Anda di halaman 1dari 16

Wireless Personal Communications 25: 277–292, 2003.

© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

A MAC Protocol for New and Handoff Calls


with Finite Population

MIAO MA 1 and ERRY GUNAWAN 2


1 Institute for Infocomm. Research (I2R), 21 Heng Mui Keng Terrace, Singapore 119613, Singapore
E-mail: miaom@i2r.a-star.edu.sg
2 School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798,
Singapore
E-mail: egunawan@ntu.edu.sg

Abstract. A MAC protocol for new and handoff calls with a finite population of users, rather than with an infinite
user population as in dynamic channel reservation scheme (DCRS), is considered. Similar to DCRS, we divide
the wireless channels into shared channels and reserved channels. The handoff calls access any available channel
with probability of one, while the new calls access a shared channel with probability of one and access a reserved
channel with a request probability. We propose three simpler formulae than that used in the existing DCRS in
setting the request probability. In addition, the handoff calls in our proposed protocol are allowed to queue in a
finite buffer. To evaluate the system performance, a mathematical model based on queuing theory, rather than a
simulation method used in DCRS, is developed. It is a general model that can be adopted for any types of request
probabilities including DCRS. Suggestions of how to get the optimum values of the design parameters are also
given.

Keywords: media access control, finite population, request probability, queue size, reserved channel.

1. Introduction

With the increasing demands for the mobile multimedia services, wireless systems have been
designed with micro-picocellular architectures in order to provide a higher capacity. However,
the reduced coverage area of a cell has led to the undesirable consequence of an increase in the
number of handoffs. As the handoff rate increases, bandwidth management and traffic control
strategy, for examples the call admission control, the handoff procedure and etc., become more
challenging problems in wireless networks.
Handoff is an action of switching a call in progress in order to maintain continuity and
the required quality of service (QoS) of the call when a mobile terminal moves from one cell
coverage to another. Due to users’ mobility, the media access control (MAC) protocol becomes
much more complicated in wireless networks. An accepted call that has not been completed
in the current cell may have to be handed off to another cell. During the process, the call may
not be able to gain a channel in the new cell to continue its service due to the limited resource
in the wireless networks, which will lead to the call dropping. Thus, the new calls and the
handoff calls have to be treated differently in terms of the resource allocation. Since the users
tend to be much more sensitive to a call dropping than to a call blocking, handoff calls are
normally assigned a higher priority over the new calls.
Recently, intensive research work on channel allocation schemes has been in progress to
reduce the handoff blocking probability. The most common method to handle handoff requests
278 Miao Ma and Erry Gunawan

with a higher priority over new arrival requests is to employ bandwidth reservation. The
guard channel scheme (GCS) [1] reduces dropped handoff calls by simply reserving a number
of channels exclusively for handoff calls. This scheme is very easy to implement but does
not adapt effectively to the varying traffic conditions. The queuing handoff scheme (QHS)
improves channel efficiency and decreases the handoff blocking probability by permitting
handoff calls to be queued in a buffer. The queuing discipline can be the first-in-first-out
(FIFO) [2] or the measurement-based (MB) [3]. The dynamic channel reservation scheme
(DCRS) [4] introduces a threshold and a request probability into the resource allocation strat-
egy, which provides a good quality of service (QoS) and improves the channel utilization.
More sophisticated schemes have been studied recently and may show good performance
[5–9].
The purpose of this paper is to propose a MAC protocol for new and handoff calls from
a finite population size of users. The work here differs from DCRS in [4] in several respects.
Firstly, three new formulae including the constant request probability scheme (ERPS), the
linear request probability scheme (LRPS) and the exponential request probability scheme
(EPRS) are proposed to set the request probability. Another important contribution of this
paper is that a mathematical model based on queuing theory, rather than simulation as in [4], is
developed to evaluate the system performance of the proposed MAC protocol. The analytical
model developed is applicable for a general request probability function, and therefore is very
useful for comparison study between the various methods of setting the request probability.
The effects of several factors, such as the request probability, the mobility factor, the number
of reserved channels and the queue size, on the system performance are also studied. Thirdly,
unlike DCRS presented in [4] where the handoff calls are dropped immediately, queuing the
handoff in a finite buffer is allowed for handoff calls. A queued handoff will be dropped if it
moves out the handoff area before being accomplished successfully. Finally, we consider the
case with a finite population of users, rather than with an infinite user population as in [4].

