IOSR Journal of Engineering (IOSRJEN)

ISSN (e): 2250-3021, ISSN (p): 2278-8719

Vol. 05, Issue 04 (April. 2015), ||V2|| PP 01-12

www.iosrjen.org

BMAP/BMSP/1 Queue with Randomly Varying Environment

Sandhya.R1, Sundar.V2, Rama.G3, Ramshankar.R4, Ramanarayanan.R5
1

Independent Researcher MSPM, School of Business, George Washington University, Washington .D.C, USA
2
Senior Testing Engineer, ANSYS Inc., 2600, Drive, Canonsburg, PA 15317,USA
3
Independent Researcher B. Tech, Vellore Institute of Technology, Vellore, India
4
Independent Researcher MS9EC0, University of Massachusetts, Amherst, MA, USA
5
Professor of Mathematics, (Retired), Vel Tech University, Chennai, INDIA.

Abstract: This paper studies two stochastic batch Markovian arrival and batch Markovian service single server
queue BMAP/BMSP/1 queue Models (A) and (B) with randomly varying k* distinct environments. The arrival
process of the queue has matrix representation {Dim : 0 m M} of order k i describing the BMAP and the
service process has matrix representation {Sni : 0 n N} of order ki describing the BMSP respectively in the
environment i for 1 i k*. Whenever the environment changes from i to j the arrival BMAP and service
BMSP change from the i version to the j version with the exception of the first remaining arrival time and first
remaining service time in the new environment start as per stationary probability vector of j version of the
BMAP and of j version of the BMSP respectively for 1 i, j k*. The queue system has infinite storing
capacity and the state space is identified as five dimensional one to apply Neuts matrix methods. In the
environment i, the sizes of the arrivals and services are governed by the matrices Dim and Sni , with respect to
environment. The service process is stopped when the queue becomes empty and is started with initial
probability vector of corresponding environment BMSP when the arrival occurs. Matrix partitioning method is
used to study the models. In Model (A) the maximum of the arrival sizes is greater than the maximum of the
service sizes and the infinitesimal generator is partitioned mostly as blocks of the sum of the products of BMAP
arrival and BMSP service phases in the various environments times the maximum of the arrival sizes for
analysis. In Model (B) the maximum of the arrival sizes is less than the maximum of the service sizes. The
generator is partitioned mostly using blocks of the same sum-product of batch Markovian phases times the
maximum of the service sizes. Block circulant matrix structure is noticed in the basic system generators. The
stationary queue length probabilities, its expected values, its variances and probabilities of empty levels are
derived for the two models using matrix methods. Numerical examples are presented for illustration.
Keywords: Batch Arrivals, Batch Services, Block Circulant Matrix, Neuts Matrix Methods, Phase Type
Distribution.

I.

INTRODUCTION

In this paper two batch arrival and batch service BMAP/BMSP/1 queues with random environment
have been studied using matrix geometric methods. Numerical studies on matrix methods are presented by Bini,
Latouche and Meini [1]. Multi server model has been of interest in Chakravarthy and Neuts [2]. Birth and death
model has been analyzed by Gaver, Jacobs and Latouche [3]. Analytic methods are focused in Latouche and
Ramaswami [4] and for matrix geometric methods one may refer Neuts [5]. For M/M/1 bulk queues with
random environment models one may refer Rama Ganesan, Ramshankar and Ramanarayanan [6] and M/M/C
bulk queues with random environment models are of interest in Sandhya, Sundar, Rama, Ramshankar and
Ramanarayanan [7]. PH/PH/1 bulk queues without variation of environments have been treated by Ramshankar,
Rama Ganesan and Ramanarayanan [8] and with variations of environment have been analyzed by
Ramshankar,Rama, Sandhya, Sundar and Ramanarayanan [9]. BMAP/M/C queue with bulk service and random
environment has been studied by Rama, Ramshankar, Sandhya, Sundar and Ramanarayanan [10]. The models
considered here are general compared to existing models. Batch Markovian service process (BMSP) considered
in this paper is similar to batch Markovian arrival process (BMAP) and BMAP has been studied by
Lucantony[11] and has been analyzed further by Cordeiro and Kharoufch [12] . When the queue becomes empty
the service process is stopped and it is started with starting probability vector when the arrivals occur in the
empty queue. Usually bulk arrival models have M/G/1 upper-Heisenberg block matrix structure. The
decomposition of a Toeplitz sub matrix of the infinitesimal generator is required to find the stationary
probability vector and matrix geometric structures are rarely noted. In such analysis the recurrence relation
method to find the stationary probabilities is stopped at a certain level in most general cases using a terminating
method very well explained by Qi-Ming He [13] and this stopping limitation of terminating method converts an
infinite arrival system to a finite arrival one. In special cases generating function has been identified by Rama

