Optimal Policy of Fuzzy Multiple Vacation Queueing System

International Journal of Computer Application
Available online on http://www.rspublication.com/ijca/ijca_index.htm
OPTIMAL POLICY OF FUZZY
M (Xm, N ) / Gsos / 1
Issue 4, Volume 1 (February 2014)

ISSN: 2250-1797
MULTIPLE VACATION
QUEUEING SYSTEM
Mrs.K.Julia Rose Mary
Ms.C.Shanmugapriya
Assistant professor,
PG student,
Department of Mathematics,
Department of Mathematics,
Nirmala College for women,
Nirmala College for women,
CBE-18.
CBE-18.
____________________________________________________________________
ABSTRACT
In this paper we investigate the optimal operating policy of fuzzy M (Xm, N ) / GSOS / 1
multiple
vacation queueing
model
in
fuzzy
environment.
Non- linear programming
technique(NLP) is used to construct the membership function of the system characteristics, by

considering the various cost as fuzzy numbers. Further with the help of Zadeh extension
principle and -cuts we find the upper and lower bounds of the system characteristics, which
help in deriving the optimum cost for FM (Xm, N ) / GSOS / 1 queueing model.
__________________________________________________________________
INTRODUCTION
The first study of queueing system with second optional service (SOS) was done by
Madan(8). He cited important applications of the model in many rea l life situations. Later
Medhi(10) and Choudhury and Madan(2) generalized the results of Madan by deriving the steady
state queue size distribution at the stationary point of time for general SOS service time.
Madan et al.,(9) considered an extended
M X / G / 1 queue with two types of general
heterogeneous service and modified Bernoulli schedule vacation.

Among the various paradigmatic changes in Science and Mathematics one such change
concerns the concept of uncertainty. Based on uncertainty, Zadeh introduced the concept of fuzzy
sets. Fuzzy sets and fuzzy logic is a true magnum opus and they are remarkable achievements.
R S. Publication (rspublication.com), rspublicationhouse@gmail.com
Page 199


ISSN: 2250-1797
Fuzzy logic has rapidly become one of the more successful of todays technologies for
developing sophisticated control systems. Further, fuzzy queues are potentially much more useful
and realistic than the commonly used crisp queues. Li and Lee(7) investigated analytical results
for two typical fuzzy queues, M / F / 1/ , FM / FM / 1/ where F represents fuzzy time and
FM represents fuzzy field exponential distribution. Negi and Lee(11) proposed a procedure using
-cuts and two variable simulation to analyse fuzzy queue. Using parametric programming
Kao et al.,(6) constructed the membership function of the system characteristics for fuzzy queues
and successfully applied them to four simple fuzzy queues namely M / F / 1/ , F / M / 1/ ,
F / F / 1/ , FM / FM / 1/ .
Chen(1) had presented a model using the fuzzy membership function and considered the
fuzzy arrival rate and fuzzy cost regarding the fuzzy parametric planning method.
R.Kalayanaraman et al.,(5) discussed a single server fuzzy queue with unreliable server using the
fuzzy concepts and extension principle. Jeeva and Rathnakumari(3) analyzed the bulk arrival
single server Bernoulli feedback queue with fuzzy vacations and fuzzy parameters. Recently
W.Ritha and Lilly Robert(12) investigated profit analysis of fuzzy M / E K / 1 queuing system
using nonlinear zero one programming approach and they provided the technique to find the
optimal solutions at different possibility levels of .
The availability of numerous literature reveals that, in the field of queuing theory, fuzzy
logic plays an important role. Thus in this paper we apply the techniques of -cuts and Zadehs
extension principle, to transform a family of conventional queue into fuzzy queue. Then, with
X
the aid of membership functions the optimum cost for the FM ( m, N ) / GSOS / 1 / MV queuing model
is analyzed.
MODEL DESCRIPTION
M (Xm, N ) / Gsos / 1 / MV queueing model deals with Non-Morkovian single server batch arrival
queueing system in which the server provides two types o f heterogeneous service facility. It is
assumed that all the arriving customers require the First Essential Service(FES) and only some of
them demand the Second Optional Service(SOS) with probability r. Each cycle is made up of
Page 200


