Sie sind auf Seite 1von 52

Minimization of Handoff Cost

in Next Generation Wireless


Cellular Network Using Dual
Homed RNCs:
A Metaheuristic Based
Approach
By- Bedadipta Bain
Examination roll number
M4CTL14-04
Under the guidance of
Dr. Madhubanti Maitra

Contents
Introduction
Problem formulation
Solution methodology
Performance evaluation
Conclusions

Introduction
Optimal arrangement of network play the key roll
in
optimization of total cost of operation for modern cellular network
[3].
Re-planning of network has to be done time to time because of
the dynamic behaviour of subscriber .
Therefore, post deployment tuning process is a key factor for
minimizing the total cost [2].
As a result dual-homing cell to switch assignment is gaining
attention as a post deployment tuning process for modern cellular
network.

Introduction
[Cell to Switch Assignment (CSA) ]
Assigning each cell to switch such a way that the total cost would
be minimum
2

High complex handoff


4

1
High complex handoff

3
2
1

Motivation and
Most of the researchObjective
endeavors, which have dealt with the CSA
problem of 2G/2.5G network architecture [3]-[19], have only
considered the aspects of single homing arrangement.
A few literatures have discussed the aspects of dual-homing
arrangement in a 2G/2.5G network.
The UMTS network consists of three-level equipments, and hence
finding the optimal arrangement in a UMTS network becomes much
more challenging than that of the 2G/2.5G wireless networks [1].
Moreover, UMTS network is specially designed for both the voice
and Data services Therefore, the cost function should be formulated
in such a unified manner that it would take care of costs incurred for
completion of successful handoff between the two adjacent RNCs
S. K. Sadhukhan et al. [3], [26] have made a research endeavor to
extend the beneficial features of dual-homing assignment technique
in 3G-network architecture.

Motivation and
They have proposed
a state space based search technique,
Objective
namely Optimal Dual Home (ODH) in [3], to find out the optimal
solution of the CSA problem in 3G network using a dual-homing
arrangement.
They also proposed other stander metaheuristics technique like
TS,SA and ACO[26].
Therefore, in order to provide a plausible solution to the CSA
problem, a novel state space based search method, namely Best
Contributor Search (BCS) has been proposed to address the
assignment problem of a dual-homed UMTS network.

Although, extensive simulation studies reveal the fact that there is


a tradeoff between the cost of the solution obtained by BCS and ACO.
To overcome the shortcoming of BCS, the author has further
modified BCS to develop an advanced optimization algorithm,
6
namely Modified BCS (MBCS), to find out an optimal arrangement and

CSA UMTS: Problem


Formulation
[assumptions]
Without any loss of generality,
we will go by these following
assumptions.
I.
There are M number of MSCs , S number of SGSN and N
number of RNCs.
II. Locations of switches are known and fixed.
III. MSCs and SGSNs are collocated.
IV. Number of MSCs is equal to the number of SGSN.
V.
Already a single home
is available.
i 1,Larrangement
, N

i denote the ith RNC wherej 1,L , M

s 1,L , S
j denote the jth MSC where

s denote the sth SGSN where

CSA UMTS: Problem


Formulation
H
h H
, where
ii'

MSC

MSC

hii' ' H SGSN

N N

[important variables]

H SGSN
, where

cij CMSC , where


cis CSGSN , where

MSC

N M

SGSN

N S

N N

C CMSC CSGSN
CSIC j CSIC j 1,L , M
DSICs DSIC s 1,L , S

CSA UMTS: Problem


Formulation
[Assignment variable ]

In dual homing arrangement, let us define

xij =1 if i is assigned to j
0 Otherwise
xis =1 if i is assigned to s
0 Otherwise
For single home network, let us define
zij =1 if i was assigned to j
0 Otherwise
z 'is =1 if i was assigned to s
0 Otherwise

CSA UMTS: Problem


Formulation

Let us also define


And
similarly
And
Therefore

yii ' xij zi ' j


yi'i yii '
y 'ii ' x 'is z 'i 's
y 'i'i y 'ii'

yii ' 1 if i and i ' are connected with same MSC


0 Otherwise
y 'ii ' 1 if i and i ' are connected with same SGSN
0 Otherwise

10

CSA UMTS: Problem


Formulation
[Cost function ]