International organization of Scientific Research

1|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

and Ramanarayanan [14]. However the method of partitioning of the infinitesimal generator along with
environment, BMAP phases and BMSP phases used in this paper is presenting matrix geometric solution for
finite sized batch arrivals and batch services models. The M/PH/1 and PH/M/C queues with random
environments have been studied by Usha [15] and [16] without bulk arrivals and bulk services. It has been
noticed by Usha [15, 16] that when the environment changes the remaining arrival and service times are to be
completed in the new environment. The residual arrival time and the residual service time distributions in the
new environment are to be considered in the new environment at an arbitrary epoch since the spent arrival time
and the spent service time have been in the previous environment with distinct sizes of PH phase. Further new
arrival time and new service time from the start using initial PH distributions of the new environment cannot be
considered since the arrival and the service have been partly completed in the previous environment indicating
the stationary versions of the arrival and service distributions in the new environments are to be used for the
completions of the residual arrival and service times in the new environment and on completion of the same the
next arrival and service onwards they have initial versions of the PH distributions of the new environment. The
stationary version of the distribution for residual time has been well explained in Qi-Ming He [13] where it is
named as equilibrium PH distribution. Randomly varying environment BMAP/BMSP/1 queue models have not
been treated so far at any depth.
Two models (A) and (B) on BMAP/BMSP/1 bulk queue systems with infinite storage space for
customers are studied here using the block partitioning method. Model (A) presents the case when M, the
maximum of the arrival sizes of BMAP is bigger than N, the maximum of the service sizes of BMSP. In Model
(B), its dual case N is bigger than M, is treated. In general in Queue models, the state space of the system has the
first co-ordinate indicating the number of customers in the system but here the customers in the system are
grouped and considered as members of blocks of sizes of the maximum for finding the rate matrix. Using the
maximum of the batch arrival sizes and of the batch service sizes and grouping the customers as members of
blocks in addition to coordinates of the arrival and service phases for partitioning the infinitesimal generator is a
new approach in this area. The matrices appearing as the basic system generators in these two models due to
block partitioned structure are seen as block circulant matrices. The paper is organized in the following manner.
In sections II and III the stationary probability of the number of customers waiting for service, the expectation
and the variance and the probability of empty queue are derived for these Models (A) and (B). In section IV
numerical cases are presented to illustrate them.

II.

MODEL (A): MAXIMUM ARRIVAL SIZE M > MAXIMUM SERVICE SIZE N

2.1Assumptions
(i)There are k* environments. The environment changes as per changes in a continuous time Markov chain with
infinitesimal generator 1 of order k* with stationary probability vector .
(ii)In the environment i for 1 i k*, the batch arrivals occur in accordance with Batch Markovian Arrival

Process with matrix representation for the rates of batch sizes m given by the finite sequence {
, 0 m }

with phase order where 0 has negative diagonal elements and its other elements are non-negative; has
non-negative elements for 1 m where is the maximum batch arrival size in the environment i. Let

