998) pp. 75–81.
[19] F. Valois and V. Veque, QoS-oriented channel assignment strategy for hierarchical cellular networks. IEEE PIMRC 2 (2000) 1599–1603. [20] L.-C. Wang, G.L. Stuber and C.-T. Lea, Architecture design, frequency
planning, and performance analysis for a microcel/macrocell overlaying system. IEEE Tran. on Vehicular Tech. 46(4) (1997) 836–848. [21] P.A. Whiting and D.W. McMillan, Modeling for repacking in cellular
radio, in: 7th U.K. Teletraffic Symp. (1990).
Hsien-Ming Tsai was born in Tainan, Taiwan,
R.O.C., in 1973. He received the double B.S. de-grees in Computer Science & Information En-gineering (CSIE) and Communication Engineer-ing, the M.S. degree in CSIE, and the Ph.D. de-gree in CSIE from National Chiao-Tung University (NCTU), Taiwan, in 1996, 1997, and 2002, respec-tively. He is currently a research specialist in Quanta Research Institute, Quanta Computer Inc. His re-search interests are in the areas of cellular protocols
(UMTS/GPRS/GSM/DECT), cellular multimedia (MPEG-4 Audio/Speech), and embedded systems. He is an IEEE member.
Ai-Chun Pang was born in Hsinchu, Taiwan,
R.O.C., in 1973. She received the B.S., M.S. and Ph.D. degrees in Computer Science and Information Engineering from National Chiao Tung University (NCTU) in 1996, 1998 and 2002, respectively. She joined the Department of Computer Science and In-formation Engineering, National Taiwan University (NTU), Taipei, Taiwan, as an Assistant Professor in 2002. Her research interests include design and anal-ysis of personal communications services network, mobile computing, voice over IP and performance modeling.
E-mail: acpang@csie.ntu.edu.tw
Yung-Chun Lin was born in Kaohsiung, Taiwan,
R.O.C., in 1978. He received the B.S. and M.S. de-grees in Computer Science and Information Engi-neering (CSIE) from National Chiao-Tung Univer-sity (NCTU), Taiwan, in 2001, 2003, respectively. He is currently pursuing the Ph.D. degree in CSIE. His research interests include design and analysis of a personal communications services network, the cellular protocols (UMTS/GPRS/GSM), and mobile computing.
Yi-Bing Lin received his BSEE degree from
Na-tional Cheng Kung University in 1983, and his Ph.D. degree in Computer Science from the University of Washington in 1990. From 1990 to 1995, he was with the Applied Research Area at Bell Communications Research (Bellcore), Morristown, NJ. In 1995, he was appointed as a professor of Department of Com-puter Science and Information Engineering (CSIE), National Chiao Tung University (NCTU). In 1996, he was appointed as Deputy Director of Microelec-tronics and Information Systems Research Center, NCTU. During 1997-1999, he was elected as Chairman of CSIE, NCTU. His current research interests include design and analysis of personal communications services network, mobile computing, distributed simulation, and performance modeling. Dr. Lin has published over 150 journal articles and more than 200 conference papers. Lin is an Adjunct Research Fellow of Academia Sinica, and is Chair Professor of Providence University. Lin serves as consultant of many telecom-munications companies including FarEasTone and Chung Hwa Telecom. Lin is an IEEE Fellow and an ACM Fellow.