Total cost can be expressed as


N

Cost ( X ) Cost ( X 0 )

hii yii ' hii y 'ii '

i 1 i 1,i i
i 1 i 1,i i

N M

N M

(1)
i 1 j 1

i 1 j 1

cij xij cis xis

Total cost after


the single
homing
arrangement

Total cable cost


After the
Dual home assignment

Total eliminated
handoff cost

11

CSA UMTS: Problem


Formulation
[Constraints ]
Link constraint
Capacity constraint
N
CSIC j CSCRi xij
i 1

(2)

DSICs DSCRi xis


i 1

Cost constraint

12

CSA UMTS: Solution


Methodology
Best contributor search (BCS): A state space based heuristics
method

Modified best contributor search (MBCS) : An advanced


optimization algorithm to overcome the shortcoming of BCS

13

Best Contributor Search


(BCS)
M1

R3

R2

M
2

R1

R4
h14

h24

R5
h15

h25
h35
Figure.1. Network layout

14

Best Contributor Search


(BCS)

QUE h14 h15 h 24 h 25 h35

R1 M2
R2 M2

15

Best Contributor Search


(BCS)
[ Time complexity ]

The worst-case scenario is when BCS has to run for the maximum
Number of iterations
Therefore, without loss of generality let us assume the followings:
1. There is no capacity constraint.
2. There is no cost constraint.
c2N
3. Initially there are
number of non-zero complex
handoff costs.

16

Best Contributor Search


(BCS)
[ Time complexity ]
Initial

M1

M2

M3

QUE h12 h13 h 23

1st iteration

QUE h12 h13 h 23

2nd iteration

QUE h13 h 23

3rd iteration
R1

R2

QUE h 23

R3

Therefore the total computation


time is 3 2 1 t 6 t

[where t is the avg. Computation


time for finding a suitable dual home17
assignment for eliminating one handoff

Best Contributor Search


(BCS)
c2N 1

Or,

c2Ncomplexity
[ Time
]
i

i 1

Or,

(3)

c2N c2N 1

1 c2N 1

N 1
c
c2N 2
2

Or,

(4)

N 2 N N 2 N 2


The worst-case time complexity

4
N

N is
of
N 2
OBCS

(5)

for
18

Best Contributor Search


(BCS)
[Drawbacks ]
Fixed arrangement of QUE

Problem in future anticipation

19

Modified Best Contributor Search


(MBCS)

20

Performance Evaluation
[Test case generation ]

The handoff cost


between two
neighbour cells
follows the uniform
distribution[2]
Capacity of RNCs
are generated from
capacity of BSCs[2]

Figure.2. Network layout

Capacity of MSC
satisfied the cost
constrained [2]

21

Performance Evaluation
[Total cost variation]
ODH is an admissible heuristics [3]. Therefore a near optimal
solution can always be expected.
TS and ACO are standard meta heuristics method. Therefore
near optimal solution is always granted
BCS find the optimal arrangement by eliminating most of the
handoff cost
MBCS makes a detail analysis of how the future step will be
effected if it assigns a RNC to a MSC in dual homed arrangement
.

22

Performance Evaluation
[Total cost variation]

Figure.3. Variation of total cost


23

Performance Evaluation
[Average run time]

The time complexity of h(s) of a state in ODH is O(r*m) [2]. Therefore,


it is time consuming for large network Size

Time complexity ACO and TS also depend on the network size but
Should be much
faster then ODH.

The computation time of BCS and MBCS is primarily depend on the


c2N is always lesser than
number of handoff costs, which

The computation time of MBCS should be grater then the computation


time of BCS

24

Performance Evaluation
[Average run time]
Table 1: Average runs time in second
Network
size

MBCS

BCS

ACO

TS

ODH

595-6-3

0.223

0.107

0.085

0.082

0.127

595-8-3

0.262

0.095

0.095

0.09

0.317

595-10-3

0.404

0.11

0.108

0.095

21.36

595-12-3

0.397

0.133

0.12

0.108

16.274

595-14-3

0.405

0.097

0.119

0.117

25.319

595-6-5

0.288

0.105

0.109

0.101

0.257

595-8-5

0.342

0.095

0.118

0.113

7.537

595-10-5

0.403

0.109

0.128

0.125

194.338

25

Performance Evaluation

Figure.4. Average computation time as a function of MSCs

26

Performance Evaluation

Fig ure.5. Average computation times as a function of RNCs


27

Performance Evaluation
[Lower Bound (LB)]

For, calculating the LB we assumed the followings:


I.