= =0
and be the stationary probability vector of the generator matrix with = 0 and e = 1.
(iii) In the environment i for 1 i k*, when the queue length L is more than or equal to the maximum batch
service size , (L ) of the environment, the batch services occur in accordance with Batch Markovian
Service Process with matrix representation for the rates of batch sizes n given by the finite sequence { , 0 n
} with phase order where 0 has negative diagonal elements and its other elements are non-negative;

has non-negative elements for 1 n . Let = =0

and be the stationary probability vector of the
generator matrix with = 0 and e = 1.
(iv)When n customers n < are waiting for service, then n customers are served with rate for 1 n n-1

and n customers are served with rate =

= ,
which is a matrix of order .
(v)The BMSP service process is stopped when the queue becomes empty and is started in the environment i for
1 i k*with initial probability vector of the i version BMSP when the arrival occurs.
(v) When the environment changes from i to j for 1 i, j k*, the arrival process and service process in the new
environment j start as per stationary (equilibrium) probability vectors of and of respectively of the j

International organization of Scientific Research

2|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

versions of arrival process BMAP, namely, { , 0 m } and of the j version of the service process,
namely, BMSP { , 0 n } in the new environment.
(vii) The maximum batch arrival size of all BMAPs, M= ma1 is greater than the maximum batch
service size N= ma1 .
2.2.Analysis
The state of the system of the continuous time Markov chain X (t) under consideration is presented as follows.
X(t) = {(0, i, j) : for 1 i k* and 1 j )} U {(0, k, i, j, j) ; for 1 k M-1; 1 i k*; 1 j ; 1 j
} U {(n, k, i, j, j): for 0 k M-1; 1 i k*; 1 j ; 1 j and n 1}. (1)
The chain is in the state (0, i, j) when the number of customers in the queue is 0, the environment state is i for 1
i k* and the arrival BMAP phase is j for 1 j . The chain is in the state (0, k, i, j, j) when the number of
customers is k for 1 k M-1, the environment state is i for 1 i k*, the arrival BMAP phase is j for 1 j
and the service BMSP phase is j for 1 j . The chain is in the state (n, k, i, j, j) when the number of
customers in the queue is n M + k, for 0 k M-1 and 1 n < , the environment state is i for 1 i k*, the
arrival BMAP phase is j for 1 j and the service BMSP phase is j for 1 j . When the number of
customers waiting in the system is r, then r is identified with (n, k) where r on division by M gives n as the
quotient and k as the remainder. The chain X (t) describing model has the infinitesimal generator of infinite
order which can be presented in block partitioned form given below.
1 0 0
0
.
.
.
2 1 0 0
.
.
.
0 2 1 0 0
.
.
=
(2)
0
0 2 1 0 0 .
0
0
0 2 1 0 0

In (2) the states of the matrices are listed lexicographically as 0, 1, 2, 3, . For partition purpose the zero states

in the first two sets given in (1) are combined. The vector 0 is of type 1 x [
=1 + ( 1) =1 ] and
0=((0,1,1),(0,1,2),(0,1,3)(0,1, 1 ),(0,2,1),(0,2,2),(0,2,3)(0,2, 2 ),(0,k*,1),(0,k*,2),(0,k*,3)(0,k*, ),
(0,1,1,1,1),(0,1,1,1,2)(0,1,1,1, 1 ),(0,1,1,2,1),(0,1,1,2,2)(0,1,1,2,1 ),(0,1,1,3,1)...(0,1,1,3,1 )(0,1,1,1 ,1)
(0,1,1,1 , 1 ),(0,1,2,1,1),(0,1,2,1,2)(0,1,2,1, 2 ),(0,1,2,2,1),(0,1,2,2,2)(0,1,2,2,2 ),(0,1,2,3,1)...(0,1,2,3,
2 )(0,1,2,2 ,1)(0,1,2,2 , 2 ),(0,1,3,1,1)(0,1,3,3 , 3 )(0,1,k*,1,1),,(0,1,k*, , ),(0,2,1,1,1),(0,2
,1,1,2)(0,2,k*, , ),(0,3,1,1,1)(0,3,k*, ),(0,4,1,1,1)(0,4,k*, )(0,M-1,1,1,1)
(0,M-1,k*, , )).
.
(3)
The vector is of type 1x[
=1 ] and is given in a similar manner as follows
=(n,0,1,1,1),(n,0,1,1,2)(n,0,1,1, 1 ),(n,0,1,2,1)(n,0,1,2,1 )(n,0,k*,1,1)(n,0,k*, , ),(n,1,1,1,1)
(n,1,k*, , ),(n,2,1,1,1)...(n,2,k*, , )(n,M-1,1,1,1),(n,M-1,1,1,2)(n,M-1,k*, , )).
(4)