2. Traffic Model and MAC Protocol

The focus of our study in this paper is restricted to the traffic flow of the new and the handoff
calls to and from a single cell, since a single cell having a base station is a representative of
the behavior of the other cells in a cellular network. We assume a finite number of users, M,
can access the base station. It is also assumed that each cell has been allocated a fixed number
of channels, which is represented as S.
In general, a real-time video call requires a relatively large amount of bandwidth compared
with a voice call. A voice call occupies one channel while a video call needs multiple ones.
Whenever the effect of the video handoff is considered, the MAC protocol becomes much
more complicated. On the other hand, the data traffic does not have real-time requirement but
it is usually bursty, hence any data packet handoff can be considered in the same manner as the
new data packet. Since the purpose of this paper is to compare and study the performances of
the schemes with different request probabilities, for simplicity, only voice traffic is considered
in each cell.
Let λ, λh , µ and µh denote the arrival rate for the new calls, the arrival rate for the handoff
calls, the average channel holding time for the new calls, and the average channel holding time
for the handoff calls, respectively. We assume that the arrival process for the new calls and the
arrival process for the handoff calls are all Poisson, and the channel holding times for the
A MAC Protocol for New and Handoff Calls with Finite Population 279

new calls and the handoff calls are exponentially distributed. Let ρ(= λ/µ) and ρh (= λh /µh )
denote the traffic load of the new and the handoff calls, respectively.
We also define,
λh = αλ , (1)
where α is a mobility factor and it is the ratio between the handoff and the new call arrivals.
Recent studies [10–12] show that the new call channel holding time and the handoff call
channel holding time might have different distributions and average values. The analytical
study adopting this realistic consideration is possible but at the expense of a greater complex-
ity. Therefore, in this paper, the distinction between the channel holding times for the new
calls and the handoff calls is not made, as it is also commonly practiced in the literature [13–
16]. First, we determined the common average channel holding time for the cell traffic, which
is the combination of the new call and the handoff call traffic. Then we used this parameter to
form the exponential distribution to approximate the channel holding time distribution. Due
to such an approximation, the one-dimensional Markov chain model can be used to derive
the analytical results for the blocking probability performance. Following this approach, the
average channel holding time is given by [12],
1 λ 1 λh 1
= · + · . (2)
µav λ + λh µ λ + λh µh
Let S denote the total number of channels in a cell; and C and G denote the shared channels
and reserved channels, respectively, that is, S = C +G. The MAC protocol allows the handoff
calls to seize any available channel with the probability of one. But the new calls access a
shared channel with a probability of one, and access a reserved channel with an acceptance
probability of pn (n = C, C + 1, . . . , S − 1), which depends on the number of busy channels
(equal to n). Obviously, when pC = pC+1 = · · · = pS−1 = 1, it becomes the non-priority
scheme, and when pC = pC+1 = · · · = pS−1 = 0, this scheme becomes the guard channel
scheme (GCS). We also observe that when pC ≥ pC+1 ≥ · · · ≥ pS−1 , the new call throughput
becomes less and less as the number of busy channels increases. The idea behind this scheme
is to gracefully decrease the new call throughput as the system traffic is building up. Thus,
when the system is approaching congestion, the admitted new call stream becomes less, and
hence more handoff calls are permitted. Due to the flexible choice of setting the new call
request probabilities, this scheme can be made very general [12].
If a new call fails on requesting a channel, or if it finds that all the channels are busy at
its arrival, it is blocked and cleared immediately. As for handoff calls, there are two different
schemes considered: blocking and queuing. For the blocking scheme, a handoff request is
blocked and cleared at once like a new call when there is no free channel. On the other hand,
for the queuing scheme, a handoff call is allowed to queue in a buffer with a finite queue size
Q. The queue discipline is assumed to be first-in-first-out (FIFO), and we have 0 ≤ Q ≤
M − S. The queued handoff call will be dropped from the queue by the system if it moves out
of the handoff area before it can be served successfully. The handoff area is the overlapped
coverage area between the cell a mobile user is leaving from and the cell it is entering to. In the
handoff area, the queued handoff can still continue communicating with the cell it is leaving
from. The length of time a mobile spends in the handoff area is called the dwell time of the
handoff call. In this paper, we assume that the handoff dwell time is always shorter than the
allowable delay for the queuing handoff in the buffer. Therefore, if a handoff call is queued in
the target cell when no channel is immediately available, the handoff dwell time is the allowed
280 Miao Ma and Erry Gunawan