There are no capacity constraints

II.

There are no cost constraints

Generally, at the global optima the total handoff cost


become ideally zero or near to zero [19].
Therefore, our LB is as
N M

cij xij

i 1 j 1

(6)

LB min
N N

Subject to
(7)

hii' yiii 0

i 1 '
i 1

28

Performance Evaluation
[Lower Bound (LB)]
Table 2: Performance of BCS as an Optimization algorithm
Network Size

LB

Solution obtained by BCS

% of deviation

595-6-3

205.6684

217.53

5.7

595-6-5

299.2

320.2

7.02

595-8-3

290.6042

345.3566

18.8

595-8-5

353.52

395.7304

11.5

595-10-3

350.1008

360.1008

595-10-5

459.274

552.7997

18

595-12-3

383.9007

398.0309

3.6

595-12-5

473.2198

489.2198

3.3

29

Performance Evaluation
[Lower Bound (LB)]
Table 3: Performance of MBCS as an Optimization algorithm
Network Size

LB

Solution obtained by MBCS

% of deviation

595-6-3

205.6684

210

2.1

595-6-5

299.2

308.8203

3.21

595-8-3

290.6042

295.67

1.7

595-8-5

353.52

357.52

1.1

595-10-3

350.1008

355.7290

1.6

595-10-5

459.274

466.8

1.6

595-12-3

383.9007

392.9007

2.34

595-12-5

473.2198

487.287

2.9

30

Performance Evaluation
[scalability]
How robust the optimal arrangement found by BCS and
MBCS?

th
percentage of scalability of iCSIC
MSC
is
CSCC

i
i
100
%ScalabilityMSC _ i
CSICi

(8)

31

Performance Evaluation
[scalability]
Table 4: Scalability; after optimizing the network by MBCS
algorithm. (number of MSCs is 5)

Network

MSC_1

MSC_2

MSC_3

MSC_4

MSC_5

595-6-5

47.0463

29.8178

100

20.6595

23.6134

595-8-5

49.9738

11.6279

14.9209

100

100

595-10-5

26.9173

77.2824

3.1002

24.322

42.7586

595-12-5

39.4032

31.0108

66.9431

50.1684

100

595-14-5

52.3927

67.8924

0.4303

44.6556

41.7004

32

Performance Evaluation
[scalability]

Table 5: Scalability; after optimizing the network by MBCS


algorithm. (number of MSCs is 3)
Network

MSC_1

MSC_2

MSC_3

595-6-3

100

77.0435

22.2133

595-8-3

46.004

52.8721

100

595-10-3

46.942

25.902

100

595-12-3

47.2222

43.1349

100

595-14-3

42.4833

67.8828

66.8019

33

Performance Evaluation
[scalability]

Table 6: Scalability; after optimizing the network by BCS algorithm.


(number of MSCs is 5)

Network

MSC_1

MSC_2

MSC_3

MSC_4

MSC_5

595-6-5

47.0463

29.8178

100

20.3659

23.6134

595-8-5

49.9738

17.9535

40.2174

100

41.3517

595-10-5

26.9179

50.2654

27.5835

24.322

42.7586

595-12-5

39.4032

31.0108

66.9431

50.1684

100

595-14-5

52.3927

67.8924

29.2599

20.4276

41.7006

34

Conclusions
Scope of the work
1. Address the dual homing CSA problem for UMTS network.
2. Developed novel heuristics method named as BCS and
MBCS
3. BCS gives the near optimal solution within a reasonable
time
4. The solution obtained by MBCS is best in terms of total
cost but slower then ACO,TS, BCS
5. BCS and MBCS , both can be applicable as solution
methodology for dual homing network arrangement in
UMTS network
Future scope of the work
1. Improvement of MBCS algorithm.
2. Would like to incorporate the random walk for
generating the handoff cost for capturing the real-life
scenario.
35
3. Address CSA problem for 4G network

