A Comprehensive Call Management Strategy for
Congestion Control in Cellular Networks
N. V. Marathe, G. S. Biradar, U. B. Desai, S. N. Merchant
Electrical Engineering Department
Indian Institute of Technology, Bombay
Powai, Mumbai, INDIA 400 076
{nvmarathe,gsbiradar,ubdesai,merchant} @ee.iitb.ac.in
Abstract Congestion in cellular networks during peak hours
and at hot spots is a major problem being faced today almost
everywhere especially in metro cities. Even though capacity
dependent phenomenon this solution proves to be inadequate.
Another solution is 'Dynamic Pricing Scheme' where the
tariff is decided based on the traffic conditions in the cell area
economically viable due to heavy infrastructure costs involved. As
the traffic demand is continuously increasing the problem of
congestion is going to remain forever and therefore needs to be
effectively addressed. Our proposed scheme tries to address the
problem of congestion in cellular networks by introducing a new
concept of call duration control coupled with dynamic pricing and
the major reasons for congestion in the network. We propose to
restrict the duration of such calls depending upon traffic
conditions. At heavy traffic conditions the network restricts the
duration of ongoing calls upto a specific time beyond which the
user has to pay at a higher tariff. At the same time a principle of
call-on-hold is also implemented so that if a newly generated call
[1], [2]. Though this scheme overcomes the shortcomings of
'Time Based Differential Tariff Scheme', it puts financial
burden on users because higher tariff is charged for the entire
duration of the call to all the users who enter the network
under heavy traffic conditions. A thinning algorithm is
presented by Y. Fang in [3] which smoothly reduces the traffic
by Y.eFanmino[3]owhchtsmoothlyreducespth trf
presented
admission rates at the time of congestion depending upon the
type andpriority ofthe service. Wewouldlike to propose here
a new comprehensive call management scheme, which tries to
address the problem of congestion in cellular networks by
introducing a concept of call duration control coupled with
dynamic pricing and the principle of call-on-hold. In our
hold in a queue hoping to get a channel soon. Our scheme reduces
the level of congestion substantially without any compromise with
system utilization and at the same time it also marginally
proposed scheme every user gets an almost guaranteed access
to the network for a specified time at the normal tariff
irrespective of the traffic conditions.
Keywords- congestion; call duration control; dynamic pricing;
It is very clear that the problem of congestion can be solved
with two approaches. One of the natural solutions appears to be
an increase in the network capacity with more base stations, but
it requires huge infrastructure investment and is not
immediately possible everywhere. This solution is also not
economically viable because the capacity of the system is not
then fully utilized at times other than peak hours. It is therefore
very obvious to work for another solution. The congestion can
also be reduced if the traffic demand is reduced by some
means. The traffic demand is proportional to the call arrival
rate as well as average call duration. Therefore if we are able to
reduce the average call duration then it will naturally reduce
the traffic demand and will solve the problem of congestion to
a satisfactory level. Our proposed solution tries to adopt this
approach and the simulations carried out show a remarkable
improvement in call blocking probability. The scheme can be
in the
base station setups of the
easily
serviceimplemented
providers. Also any existing
such solution should not
adversely
generation.
affect the system utilization and revenue
Our
r
t ofe segendition.
solution as
utilizsaio
increase seems to be a natural solution to this problem, it is not
call-on-hold principle. Long duration calls in the network is one of
does not get a traffic channel then it is not blocked but is put on
increases the revenue per unit time.
I.
INTRODUCTION
As the subscriber base of cellular phones is increasing at a
rapid rate the problem of network congestion is becoming
more and more critical. Since the infrastructure capacity is not
increased to that proportion, it is giving rise to intolerably high
value of call blocking probability especially during peak hours
and at hot spots. Some solutions have already been suggested
to this problem. Also it has been observed that the network
becomes almost inaccessible due to exceptionally high traffic
at critical times of either natural or manmade disasters.
Therefore there is an urgent need to find out a scheme which
will reduce call blocking probability without compromising
with system utilization and the revenue generated. One of the
suggested solutions is 'Time Based Differential Tariff
Scheme' where call rates are higher than the normal rates
during peak hours. But this scheme is not flexible and may
sometimes result in underutilization of the system resulting in
the loss of revenue. As the congestion is also a location
II.
on as
PROPOSED SCHEME
explained later satisfies both of these conditions.