queue waiting time. Recent paper [17] presents a more accurate modeling on the dwell time.
But due to the extra complexity it adds to the analysis, in this paper we approximately assume
the dwell time of the waiting handoff to be exponentially distributed [16, 18] with mean 1/µw ,
and we have,
µw = βµav , (3)
where β is queuing factor and it represents the ratio between the dwell time inside the handoff
area and the average channel holding time.
In this paper, we consider two basic blocking probabilities. The first one is the new call
blocking probability, Pb , which is defined as the probability that a new call arrival fails on
obtaining a channel or finds all channels are busy. The second one is the handoff blocking
probability, Ph , which is defined as the probability that a handoff call finds the queue buffer is
full or finds all channels are busy on its arrival.

3. Queuing Analysis

The state number of the system is defined as the sum of the number of calls in service and in
queue. Here we consider the case where we have a Poisson input process with a population of
possible users, rather than with an infinite user population. The system structure is such that
we have a total of M users. A user is either in the system (consisting of S servers and a queue
whose size is Q) or outside the system and in some sense “arriving”. In particular, when a
user is in the “arriving” condition then the time it takes him to arrive is a random variable
with an exponential distribution whose mean is 1/(λ + λh ) . All users act independently of
each other. As a result, when there are n users in the system (queue plus service) then there
are M − n users in the arriving state, and therefore, the total average arrival rate in this state
is (M − n)(λ + λh ) [19].
Firstly, we consider the case that a handoff will be blocked immediately whenever all
channels are busy. The total S available channels are divided into G reserved channels and C
shared channels. If there is a new call arrival, it will be attended only if the number of busy
channels at that time is less than C, otherwise the call is subjected to a certain probability of
acceptance that depends on the number of busy channels at the instant of its arrival. On the
other hand, handoff arrivals will only be blocked, hence producing a dropped handoff, if the
number of busy channels in the cell is equal to S.
The case can be modelled by a Markov process with (S +1) states, but a distinction must be
made between the system behavior when the number of busy channels is less than C and when
it is more than C but less than S. Let Pn be the steady state probability of n busy channels.
Then for 0 ≤ n < C, a transition rate from state Pn to Pn+1 is given by (M − n)(λ + λh ) and a
transition from state Pn+1 to Pn is given with a rate nµav because the channel occupancy time
is exponentially distributed. For C ≤ n < S, a transition rate from state Pn to Pn+1 is given by
(M − n)(pn λ + λh ) and a transition from state Pn+1 to Pn is given with a rate nµav because
a new call arrival can access a reserved channel with an acceptance probability pn when there
are C or more busy channels in the cell. In other words, if a new call arrives at that moment,
the call will be rejected with a probability of (1 − pn ), but if a handoff arrival occurs, it will
be attended. The state diagram that describes the system is shown in Figure 1.
The birth and death rates are expressed as,

(M − n)(λ + λh ) 0≤n<C
λn = (4)
(M − n)(pn λ + λh ) C ≤ n < S
A MAC Protocol for New and Handoff Calls with Finite Population 281

Figure 1. State transition diagram for non-queuing scheme.