References
1. M. M. Diallo, S. Pierre, and R. Beaubrun, A Tabu search approach for
assigning node Bs to switches in UMTS networks," IEEE. Trans.
Wireless. Comm, vol. 9, no. 4, pp. 1350-1359, 2010.
2. S. K. Sadhukhan, S. Mandal, S. R. Biswas, P. Bhaumik, and D.Saha,
Post-deployment Tuning of UMTS Cellular Networks through Dualhoming of RNCs, in Proc. of 1st International Conference on
Communication Systems and Networks and Workshops, Jan 2009, pp.
1-10.
3. A. Merchant and B. Sengupta, Assignment of cells to switches in PCS
networks," IEEE/ACM Trans. on Networking, vol. 3, no. 5, pp. 521-526,
1995.
4. P. S. Bhattacharjee, D. Saha, and A. Mukherjee, Heuristics for
assignment of cells to switches in a PCSN: A comparative study," in
Proc. of IEEE Int. Conf. Pers. Wireless Commun., Feb 1999, pp. 331
-334.
5. D. Saha, A. Mukherjee, and P. S. Bhattacharjee, "A simple heuristic for
assignment of cells to switches in a PCS network," Wirel. Pers.
Commun., vol. 12, no. 3, pp. 209 -224, 2000.
6. P. S. Bhattacharjee, D. Saha, A. Mukherjee, and M. Maitra, "Location
36
area planning for personal communication services networks,", in Proc.

References
7. A.Demirkol, C. Ersoy, M. U. Caglayan, and H. Delic, "Location area
planning in cellular networks using simulated annealing", in Proc.
IEEE INFOCOM, vol. 1, 2001, pp. 13 -20.
8. A. Quintero and S. Pierre, Evolutionary approach to optimize the
assignment of cells to switches in personal communications
networks," Comput. Commun., vol. 26, no. 9, pp. 927-938, 2003.
9. A. Quintero and S. Pierre, Assigning cells to switches in cellular
mobile networks: A comparative study," Comput. Commun., vol.
26, no. 9, pp.950 -960, 2003.
10.S. S. Sanza and X. Yao, Assignment of cells to switches in a
cellular mobile network using a hybrid Hopfield network-genetic
algorithm approach, Applied Soft Computing, vol. 8, no. 1, pp.
216-224, 2008.
11.S. Pierre and F. Houeto, Assigning cells to switches in cellular
mobile networks using taboo search," IEEE Trans. Syst., Man,
Cybern. B, Cybern., vol. 32, no. 3, pp. 351-356, 2002.
12.A. Quintero and S. Pierre, A memetic algorithm for assigning cells
to switches in cellular mobile networks, IEEE Commun. Lett., vol.
6, no. 11, pp. 484 -486, 2002.
13.R. Din and S. S. Tseng, A genetic algorithm for solving dual-

37

References
14. M. Dianti, S. Naik, X. Shen, and F. Karray, "A genetic algorithm
approach in cellular mobile networks", in Proc. Can. Workshop Inf.
Theory, 2003, pp.159 -162.
15. M. R. Akella, R. Batta , E. M. Delmelle, P. A. Rogerson, A. Blatt, and G.
Wilson, "Base station location and channel allocation in a cellular
network with emergency coverage requirements," Eur. J. Oper. Res.,
vol. 164, no. 2, pp. 301 -323, 2005.
16.C.Y.Lee, S. J. Kim and T.Park A design of multi-layered location
registration areas in microcellular systems", Telecommun. Syst., vol.
14, no. 14, pp.107 -120 2000.
17.S. J. Shyu, B. M. T. Lin, and T.-S. Hsiao, Ant colony optimization for
the cell assignment problem in PCS networks," Comput. Oper. Res.,
vol. 33, no. 6, pp.1713 -1723, 2006.
18.S. Menon and R. Gupta, "Assigning cells to switches in cellular
networks by incorporating a pricing mechanism into simulated
annealing", IEEE Trans. Syst., Man, Cybern. B, vol. 34, no. 1, pp.558
-565, 2004.
19.M. Maitra, A. Mukherjee, and D. Saha, "Integrated interswitch cable
and handoff cost minimizing heuristics for cell-to-switch assignment38
in wireless cellular networks," IEEE Transactions on Vehicular

References
22. U. Turke and M. Koonert, Advanced site configuration techniques for
automatic UMTS radio network design," in Proc. IEEE 61st Vehicular
Technology Conference 2005 (VTC 2005-Spring), vol. 3, 2005,
pp.1960 -1964.
23. C. Charnsripinyo and D. Tipper, Topological design of 3G wireless
backhaul networks for service assurance," in Proc. 5th Int. Workshop
on Design of Reliable Communication Networks 2005 (DRCN 2005),
2005, pp.115 -123.
24. M. St-Hilaire, S. Chamberland, and S. Pierre, "Uplink UMTS network
design-an integrated approach," Computer Networks, vol. 50,
pp.148 -151, 2006.
25.M. St-Hilaire , S. Chamberland, and S. Pierre, "A tabu search
algorithm for the global planning problem of third generation mobile
network," Computers and Electrical Engin., vol. 34, pp.470 -487,
2008.
26.S.K.Sadhukhan, S.Mandal, and D.Saha, "Dual-homing of RNCs in
UMTS Networks", IIM Calcutta, working paper series WSP No. 701,
2012.
39