The matrices1 1 have negative diagonal elements, they are of orders [

+
(

1)

]
and
=1
=1
[
respectively
and
their
off
diagonal
are
non-negative.
=1 ]
The matrices 0 2 have nonnegative elements and are of order [
=1 ] . The matrices 0 2

have non-negative elements and are of types [ =1

+ ( 1)
and
=1 ] x [ =1 ]

[ =1 ] x [ =1 + ( 1) =1 ]. Component matrices of for i=0, 1, 2 are defined

below. Let denote the Kronecker sum and Kronecker product.
Let =0 0 + (1 ), = (0 ) + ( 0 ) + (1 ), for 1 i k*
(5)
where I indicates the identity matrices of orders given in the suffixes, is of order and the last term is a
diagonal matrix of order . Considering the stationary probability starting vectors of BMAP and of BMSP
for the change of environment the following matrix of order
=1 is defined which is concerned with
change of environment during arrival and service time.
1
1,2 1,3 1,
2,1
2
2,3 2,
3 3,
= 3,1 3,2
(6)

,1 ,2 ,3
where , is a rectangular matrix of type x and all its rows are equal to (1 ), ( ) for i j ,
1 i, j k*.The matrix of arrival rates of n customers corresponding to the arrival in BMAP in the environment
i is which is a matrix of order with non-negative elements for 1 n and = 0 matrix for n > for
1 i k*.
(7)

International organization of Scientific Research

3|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

The matrix of service rates of n customers corresponding to the service in BMSP in the environment i is
which is of order with non-negative elements for 1 n and = 0 matrix for n > for 1 i k*. (8)
1 1
0
0

0
2
0
2
0

0
Let =
for 1 n M
(9)
0
0
3 3
0

0
0
0

In (9) is a square matrix of order

=1 ; is a square matrix of order for 1 j k* and
=0 matrix for n > .The (i, j) component 0 appearing in (9) is a block zero matrix of type x .
1 1
0
0

0
2
0
2
0

0
3
Let =
for 1 n N
(10)
0
0
3
0

0
0
0

In (10) is a square matrix of order

=1 ; is a square matrix of order for 1 j k* and 0
appearing as (i, j) component of (10) is a block zero rectangular matrix of type x .The matrix for i
= 0,1,2 are as follows.
0 =

0
0
0
2 =
0 0 1 2
1
1

0
0
0
0 0 0

3
2
2 1
0
0
0

3 2
0
0
0
(11)
0 0 0
0
1

(12)
0 0 0
0
0

3
4

0
0
0 0 0
0
0
0
2
3
1

1
2
2 1
0 0 0
0
0
0

1
2
2 1

2
1
1

1
3 2 1
3
2
2
1

4 3 2
4
3

1 2

1
2
2 1
1 =
(13)
0

1
1

1
3 2
0
0

2
1

4 3

0
0
0

1
2

1
0
0
0

For defining the matrices for i = 0,1,2 the following component matrices are required
1 1
0
0

0
0
2 2
0

0
3
=
for 1 n M
(14)
0
0
3
0

0
0
0

k

is a rectangular matrix of type (

i=1 ( ) for 1 n M ; is a rectangular matrix of
=1 )x
order and 0 appearing as (i, j) component of (14) is a block zero rectangular matrix of type x for

1 i, j k*. Let , = ,+1

e for 1 n N -1 is a matrix of type x for 1 i k* where ,
=

= for 1 j N and let

1,
0
0
0
0
2, 0
0
=
for 1 n N-1.
(15)

0
0
0 ,

This is a rectangular matrix of type ( =

=1 )
=1 and 0 appearing in the (i, j) component is a
rectangular 0 matrix of type x for 1 i, j k*.