µn = nµav 0 ≤ n < S. (5)

Considering the state transition diagram of Figure 1, the steady state probability, Pn is
determined as follows.
First, for the range 0 < n ≤ C we have,
  
M λ + λh n
Pn = P0 . (6)
n µav
For the region C < n ≤ S we have,
  C 
n−1  
M λ + λh pi λ + λh
Pn = P0 . (7)
n µav µav
i=C

Using the normalization equation,



S
Pn = 1 (8)
n=0

then we get P0 ,
 C       n−1  −1
 M λ + λh n S
M λ + λh C  pi λ + λh
P0 = 1 + + . (9)
n µav n µav µav
n=1 n=C+1 i=C

The probability that a new call arrival fails on obtaining a channel or finds all channels are
busy, Pb , is computed as,

S−1
Pb = (1 − pn )Pn + PS , (10)
n=C

where PS is the probability that all S channels are busy.


The probability that a handoff call finds all channels are busy on its arrival, Ph , can be
computed as,
Ph = PS . (11)
Secondly, we study the case that handoff is permitted to be queued in a finite buffer when-
ever all channels are busy at its arrival. Since the queue size is Q, and the maximum waiting
time is exponentially distributed with a mean value 1/µw . The state transition diagram is
shown in Figure 2. If there is a handoff arrival, it will be attended if the number of busy
channels is less than S, otherwise the handoff will be queued in a buffer. In other words, the
282 Miao Ma and Erry Gunawan

Figure 2. State transition diagram for queuing scheme.

total number of calls (those in the queue and in service) in the system is no more than (S + Q),
otherwise a dropped handoff is produced.
A Markov process with (S + Q + 1) states can model this case. The only difference is that,
another distinction must be made between the system behavior when the number of the calls
in service plus in queue is more than S but less than (S + Q). Then for S ≤ n < S + Q, a
transition rate from state Pn to Pn+1 is given by (M − n)λh and a transition from state Pn+1 to
Pn is given with a rate Sµav +(n−S)µw because the new call arrivals are blocked immediately
when there are S or more calls in service plus in queue in the cell, and if a handoff arrives at
that moment, it is allowed to queue in a buffer with queue size Q. And the queued handoff
calls are blocked as they move out of a handoff area before being accomplished.
The birth and death rates of Figure 2 are expressed as,

 (M − n)(λ + λh ) 0≤n<C
λn = (M − n)(pn λ + λh ) C ≤ n < S (12)

(M − n)λh S ≤n<S +Q

nµav 0≤n<S
µn = (13)
Sµav + (n − S)µw S ≤ n < S + Q

Considering the state transition diagram of Figure 2, the steady state probability, Pn is
calculated as follows:
First, for the range 0 < n ≤ C we have,
  
M λ + λh n
Pn = P0 . (14)
n µav
For the region C < n ≤ S we have,
  C 
n−1  
M λ + λh pi λ + λh
Pn = P0 . (15)
n µav µav
i=C

For the region S < n ≤ S + Q we have,


 C S−1
  n−S
 
M! λ + λh pi λ + λh λh
Pn = P0 . (16)
(M − n)!S! µav i=C
µav j =1
Sµav + j µw

Using the normalization equation,


S+Q
Pn = 1 (17)
n=0
A MAC Protocol for New and Handoff Calls with Finite Population 283

then we get P0 ,
 C       n−1  
 M λ + λh n  S
M λ + λh C  pi λ + λh
P0 = 1 + +
n µav n µav µav
n=1 n=C+1 i=C

   S−1   n−S   −1 (18)


λ + λh C  pi λ + λh 
S+Q
M! λh
+ .
n=S+1
(M − n)!S! µav i=C
µav j =1
Sµav + j µw

The probability that a new call arrival fails on obtaining a channel or finds all channels are
busy, Pb , is computed as,

S−1 
S+Q
Pb = (1 − pn )Pn + Pn . (19)
n=C n=S

The probability that a handoff call finds the queue buffer is full on its arrival, Ph , can be
computed as,
Ph = PS+Q . (20)
In order to evaluate the relative penalty from Pb and Ph , we define an overall blocking
probability (or cost function), B, as follows,
B = γ Pb + (1 − γ )Ph , (21)
where γ is a weighting factor, and 0 ≤ γ ≤ 1 . The value of γ depends on the emphasis laid
on the quality of service requirements for Pb and Ph . Usually, interruptions of handoff calls
upset customers much more than blocking of new calls. In other words, Ph is more significant
than Pb , hence the value of γ should be confined to be less than 0.5.