The problem of congestion is observed at certain places
called as hot spots and at certain times called as peak hours and
1-4244-06 14-5/07/$20.00 ©)2007 IEEE.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
the system is not fully utilized at other times. It is therefore
needed to flatten the traffic demand curve. A scheme needs to
be worked out such that the users are discouraged to continue
the non-urgent communication beyond certain duration at the
time of heavy traffic demand. They instead can continue with
such communication at other times. This can be achieved by
effective call management strategy proposed by us. In our
proposed scheme we try to address the problem of congestion
with the following approach. A new concept of call duration
control is introduced here and then it is coupled with dynamic
pricing making the system flexible. We also introduce the
principle of call-on-hold which has been discussed in [4], [5],
[6] so that when a newly generated call finds that all the traffic
channels are occupied, then the call is not blocked but is kept
on hold in a queue hoping some channel to become available
shortly. In our scheme when a newly generated call finds that
all the traffic channels are occupied, then the system decides to
issue an advance termination notice of 15 seconds to the
currently longest duration call in the network. The newly
arrived call is then kept on hold in a queue. The traffic channel
becoming free because of termination of any of the ongoing
call in the network is assigned to the newly arrived call, which
was kept on hold at the top of the queue. The caller of currently
longest call to whom termination notice of 15 seconds is issued
has an option to extend the call beyond this limit for which he
has to pay at a higher rate than the normal rate. If this option is
not exercised during the notice period, then the system
terminates the call after the notice period is over. Ideally in this
system call blocking probability becomes zero irrespective of
traffic conditions if no limits are put on the minimum duration
of the ongoing call selected for termination and the maximum
waiting time in the queue for a call put on hold. But to make
the system practical some reasonable limits are required to be
put on both of these quantities. In the proposed system the
longest ongoing call in the network is selected for termination
only if it has completed a certain minimum time. Also when a
call which has been put on hold does not get a traffic channel
within the specified maximum waiting time after its arrival
then it is discarded. We have implemented our scheme for a
single cell environment and obtained the results. We have not
considered the handoff issue here because there will not be any
significant changes in the results due to the handoff
management schemes as they operate independently and will
not have any effect on our call management strategy.
III. SIMULATION MODEL
We have considered a single cell environment for
simulation and number of traffic channels available in each cell
is taken to be equal to 40. The average call duration is
considered to be 180 seconds with an exponential distribution.
The call arrival rate distribution is considered to be Poisson
distribution and the mean call arrival rate is varied in between
600 to 1200 calls per hour for consideration of different traffic
conditions. This is a typical model for a practical cellular
system. In our scheme, at the time of the arrival of the call if a
traffic channel is available then it is allotted to the caller and
the call becomes operational otherwise the call is put on hold.
Each time when a new call finds that no traffic channel is
available to it then currently longest ongoing call is selected for
termination
-1
.,
complete a certain minimum time 'Tmin ' in the network before
such forced termination. Simulations are to be carried out for
different values of 'Tmin'. Such call is to be marked as 'OVER'
and hence it is not to be selected as the longest call for
termination anymore in future. As the user has an option to
continue beyond the prescribed time by paying at a higher rate
certain random proportion of such calls continue even after the
expiry of notice period of 15 seconds. The probability of such
calls is taken as 'Pa'. The additional time of these continued
calls is charged at a higher tariff which is 'k' times the normal
rate. Naturally 'Pa' will be lower for higher value of 'k' and
vice versa. Simulations are to be carried out for different
combinations of 'k' and 'Pa'. Calls put on hold are kept in a
queue and the traffic channels becoming free because of
termination of an ongoing call in the network are assigned
these calls in their order of arrival. A threshold waiting time
'Tw' is fixed and if a call in the queue does not get a channel in
this much time after its arrival then it is discarded and such a
call is marked as 'BLOCKED' call. Four important parameters
are calculated from this simulation and these are call blocking
probability, system utilization, total revenue generated per unit
time and average delay. All these parameters are defined in the
next section.
IV.
SIMULATION RESULTS
Simulations were carried out for existing normal method
with uniform tariff, unrestricted call duration and no provision
of call-on-hold. Simulations were also carried out for the
proposed scheme with different values of 'Tmin', 'Pf' and 'k'. In
all we have considered total six cases for different
combinations of these parameters. The minimum duration of
the longest call to be terminated i. e. 'Tmin' is taken as 60, 120
and 180 seconds. The termination notice period is taken to be
15 seconds. The ratio of peak tariff rate to normal tariff rate
i e 'k' is taken as 2 and 3. A reasonable assumption is made
regarding the probability of continuation of the call after the
expiry of termination notice period. For k 2, this probability
is taken as 0.5 and for k 3 this probability P. is taken as
0.3. This is a quite reasonable assumption and in practice this
probability may even be less. Actual tariff can be decided by
studying the usage profile to ensure appropriate reduction in
call blocking probability. The maximum waiting time 'Tw' is
uniformly taken as 15 seconds and is again a reasonably good
figure, which can also be varied by the operator depending
upon the requirement. From the simulation six different
parameters are found out which are listed below.