ISSN: 2250-1797
buildup period, setup period, dormant period, and busy period. The vacation time, setup time and
service time of both phases are assumed to be independent of each other and follow general
distributions with finite moments. The vacation policy is multiple working policy. According to
this policy, the server leaves the system for vacation of random length V1 as soon as the system
empties. After returning from the vacation if the server finds bounded or more customers in the
system, he immediately starts a setup operation of random length D. Otherwise, he takes another
vacation V2 and so on until he finds finally atleast m-customers. The random variables
V1 ,V2 ,......... . are assumed to be independent and identically distributed with generic
representation V .
M X / Gsos / 1 / SV queueing model is proposed by Julia and Afthab Begum(4). They derived
optimal policy also. Based on that the optimal operating cost for M (Xm, N ) / Gsos / 1 / MV is given by
TC (m, N )
Cy
Ch LRsos( m, N ) Cset Pset Cdor Pdor Cbuild Pbuild Cbusy Pbusy Cv pv (1)
E(cycle)
where the costs C y , Ch , Cset , Cdor , Cbuild , Cbusy andCv denote cycle cost, holding cost, setup cost,
standby
cost,
buildup
cost,
operating
cost
and
reward
cost
respectively.
Also
Pv , Pset , Pdor , Pbusy andLRsos( m, N ) denote the probability that the server is in vacation state, setup state,
dormant state, busy state and the expected system size of the model respectively.
Now by extending the M X / Gsos / 1 queue system to fuzzy environment, the fuzzy set are
defined for cycle cost C y , holding cost Ch , setup cost C s , standby cost Cd ,buildup cost C B ,
operating cost C Bu and reward cost Cv .These fuzzy sets are represented by C y , C h , C s , C d ,
C B , C Bu , and C v . With the help of these costs, we determine the optimum cost for fuzzy
M (Xm, N ) / Gsos / 1 multiple vacations queueing system.
Page 201


ISSN: 2250-1797
RESULTS
The membership function of the above fuzzy sets are denoted by c y x , ch ( y) , cs (z ) ,
c (s) , c (t ) , c (u) and c (v) . Then we have,

d
Bu
c y = { x, c y (x) ; x X } --------------(2)
ch = {y, ch ( y) ; y Y } -----------------(3)
cs = {z, cs (z ) ; z Z } ---------------(4)
cd = {s, cd (s) ; s S }-------------------(5)
c B = {t, cB (t ) ; t T } ----------------(6)
c Bu = {u, cBu (u) ; u U }---------------(7)
cv = { v , cv (v) ; v V } -------------(8)
where X, Y, Z, S, T, U and V are crisp universal sets of cycle cost, holding cost, setup cost,
standby cost, buildup cost, operating cost and reward cost.
Let f(x,y,z,s,t,u,v) denote the system characteristics of interest. Since c y , ch , cs , cd , c B , c Bu
and cv are fuzzy numbers, f( c y , ch , cs , cd , c B , c Bu , cv ) is also a fuzzy number. By following
Zadehs extension principle
the membership
function of the system characteristics
f( c y , ch , cs , cd , c B , c Bu , cv ) is defined as
f (c
y ,ch ,cs ,cd ,cB ,cBu ,cv )
( z)
sup min
xX , yY , zZ , sS ,tT ,uU ,vV
{ c y ( x), ch ( x), cs ( x), cd ( x), cB ( x), cBu ( x), cv ( x)}

(9)
where z= f(x,y,z,s,t,u,v) and thus Tc (m, N ) of equation (1) becomes,
z=f(x,y,z,s,t,u,v) =
x
y{LRsos( m, N ) } z{ pset } s{ pdor } t{ pbuild } u{ pbusy } v{ pv } ----(10)
E (cycle)
Then the membership function for the optimum cost yields,
T ( m , N ) ( z )
c
sup min
{ c y ( x), ch ( y), cs ( z), cd (s), cB (t ), cBu (u), cv (v)} (11)
Page 202


ISSN: 2250-1797
Unfortunately, the membership function is not expressed in the usual form, there by making
very difficult to imagine its shape. Thus in this paper we approach the problem by using a
mathematical programming technique, Parametric Non Linear Programming(PNLP). Hence
PNLP techniques are developed to find the - cuts of
f( c y , ch , cs , cd , c B , c Bu , cv ) based on the
extension principle.
PARAMETRIC NON LINEAR PROGRAMMING