4. Case Study

For the proposed MAC protocol, the request probability reflects the possibility of the system
to permit new calls to access the reserved channels. It has the effect on determining the good
functionality and performance of the proposed MAC protocol. Setting the request probability
to different values and functions will lead to different channel assignment schemes which
have different design goals, i.e., different tradeoff combinations between the new and handoff
call performances. Five different formulae of setting request probability are discussed and are
summarized in Table 1. The first two (GCS and DCRS) are the existing schemes and the last
three (CPRS, LRPS and ERPS) are the proposed new formulae to set the request probability.
The existing guard channel scheme (GCS) [1] is the special case when the request prob-
ability is set to zero, which gives the highest priority to the handoff calls than the new calls
since the new calls are strictly prohibited to access the reserved channels. It is expressed as,
pn = 0, C ≤ n < S. (22)
The dynamic channel reservation scheme (DCRS) [4] sets the request probability as a
function of mobility factor α, and it is calculated by the heuristic formula as follows,

S−n 2π(n − C) 1/2
pn = max 0, α + (1 − α) cos , C ≤ n < S. (23)
G 4G
284 Miao Ma and Erry Gunawan
Table 1. Request probabilities for different schemes.

Scheme Request probability, pn (C ≤ n < S)

GCS 0.0
     
2π(n−C) 1/2
DCRS max 0, α S−n
G + (1 − α) cos 4G
CRPS 0.3
LRPS 1.0 − n−C
G α
ERPS 1.0 − (n−C)
G

The proposed three new formulae to set the request probability use much simpler functions
than that of the DCRS given in Equation (23). The first one called the constant request proba-
bility scheme (CRPS) sets pn to a constant fractional value, rather than zero as in GCS. As an
illustration pn is set to an arbitrary 0.3, and we have,
pn = 0.3, C ≤ n < S. (24)
The second one called the linear request probability scheme (LRPS) sets pn as a linear
function of n, where the value of pn decreases linearly with an increasing n. It is expressed as,
n−C
pn = 1.0 − , C ≤ n < S. (25)
G
The third called the exponential request probability scheme (ERPS) sets pn as an
exponential function of n. It is calculated by,
(n − C)α
pn = 1.0 − , C ≤ n < S. (26)
G
The rational for the second and the third new schemes is that when the traffic load becomes
heavy, the system should give more opportunities to the handoff calls rather than the new calls
to access the channel by setting PC ≥ PC+1 ≥ · · · ≥ PS−1 as discussed in Section 2.
Figure 3 shows an example of pn determined by Equation (26), with S = 40 and G = 8. It
is observed that, when α < 1.0, pn decreases gradually, providing more opportunity to the new
calls to access the reserved channels compared with LRPS. On the other hand, when α > 1.0,
pn decreases abruptly, providing more access chances to handoff calls compared with LRPS.
When α = 1.0, pn then reverts to Equation (25) and decreases linearly and coincides with that
of LRPS.
If it is desired, the acceptance probability could be combined with some other sophisticated
methods, for example, measurement-based methods, fuzzy logic control, neural networks and
etc. In addition, the number of reserved channels as well as the queue size can also be set
adaptively to different traffic loading conditions and mobility characteristics, etc.

5. Numerical Results

In this paper, five different strategies namely GCS, CRPS, LRPS, ERPS and DCRS are studied
and compared. The analysis introduced in Section 3 is used to generate the numerical perfor-
mance results. The new call blocking probability Pb , the handoff call blocking probability Ph ,
A MAC Protocol for New and Handoff Calls with Finite Population 285

Figure 3. Request probability for ERPS (S = 40, G = 8).