1 Call Blocking Probability (PB)
T
This is the most important parameter of merit and is defined
as the ratio of successful calls to the total number of generated
calls. This is expressed in percentage in our results.
2. System Utilization (q)
This is the average occupancy of the system resources i. e.
the traffic channels in this case expressed in percentage.
3. Revenue per Unit Time (S)
This parameter corresponds to the total revenue generated
b tthe service
p
by s
provider per unit time. This iS calculated on the
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
ParmetVlusU5. < .~ . .
basis of the used airtime by the user multiplied by the tariff. In
our case the normal tariff per second is considered as unity and
tariff for extended time is 'k' units.
4. Average Delay (TD)
As the concept of call-on-hold is introduced here, there can
be some delay in arrival of the call and channel assignment to
it. This parameter indicates the average of such delays for all
successful calls.
5. Average Call Duration (Tav)
This is the average time duration of all successful calls in
the system.
6. Call Termination Ratio (RT)
As our system terminates some ongoing calls in the event
of congestion, the ratio of such terminated calls by the system
to the total number of successful calls is determined. Again this
is expressed in percentage in our results.
Six different cases have been considered for different
combinations of the parameters Tmin, k and P, as shown in
Table I. Graphs are plotted for various parameters against the
call arrival rate as shown in figures 1 to 7. Figures 1 and 2
show the graph of the critical parameter i. e. call blocking
probability (PB) against the call arrival rate for the existing
normal method and different cases of proposed scheme.
Figures 3 to 5 show the comparison of other parameters for
normal method and proposed scheme. The graphs shown are
for existing normal method and for the proposed scheme with
Tmin = 120 seconds, k = 3 and P,= 0.3 (Case III). As the
parameters average delay and termination ratio are not defined
for the normal method figures 6 & 7 show graphs of these
parameters against the call arrival rate for the proposed scheme
only (Case III). As the results for different cases of the
proposed scheme are very close to each other those are listed in
a tabular form in Tables II to VII.
In the proposed scheme there will be a reduction in the
duration of some calls as compared to the normal method. The
change in the nature of traffic is reflected in the distribution of
the call duration. Histogram shown in Figure 8 indicates the
exponential distribution of the call duration in normal method.
Histograms in Figures 9 and 10 show the change in the
distribution of call duration for the proposed scheme Case I and
Case III respectively. All the histograms are plotted for the
highest level of traffic with the call arrival rate of 1200 calls
per hour.
TABLE I.
Case No.
CASE PARAMETER DETAILS
Parameter Values
60
kP.
3
Case II
60
2
Case III
Case IV
Case V
120
3
120
2
180
3
Case VI
180
2
Case I
T.j. (in seconds)
0.3
0.5
Call Blocking Probability for Cases. I
V
3
NbNorrnal lethod
-. - Case (Proposed)
-, | Case III (Proposed)
-e--- ase (Posed)
2<
20
ii.
.
...
s15
5
0
:
-..
.
.*
- s
700
80
900
1666
11W0
Call Arrival Rate (Number of calls per hour)
I0
666
Figure 1. Call Blocking Probability (PB) for Cases I, III &V
Call Blocking Probability for Cases
30
Normal Mdethod
--Case 1:(Proposed)
-
;
.
6660
(Proposed)
Case VI (Proposed)
-
o..6.
0
CCase IV
*
2.2
II, TV & VI
..
7
700
..
W
..
9
900
16
l00
..c
..
W
1100
Call Arrival Rate (Number of calls per hour)
0.3
0.5
10
Figure 2. Call Blocking Probability (PB) for Cases II, IV &VI
0.3
0.5
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
200
System Utilization in 00 for Case III
166
Normal Method
Proposed Scheme
90~
~
~
~
Call Duration in Seconds for Case HIl
Average
Normal Method
190
---Proposed Scherme
66
~
~
~
~
166
166
146
0~~~~~~~~~~~~~~~~~~~~~~10
666
766
966
666
Figure 3. System Utilization
126666
76 6606066
---
Average Delay in Seconds for Case III
Normal Method
sed Schem
e
.0 2.6
16
Figure 5. Average Call Duration (Tav)
(%)
~Propo
H
......