To construct the membership function T ( m, N ) ( z) we have to derive the - cuts of
c
Tc (m, N ) . Thus, - cuts of c y , ch , cs , cd , c B , c Bu
and cv which are represented by crisp
intervals as,
c y ( ) [ xL , xU ] [min{x / cy ( x) }, max{x / cy ( x) }] (12a)

ch ( ) [ yL , yU ] [min{ y / ch ( y) }, max{ y / ch ( y) }] (12b)
cs ( ) [ zL , zU ] [min{z / cs ( z ) }, max{z / cs ( z ) }] (12c)
cd ( ) [sL , sU ] [min{s / cd ( s) }, max{s / cd ( s) }] (12d )
c B ( ) [tL , tU ] [min{t / cB (t ) }, max{t / cB (t ) }] (12e)
c Bu ( ) [uL , uU ] [min{u / cBu (u ) }, max{u / cBu (u) }] (12 f )
cv ( ) [vL , vU ] [min{v / cv (v) }, max{v / cv (v) }] (12g )
As a result the bound of these intervals can be described as functions of and can be
obtained as
xL min cy1 ( )
zL = min cs1 ( )
xU = max cy1 ( ) ;
zU = max cs1 ( ) ;
yL = min ch1 ( )
sL = min cd1 ( )
yU = max ch1 ( )
sU = max cd1 ( )
Page 203

tL = min cB1 ( )
tU = max cB1 ( ) ;
vL = min cv1 ( )
vU = max cv1 ( )
uL = min cBu1 ( )

ISSN: 2250-1797
uU = max cBu1 ( )
Thus by making use of the - cuts for total optimum cost (TOC), we construct the
membership function of equation (11) which is parametrized by . Now to derive the
membership function of TOC we have to find the left and right shape functions of T ( m, N ) ( z) .
c
By using Zadehs extension principle T ( m, N ) ( z) is the minimum of c y (x) , ch ( y) , cs (z ) ,

c
c (s) , c (t ) , c (u) and c (v) . Hence, to derive the membership function T ( m, N ) ( z) , we

c
d
Bu
need atleast one of the following cases to hold such that T ( m, N ) ( z) = .

c
Case (1): c y (x) , ch ( y) , cs (z) , cd (s) , cB (t ) , cBu (u) , cv (v)

Case (3): c y (x) , ch ( y) , cs (z ) , cd (s) , cB (t ) , cBu (u) , cv (v)
This can be accomplished by using parametric nonlinear programming technique(PNLP).
Thus, by using PNLP the lower and upper bounds of the - cuts of case(1) is given by
x
[Tc (m, N )]L1 min z
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild } u{Pbusy } v{Pv } (13a)
E (cycle)
x
[Tc (m, N )]U 1 max z
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild } u{Pbusy } v{Pv } (13b)
E (cycle)
Page 204


ISSN: 2250-1797
For case (2) as
x
[Tc (m, N )]L2 min z
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild } u{Pbusy } v{Pv } (13c)
E (cycle)
x
[Tc (m, N )]U 2 max z
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild } u{Pbusy } v{Pv } (13d )
E (cycle)
Similarly we can formulate the other cases also.

Thus by the definition of c y ( ) , ch ( ) , cs ( ) , cd ( ) , c B ( ) , cBu ( ) , cv ( ) we find that
they can be replaced by x [ xL , xU ] , y [ yL , yU ] , z [ zL , zU ] , s [sL , sU ] , t [tL , tU ] ,
u [uL , uU ], v [vL , vU ] . Then the -cuts form a nested structure which are given by
[ xL1 , xU1 ] [ xL2 , xU2 ] ;
[ yL1 , yU1 ] [ yL2 , yU2 ] ;
[ zL1 , zU1 ] [ zL2 , zU2 ] ;
[tL1 , tU1 ] [tL2 , tU2 ] ; [uL1 , uU1 ] [uL2 , uU2 ] ; [vL1 , vU1 ] [vL2 , vU2 ]
[ sL1 , sU1 ] [sL2 , sU2 ]
where
0 1 2 1
Hence we find that the feasible regions defined by 2 are larger than those defined by 1 .
Thus [Tc (m, N )]L1 [Tc (m, N )]L2 and [Tc (m, N )]U1 [Tc (m, N )]U 2 . Thus lower bounds have the
unique smallest element and upper bounds have the unique largest element. Then we have,
[Tc (m, N )]L
[Tc (m, N )]U
min
x
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild }
E (cycle)
u{Pbusy } v{Pv } (14a)
max
x
y{LRsos( m, N ) } z{Pset } s{Pdor } t{Pbuild }
E (cycle)
u{Pbusy } v{Pv } (14b)
The crisp interval [ [Tc (m, N )]L , [Tc (m, N )]U ] obtained from (14a) and (14b) represent the
-cuts of Tc (m, N ) , and further we have [Tc (m, N )]L [Tc (m, N )]U . In otherwords, [Tc (m, N )]L
increases and [Tc (m, N )]U decreases. If both [Tc (m, N )]L and [Tc (m, N )]U in (14a) and (14b) are
invertible with respect to , then a left shape function L( z) [(Tc (m, N )L ]1 and a right shape
function R( z) [(Tc (m, N )U ]1 can be derived and hence the membership function T ( m, N ) ( z) is
c
given by
Page 205