International organization of Scientific Research

4|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

1 1,1
e

2,1
e

0
2
0
0
(16)

0
0
0 1 ,1 e

In (16), U is a rectangular matrix of type ( =

=1 )
=1 and 0 appearing in the (i, j)

component is a rectangular 0 matrix of type x for 1 i, j k*. ,

is a rectangular matrix of type
for 1 i k*.The matrix 0 is same as that of 0 when in the first row of 0 is replaced by .
The matrix 1 is given below. The matrix 2 is same as that of 2 when the first block column with 0 is

considered as
=1 columns block instead of =1 columns block of 2 . To write 1 the block for 0 is to
be considered which has queue length, L= 0, 1, 2M-1. When L = 0 there is only arrival process and no service
process. The change in environment from i to j switches BMAP j version as started by its stationary probability
vector in the new environment for 1 i j, k*. When an arrival occurs and queue length becomes L in the
environment i next arrival time starts and the service time starts with starting probability vector for 1 i k*.
In the 0 when L =1, 2, , M-1 all the processes arrival, service and environment are active as in other blocks
for n > 0. Considering the change of environment switches on the BMAP as per stationary probability vector in
the new environment when the queue is empty, the following matrix of order
=1 is defined which is
concerned with change of environment during arrival.
1
1,2 1,3 1,
2,1
2
2,3 2,
3
3,
= 3,1 3,2
(17)

,1 ,2 ,3
Here = 0 + (1 ), and , is a rectangular matrix of type x whose all rows are equal to (1 ),
presenting the rates of changing to phases in the new environment for i j and 1 i, j k*.

1
2 2 1 2
1

1
3 2 1
3
2
1
1

4 3 2
4
3

1 1 2

1
2
2 1
1 =
(18)
0

1
1

1
3 2
0
0

2
1

4 3

0
0
0

1
2

1
0
0
0

Let U =

The basic generator of the bulk queue which is concerned with only the arrival and service is a matrix of order

[
(20)
=1 ] given above in (19) where =0 + 1 + 2

Its probability vector w gives, =0 and w. e = 1

(21)
It is well known that a square matrix in which each row (after the first) has the elements of the previous row
shifted cyclically one place right, is called a circulant matrix. It is very interesting to note that the matrix is
a block circulant matrix where each block matrix is rotated one block to the right relative to the preceding block

partition. In (19), the first block-row of type [

=1 ] x[ =1 ] is, = ( + , 1 , 2 ,
, 2 , 1 , + , , 2 + 2 , 1 + 1 ) which gives as the sum of the blocks +
+ 1+ 2 ++2+ 1++++2+2+ 1+1= which is the
matrix given by

International organization of Scientific Research

5|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

1
1,2 1,3 1,
2,1
2
2,3 2,
3 3,
= 3,1 3,2
(22)

,1 ,2 ,3
where using (5) and (6), = ( ) + ( S i ) + (1 ), for 1 i k*. The stationary probability
vector of the basic generator given in (19) is required to get the stability condition. Consider the vector w = (
1 1 1 , 2 2 2 ,, ) where = (1 , 2 , , ) is the stationary probability vector
of the environment, are the stationary probability vectors of the i version BMAP and i version BMSP
and respectively. It may be noted ( )[( ) + ( )] =0. This gives ( ) =
(1 ), ( ) = (1 ), ( ) for 1 i k*. Now the first column of the matrix multiplication
of w is 1 (1 )1,1 1,1 1,1 + 2 (1 )2,1 11 11 [(2 2 )] +.....+ (1 ),1 11 11 [( )] = 0
since ( ) = 1 and 1 =0. In a similar manner it can be seen that the first column block of w is
1 (1 )1,1 1 1 + 2 (1 )2,1 1 1 [(2 2 )] +.....+ (1 ),1 1 1 [( )] = 0 and ith column block is 1 (1 )1, [(1 1 )] + 2 (1 )2, [(2 2 )] +.....+ (1 ),
+ + (1 ), [( )]= 0. This shows that + + 1 + 2 ++2 +
1 + + ++2 + 2 + 1 + 1 = w =0. So (w, w,,w) .W= 0 = (w, w,
.w) W where W is the transpose W. This shows (w,w...w) is the left eigen vector of and the