16
Call Arrival Rate (Nu~mber of calls per hour)
I4
46
38
1166
Total Reveniue per Unit Time for Case II
42
~~
161]i
Call Arrival Rate (Numrber of calls per hour)
..................................
~~~~~~~..........
a
~~~~~~~~~~~~~~~26
34
P432
666
1
766
666
966
1666
1166
Call Arrival Rate (Number of calls per hour)
Figure 4. Total Revenue per Unit Time (S)
1266
~
666
~ ~
766
666
Call Arrival Rate
966
1666
1166
(Numtber of calls per hour)
Figure 6. Average Delay (TD)
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
1266
Histogram of Call Duration for Case I
Call Teninination Ratio in 00 for Case III
r03
.50
45~~~~~~~~~~~~~~~0
~
~
~ ~ ~ ~~~~~~~~~0
0~
40
30
..
c056
o04
~~~~~30......
..-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~............05.... ...
..
210
~~~~~~~~~~~~~~03
0
3......
rJ2~~~~~~~~~~~~~~~~~~~~~~~~
0
0
3~ ~ ~ ~ ~5~
07
u
a
2~~~~~~~~~~~~~~~C 5
C
Cu
03
0-
L
50
10
I
10 20
20
30
30
0
5
Call
Duration in ...........Seconds.....
Figue
8.HisograofCall
0
0
10
10
20
5
0
5
0
5
Call.. Dur
in.......
tio
Second..............
Duration for NeormlMtonFgrd0sHsormo
Call Durattonion feor CseI
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
0
TABLE II.
Call
Arrival
Rate
600
Normal
Method
0.17
CALL BLOCKING PROBABILITY
Call Blocking Probability ( PB) in %
Case
Case
II
0.01
I
0.01
Case
III
0.01
Case
IV
0.01
AVERAGE DELAY
TABLE V.
Case
V
0.01
Case
VI
0.02
Call
Arrival
Rate
600
Normal
Method
Average Delay in Seconds (TD)
Case
Case
Case
Case
Case
Case
0
I
0.01
II
0.02
III
0.01
IV
0.01
V
0.01
VI
0.01
700
1.87
0.02
0.07
0.02
0.13
0.03
0.08
700
0
0.14
0.19
0.13
0.21
0.17
0.16
800
6.52
0.06
0.39
0.06
0.32
0.13
0.30
800
0
0.50
0.72
0.51
0.79
0.56
0.71
900
13.73
0.11
0.60
0.15
0.58
0.28
1.38
900
0
1.05
1.57
1.09
1.62
1.17
1.82
1000
20.46
0.22
1.33
0.32
2.20
1.13
4.54
1000
0
1.62
2.78
1.66
2.86
2.12
3.70
1100
27.80
0.49
2.12
0.72
4.79
4.37
10.25
1100
0
2.32
3.63
2.52
4.52
4.22
6.13
1200
32.74
0.80
3.27
1.92
9.49
9.42
16.50
1200
0
2.84
4.61
3.76
6.49
6.53
7.94
SYSTEM UTILIZATION
TABLE III.
System Utilization (r) in %
74.23
I
74.21
II
74.01
III
74.22
IV
73.96
V
73.61
VI
74.23
Call
Arrival
Rate
600
700
83.68
84.17
84.11
83.61
84.62
84.12
83.75
800
90.38
90.23
90.83
90.26
90.95
90.15
900
94.03
93.67
94.38
93.76
94.49
93.96
1000
95.69
95.33
96.34
95.43
96.72
96.00
Call
Arrival
Rate
600
Method
Normal
Case
Case
Case
Case
AVERAGE CALL DURATION
TABLE VI.
Case
Case
Normal
Average Call Duration in Seconds (T,,)
Case
Case
Case
Case
Case
Case
180
I
180
700
180
176.5
176.8
175.6
178.0
176.8
176.1
90.85
800
180
166.5
168.1
166.4
168.1
156.6
167.8
94.63
900
180
154.3
155.8
154.3
156.3
154.8
157.8
97.27
1000
180
141.3
145.0
141.6
146.5
143.6
150.7
Method
II
180
III
180
IV
180
V
180
VI
180
1100
96.91
96.57
97.70
96.85
98.07
97.94
98.85
1100
180
131.3
135.0
131.9
139.3
138.5
148.6
1200
97.46
97.31
98.44
97.83
99.08
99.01
99.41
1200
180
121.5
126.5
123.9
135.8
135.8
147.9
TABLE IV.