ISSN: 2250-1797
L( z ) [Tc (m, N )]L 0 z [Tc (m, N )]L 1
Tc ( m, N ) ( z ) 1
[Tc (m, N )]L 1 z [Tc (m, N )]U 1
U
U
R( z ) [Tc (m, N )] 1 z [Tc (m, N )] 0
where [Tc (m, N )]L and [Tc (m, N )]U can also be solved analytically.
CONCLUSION
Many of the queuing systems have been analyzed by using the fuzzy set theory which
provides wider application in many fields. Thus in this paper by applying the techniques of -cut
and Zadehs extension principle, we derive the membership function for total optimum cost
(TOC). Moreover we find that, the fuzzy cost based queueing systems are much more useful and
realistic than the conventional crisp queueing models.
REFERENCES
1). Chen.S.P.(2005), Parametric nonlinear programming approach to fuzzy queues with bulk
service, Eropean Journal of Operations Research,volume.163,pp:434-444
2) Choudhury.G and Madan K.C.(2005),A two stage arrival queueing system with a modified
Bernoulli schedule vacation under N-policy,Math comput model,42,pp:71-85
3) Jeeva.M and Rathnakumar.E.(2012), Bulk arrival single server, Bernoulli feeback queue with
fuzzy vacations and fuzzy parameters,ARPN,volume.2 No.5,pp:492-499
4) K.Julia Rose, M.I Afthab Begum and Jemila Parveen.(2011),Bi- level threshold policy of
M X / G1G2 / 1
queue
with
early
setup
and
single
vacation
,Int.J.Operational
Research(IJOR).volume.10 No.4,pp:469-493
5) Kalayanaraman.R, Thilligovidan.N and Kannadasan.G.(2010), Single server fuzzy queue
with unreliable server, IJCC volume.8 No.1,pp:1-4
6) Kao.C, Li.C and Chen.S.(1999), Parametric programming to the analysis of fuzzy queues,
fuzzy sets and system,107,pp:93-100
Page 206


ISSN: 2250-1797
7) Li.R.J and Lee.I.S.(1989), Analysis of fuzzy queues, computers and mathematics with
applications,volume.17(7):pp:1143-1147
8) MadanK.C.(2000), An
M / G / 1 queue with second optional service, Queueing
systems,34,pp:37-46
9) Madan.K.C, Al-Rawi.Z.R and Al-Nasser.A.D.(2005), on M X / G1G2 / G( BS ) / Vs , vacation
queue with two types of general heterogeneous services, Journal of applied mathematics and
decision sciences,3,pp:123-135
10) Medhi.J.(2002), A single server poisson arrival input queue with a second optional
channel, Queueing systems,42,pp:239-242
11) Negi.D.S and Lee.I.S.(1992), Analysis and simulation of fuzzy queues, fuzzy sets and
systems,46,pp:321-330
12) Ritha.W and Lilly Robert.(2012), Profit analysis of fuzzy M / E K / 1 queueing system nonlinear zero one programming approach,IJAIR volume.1 issue 2, pp:1-14
Page 207

Optimal Policy of Fuzzy Multiple Vacation Queueing System

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Optimal Policy of Fuzzy Multiple Vacation Queueing System

Enviado por

Direitos autorais:

Formatos disponíveis

International Journal of Computer Application

Available online on http://www.rspublication.com/ijca/ijca_index.htm