Table 2. System parameters (M = 80, S = 40, β = 5.0, γ = 0.3,


1/λ = 250, 1/µ = 150, ρ = 0.6, 1/µh = 100).

Figures 1/λh ρh α G Q

Figures 4–6 100–1000 0.1–1.0 0.25–2.5 8 0


Figures 7–9 200 0.5 1.25 1–10 0
Figures 10–12 200 0.5 1.25 8 0–8

as well as the overall blocking probability B, are evaluated and compared for various system
parameters. We choose the system parameters as listed in Table 2.
Figures 4–6 illustrate the effect of the mobility factor. For each scheme, the blocking
probabilities for the new and the handoff calls increase along with the increase in the mobility
factor, since a higher mobility means there are more handoff call arrivals. Compared with the
GCS, the CRPS improves the new call performance with the expense of sacrificing the handoff
call blocking probability, as it gives the new calls 30% possibility to access the reserved chan-
nels. The overall blocking probability of CRPS is slightly better than GCS. The comparison
between LRPS, ERPS and DCRS are given as follows. When α = 1.0, the three different
schemes have the same new and handoff blocking performances. This is due to the fact that
pn for the three schemes converge to the same function when α = 1.0. When α < 1.0, LRPS
achieves the best Ph performance as well as the worst Pb performance. This phenomenon is
reasonable since Figure 3 suggests that pn of LRPS decreases most abruptly while that of
ERPS decreases most gradually. Similarly, when α > 1.0 , the order of Ph performance from
the best to the worst is: ERPS, DCRS and LRPS, which is consistent with the rate of decrease
of pn shown in Figure 3. Figure 6 demonstrates that the overall blocking probability for each
scheme degrades with increasing α. This is because the handoff traffic load is getting heavy
286 Miao Ma and Erry Gunawan

Figure 4. New call blocking probability versus the mobility factor.

Figure 5. Handoff call blocking probability versus the mobility factor.

(see Table 2). Table 3 summarizes the performance comparison between LRPS, ERPS and
DCRS.
The effect of the number of reserved channels to the system performance is illustrated in
Figures 7–9. Obviously, by providing more reserved channels to the handoff calls, the new
call blocking probability increases and the handoff call blocking probability decreases for all
the schemes. However, Figure 7 also suggests that LRPS achieves a much better new call
performance than that of ERPS and DCRS [4]. This is due to the fact that LRPS brings more
opportunities to new calls to access the reserved channels when α = 1.25 (see Table 2).
A MAC Protocol for New and Handoff Calls with Finite Population 287

Figure 6. Overall blocking probability versus the mobility factor.

Table 3. Performance comparison of LRPS, DCRS and ERPS.

Schemes Pb Ph
α < 1.0 α = 1.0 α > 1.0 α < 1.0 α = 1.0 α > 1.0

LRPS Bad Same Good Good Same Bad


DCRS Average Same Average Average Same Average
ERPS Good Same Bad Bad Same Good

Meanwhile, as expected, GCS obtain the best handoff blocking performance as shown in
Figure 8, since it strictly prohibits the new calls to access any reserved channel. On the other
hand, the overall blocking probability, shown in Figure 9, demonstrates a way to balance
these two conflicting performance measures. We observe that in this particular situation the
optimum value of G is 4 for LRPS, ERPS and DCRS and the optimum value of G is 2 for
GCS and CRPS.
Figures 10–12 deal with the variation of the queue size. For each scheme, as the queue
size increases, the handoff call blocking probability improves significantly without deteri-
orating the new call performance. However it reaches a stable condition when the queue
size is larger than four. Therefore, there is an optimum queue size necessary to improve the
handoff performance which can be determined easily since increasing the queue size beyond
the optimum value does not improve the handoff performance any longer. Similar to Figure 7,
LRPS achieves the best new call blocking performance since we also have α = 1.25 (see Ta-
ble 2). The handoff performance is the best for GCS, which is consistent with the observation
in Figure 8. Particularly, Figure 12 demonstrates that for each scheme the overall blocking
probability no longer improves either when the queue size increases beyond the optimum
value.
288 Miao Ma and Erry Gunawan

Figure 7. New call blocking probability versus the number of reserved channels.