Call
Arrival
Rate
600
REVENUE GENERATION
Call
Revenue per Unit Time (S)
Normal
Method
Case
Case
29.73
I
29.73
II
29.65
Case
III
29.74
Case
CALL TERMINATION RATIO
TABLE VII.
Case
Case
Arrival
IV
29.62
V
29.48
VI
29.74
Rate
600
Normal
Method
0
Call Termination Ratio (RT) in %
I
0.26
Case
II
0.42
Case
III
0.29
Case
IV
0.14
Case
V
0.21
Case
VI
0.32
700
33.52
33.79
33.80
33.57
34.02
33.78
33.65
700
0
3.02
3.77
2.77
4.17
3.52
3.35
800
36.20
36.48
36.88
36.52
36.99
36.49
36.88
800
0
10.70
13.54
10.78
14.89
11.31
13.49
900
37.66
38.33
38.90
38.32
38.96
38.46
38.93
900
0
21.34
27.44
21.46
27.53
21.48
25.17
1000
38.34
39.43
40.51
39.50
40.62
39.63
40.44
1000
0
31.51
42.83
31.54
40.61
29.29
31.92
1100
38.84
40.57
41.86
40.73
41.60
40.80
41.26
1100
0
44.14
54.65
41.82
46.38
33.76
32.97
1200
39.06
41.41
42.75
41.50
42.23
41.41
41.71
1200
0
50.16
61.91
47.47
47.20
34.61
33.31
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.
V.
DISCUSSIONS AND CONCLUSION
From the results shown in graphs and tables it is very much
clear that our proposed scheme shows a substantial reduction in
call blocking probability without comprising with the system
utilization. In fact there is a marginal rise in the revenue
generated per unit time as compared with existing method.
Thus it takes into consideration the service provider's
viewpoint as well. This is a very important feature of the
proposed scheme making it commercially viable for
implementation. Average delay in establishment of a call is
also within a tolerable range and always user will prefer a
scheme having a delayed connection than a total rejection.
Even though some users may not like the concept of call
termination by the network, users at large will prefer this
scheme once they find that there is a substantial reduction in
call blocking probability. Thus in this paper we have presented
a new comprehensive call management strategy, which tries to
distribute the available resources in a more equitable manner
ensuring fair treatment to all the users. At the same time the
scheme is also flexible and allows a user to utilize the resources
without any limit by paying at a higher tariff. Selection of a
particular case of the proposed scheme is to be decided by the
service provider.
The cellular traffic is continuously increasing and despite
corresponding expansion of the infrastructure to increase the
capacity, some such situations are always bound to occur when
the traffic demand temporarily increases than the capacity of
the system. In such situations there will be a need of an
effective strategy to deal with this situation and our proposed
scheme provides an effective solution, which can be easily
implemented with the existing setup.
REFERENCES
[1]
Jiongkuan Hou, Jie Yang, and Symeon Papavassiliou, "Integration of
Pricing with Call Admission Control for Wireless Networks,"
Proceedings of IEEE VTC 2001 Fall vol 3, pp. 1344-1348, Oct. 2001.
[2]
Jiongkuan Hou, Jie Yang, and Symeon Papavassiliou, "Integration of
Pricing with Call Admission Control to Meet QoS Requirements in
Cellular Networks," IEEE Transactions on Parallel and Distributed
Systems, vol 13, no. 9, pp. 898-910, Sept. 2002.
[3]
Y. Fang, "Thinning Algorithms for Call Admission Control in Wireless
Networks," IEEE Transactions on Computers, Vol. 52, No. 5, pp. 685687, May 2003.
[4]
Y. B. Lin, S. Mohan and A. Noerepel, "Queueing Priority Channel
Assignment Strategies for PCS Hand-Off and Initial Access," IEEE
Transactions on Vehicular Technology, Vol. 43, No. 3, August 1994.
[5]
Kshirasagar Naik, David S. L. Wei, "Call-on-Hold for Improving the
Performance of Dynamic Channel-Assignment strategies in Cellular
Networks," IEEE Transactions on Vehicular Technology, vol 53, no. 6,
pp. 1780-1793, Nov. 2004.
[6] D. Hong and S.S. Rappaport, "Traffic Model and Performance Analysis
for Cellular Mobile Radio Telephone Systems with Prioritized and
Nonprioritized Handoff Procedures," IEEE Transactions on Vehicular
Technology, vol. 35, pp. 77-92, Aug. 1986.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on December 4, 2008 at 01:46 from IEEE Xplore. Restrictions apply.