40

Introduction
[Type of Cell to Switch Assignment]

41

Modified Best Contributor Search


(MBCS)
QUE for BCS
QUE h14 h15 h 24 h 25 h35

Iteration-1
QUE h14 h 24 h15 h 25 h35

Iteration-2
QUE h15 h 25 h14 h35 h 24

Iteration-3
QUE h 24 h 25 h14 h35 h15

Iteration-6
QUE h14 h15 h34 h35 h 25

42

UMTS Architecture

43

Types of Handoff in UMTS


Network
M1

S1

R1

M2

R2

S2

R3

B1
44

B2

B3

B4

CSA UMTS: Problem


Formulation
[Back up]
M1

R2

R4
h 24

M3

M2

R5
h14
h 23

R1
h15

R3

R6

h35

h34
h 26

45

CSA UMTS: Problem


Formulation
y21 y12 1
y23 y32 1
y26 y62 1

[Back up]
M1

M3

M2

Total cos t Inital cos t

i 1 i' 1,i' i

R2

R4
h 24

R5
h14

R1
h15

R3

h ii' yii' C23

R6

h35

h34
h 23 0

h 26 0

46

Best Assignment
Step.1- Best _ Assignment and AV = 0
M1

M
2

Step.2Step.3Step.4-

R3

R2

R1

R4
h14

h24

R5 Step.5h15

h25
h35
Figure.1. Network layout

h QUE 1 h14

Cost R1 M2 2 h14 h15 C12

Tmp1 Cost R1 M2

EH1 h15

and
Cost R4 M1 2 h14 h24 C41

Tmp2 Cost R4 M1

Step.6-

and

EH 2 h 24

h 24
Step.7- If R1 is assigned to M2 then
another
assignment will need to
eliminate
TCost1 Cost R1 M2 Cost R2 M2

47

Best Assignment
Step.8- Similarly R4
for M1
M1

M
2

TCost 2 Cost R4 M1 Cost R5 M1

Step.9-TCost max TCost1, TCost 2

R3

R2

R1

R4
h14

h24

h25

h15

Step.10- if TCost > AV


{
AV= TCost;
R5
Best_Assignment =
Assignment
related with TCost
}

h35
Figure.1. Network layout

Step.11- Repeat from step.2 for rest of the


element of QUE
Step.12- Return the Best_Assignment .
48

Fixed arrangement of QUE


Initial arrangement
h
QUE h15
h15
h 24
h35
14
25
35
25

Optimal solutio

Solution obtained
by BCS

49

Problem in future
anticipation
M

M1

R3

R2

R1

R4
h14

h24

R5
h15

h25
h35
Figure.1. Network layout

50

Modified Best Contributor Search


(MBCS)
M1

M
2

Cost R1 M 2 Cost R2 M 1 ,

h14

max

R3

Cost R1 M 2 Cost R2 M 2

Cost R5 M 1

R5 M 1
Cost
M 2 Cost

R2 R1R1
R4
R5

h14

h24

h25
h35

h15

R1 M2

R4 M1

Left _ Child h15

Right _ Child h 24

h24
R2 M2

h15
R4 M1

Left _ Child h 25

R1 M2

R5 M1

Right _ Child h 25 h35

h35
R5 M1

51

Performance Evaluation
Table.4.Percentage of cost reduction by different algorithms
Network size

ODH

TS

ACO

BCS

595-6-3

46.31269

61.2863

61.2863

61.2863

595-8-3

49.75319

49.75319

49.17388

49.75319

595-10-3

36.72611

52.44604

52.44604

52.44604

595-12-3

13.9369

34.52987

38.68264

36.42574

595-14-3

30.29189

45.27683

43.04686

45.25332

595-6-5

23.35128

51.30867

49.587

51.30867

595-8-5

37.63484

45.9327

49.92972

47.31012

MBCS
61.2863
50.49
52.44604
38.24

45.25332
53.3086
52.06

52

Das könnte Ihnen auch gefallen