Figure 8. Handoff call blocking probability versus the number of reserved channels.

6. Conclusions

A MAC protocol for new and handoff calls with a finite population is proposed. Wireless
channels are divided into shared channels and reserved channels. The handoff calls access
any available channel with probability one, while the new calls access a shared channel with
probability one and access a reserved channel with a request probability. An analytical model
based on the queuing theory is developed to evaluate the system performance. This is a general
model that is suitable for any request probability values or functions. Three simpler formulae
A MAC Protocol for New and Handoff Calls with Finite Population 289

Figure 9. Overall blocking probability versus the number of reserved channels.

Figure 10. New call blocking probability versus the queue size.

than that used in the existing DCRS in setting the request probability are proposed. Compared
with DCRS, LRPS achieves a better handoff performance at the low mobility case, while at
the high mobility case ERPS has a better handoff performance. The effects of the important
parameters on the system performance are also investigated. Increasing the number of the
reserved channels helps decreasing the handoff call blocking probability dramatically but with
the expense of the new call performance. Therefore the optimum value of the number of the
reserved channels should be made by minimizing the overall blocking probability, which is a
good balance between the two conflicting design goals. Queuing handoff calls significantly
290 Miao Ma and Erry Gunawan

Figure 11. Handoff call blocking probability versus the queue size.

Figure 12. Overall blocking probability versus the queue size.

improves the handoff call performance without degrading the new call performance. The
optimum queue size can be easily obtained since beyond which the handoff call blocking
probability remains the same.
A MAC Protocol for New and Handoff Calls with Finite Population 291

Acknowledgements

The authors are thankful to unknown referee for their helpful comments and suggestions for
the great improvement of the present paper.

References

1. D. Hong and S. Rappaprot, “Traffic Modeling and Performance Analysis for Cellular Mobile Radio Tele-
phone Systems with Prioritized and Nonprioritized Handoff Procedures”, IEEE Transacations on Vehicular
Technology, Vol. VT-35, pp. 77–92, 1986.
2. S. Tekinay and B. Jabbari, “Handover and Channel Assignment in Mobile Cellular Networks”, IEEE
Communincations Magazine, Vol. 29, pp. 42–46, 1991.
3. S. Tekinary and B. Jabbari, “A Measurement Based Prioritization Scheme for Handovers in Cellular and
Microcellular Networks”, IEEE Journal on Selected Areas in Communications, Vol. SAC-10, pp. 1343–
1350, 1992.
4. Y.C. Kim and D.E. Lee et al., “Dynamic Channel Reservation Based on Mobility in Wireless ATM
Networks”, IEEE Communincations Magazine, Vol. 11, pp. 47–51, 1999.
5. W. Li and A.S. Alfa, “Channel Reservation for Handoff Calls in a PCS Network”, IEEE Transactions on
Vehicular Technology, Vol. VT 49-1, pp. 95–104, 2000.
6. R. Fantacci, “Performance Evaluation of Prioritized Handoff Schemes in Mobile Cellular Networks”, IEEE
Transactions on Vehicular Technology, Vol. VT 49-2, pp. 485-493, 2000.
7. D.L. Lee and D.H. Cho, “Performance Analysis of Channel-Borrowing Handoff Scheme Based on User
Mobility in CDMA Cellular Systems”, IEEE Transactions on Vehicular Technology, Vol. VT 49-6, pp. 2276–
2285, 2000.
8. C.H. Paik and G.W. Jin et al., “Integrated Call Control in a CDMA Cellular System”, IEEE Transactions on
Vehicular Technology, Vol. VT 50-1, pp. 97–108, 2001.
9. J.W. Chang and D.K. Sung, “Adaptive Channel Reservation Scheme for Soft Handoff in DS-CDMA Cellular
Systems”, IEEE Transactions on Vehicular Technology, Vol. VT 50-2, pp. 341–353, 2001.
10. Y. Fang, I. Chalamtac and Y.B. Lin, “Channel Occupancy Times and Handoff Rate for Mobile Computing
and PCS Networks”, IEEE Transactions on Computer, Vol. 47, No. 6, pp. 679–692, 1998.
11. Y. Fang and I. Chalamtac, “Teletraffic Analysis and Mobility Modeling for PCS Network”, IEEE Transaction
on Communications, Vol. 47, No. 7, pp. 1062–1072, 1999.
12. Y. Fang and Y. Zhang, “Call Admission Control Schemes and Performance Analysis in Wireless Mobile
Networks”, IEEE Transactions on Vehicular Technoloby, Vol. 51, No. 2, pp. 371–382, 2002.
13. T.S. Yum and K.L. Yeung, “Blocking and Handoff Performance Analysis of Directed Retry in Cellular
Mobile Systems”, IEEE Transactions on Vehicular Technology, Vol. 44, No. 3, pp. 645–650, 1995.
14. J. Wu and J. Lin, “Performance Analysis of Voice/Data Integrated CDMA System with QoS Constraints”,
IEICE Transactions on Communications, Vol. E79-B, No. 3, pp. 384–391, 1996.
15. E.D. Re, R. Fantacci and G. Giambene, “Handover Queuing Strategies with Dynamic and Fixed Channel Al-
location Techniques in Low Earth Orbit Mobile Satellite Systems”, IEEE Transactions on Communications,
Vol. 47, No. 1, pp. 89–102, 1999.
16. F. Barcelo, “A Scheme to Handle Fresh and Handoff Traffic Based on State-Dependent Rejection”, in
Proceedings of IEEE GLOBECOM, San Francisco, U.S.A., 2000, pp. 1522–1527.
17. M. Ruggieri, F. Graziosi and F. Santucci, “Modeling of the Handover Dwell Time in Cellular Mobile
Communications Systems”, IEEE Transactions on Vehicular Technology, Vol. 47, No. 2, pp. 489–498, 1998.
18. C.J. Chang, T.T. Su and Y.Y. Chiang, “Analysis of a Cutoff Priority Cellular Radio System with Finite
Queuing and Reneging/Dropping”, IEEE/ACM Transactions on Networking, Vol. 2, No. 2, pp. 166–175,
1994.
19. L. Kleinrock, Queuing System, Wiley, New York, 1976.
292 Miao Ma and Erry Gunawan

Miao Ma was born in Anshan city, P. R. China on November 28, 1970. She received the
B.Eng. degree in 1992 and the M.Eng. degree in 1996, both from the department of Electrical
Engineering at Harbin Institute of Technoglogy (HIT), China. From July 1999 to July 2002
she studied towards Ph.D. degree in the School of Electrical and Electronic Engineering at
Nanyang Technological University (NTU), Singapore.
She worked as an assistant engineer at AnGang Telecommunication company, Anshan
city from 1992 to 1994, and a research and development engineer at HuaKe Communication
company, Beijing from 1996 to 1999. Since August 2002 upon her graduation from NTU, she
has been a Senior Engineer at Institute for Infocomm Research (I2R), Singapore. Her research
interests include media access control (MAC), resource management, wireless QoS, mobility
management and performance evaluation.

Erry Gunawan received the B.Sc. degree in electrical and electronic engineering from the
University of Leeds, U.K., in 1983 and the M.B.A. and the Ph.D. degrees, both from Bradford
University, in 1984 and 1988, respectively. From 1984 to 1988, he was a Satellite Commu-
nication System Engineer at Communication Systems Research Ltd, Ilkley, U.K. In 1988, he
moved to Space Communication (SAT-TEL) Ltd, Northampton, U.K. He joined the School
of Electrical and Electronic Engineering, Nanyang Technological University, Singapore, in
1989, and currently, he is an Associate Professor in the same school. His research interests in-
clude the fields of digital communications, mobile and satellite communications, error coding,
and spread spectrum. He has published over 60 international research papers and has been a
consultant to a local company on the study of DECT system and BLUETOOTH.

Anda mungkin juga menyukai