corresponding probability vector is

w =
, , ,..,
where w is given by

w = ( 1 (1 1 ), 2 (2 2 ),, ( ) )
(23)
Let = ( , ) and = ( , ) be the stationary probability components of the arrival and service processes.
Neuts [5], gives the stability condition as, w 0 < 2 where w is given by (23). Taking the sum
cross diagonally in the
0 2
matrices, it
can be
seen
using (9)
that
1

=1 =1 (
=1 ( )( ) =

1
1

) =
2 =
=1 =
=1 =1 ( )(
=1 (

1
1

) =
=1 =1 ( ) =
=1 ( =1 ) . This gives the stability condition

as
(24)
=1
=1 ( =1 ) <
=1

The sum ( =1 ) and =1 are known as the fundamental rates or the stationary rates of

w 0 =

=1

=1

<
=1 )

arrival / service of the BMAP/ BMSP processes corresponding to the environment i for 1 i k*. This result
(24) is the stability condition for the random environment BMAP/BMSP/1 queue with random environment
where maximum arrival size is greater than the maximum service size. When (24) is satisfied, the stationary
distribution of the queue length exists Neuts [5]. Let (0, i, j) : for 1 i k* 1 j ); (0, k, i, j, j) ; for 1
k M-1; 1 i k*; 1 j ; 1 j and (n, k, i, j, j): for 0 k M-1; 1 i k*; 1 j ; 1 j
and n 1 be the stationary probability vectors of Markov chain X(t) states in (1) for this model.
Let 0 =((0,1,1),(0,1,2)(0,1,1 ),(0,2,1),(0,2,2)(0,2,2 )(0,k*,1),(0,k*,2)(0,k*, ),
(0,1,1,1,1),(0,1,1,1,2)(0,1,1,1 , 1 ),(0,1,2,1,1),(0,1,2,1,2)(0,1,2,2 , 2 ),(0,1,3,1,1),(0,1,3,1,2)
(0,1,3,2 , 2 )(0,1,k*,1,1),(0,1,k*,1,2)(0,1,k*, , ),(0,2,1,1,1)(0,2,k*, , )

(0,M-1,1,1,1),(0,M-1,1,1,2)(0,M-1,k*, , )) be a vector of type 1x[

=1 + ( 1) =1 ].
Let =((n,0,1,1,1),(n,0,1,1,2)(n,0,1,1 , 1 ),(n,0,2,1,1),(n,0,2,1,2)(n,0,2,2 , 2 ),(n,0,3,1,1),
(n,0,3,1,2)(n,0,3,3 , 3 )(n,0,k*,1,1),(n,0,k*,1,2)(n,0,k*, , ),(n,1,1,1,1),(n,1,1,1,2)
(n,1,1,1 , 1 ),(n,1,2,1,1),(n,1,2,1,2)(n,1,2,2 , 2 ),(n,1,3,1,1),(n,1,3,1,2)(n,1,3,3 , 3 )(n,1,k*,1
,1),(n,1,k*,1,2)(n,1,k*, , ),(n,2,1,1,1)(n,2,k*, , ),(n,3,1,1,1)(n,3,k*, , )
(n,M-1,1,1,1),(n,M-1,1,1,2)(n,M-1,k*, , )) be a vector of type 1x[
The stationary
=1 ].
probability vector = (0 , 1 , 3 , ) satisfies the equations =0 and e=1. (25)
From (25), it can be seen 0 1 + 1 2 =0.
(26)
0 0 +1 1 +2 2 = 0
(27)
1 0 + 1 ++1 2 = 0, for n 2.
(28)
Introducing the rate matrix R as the minimal non-negative solution of the non-linear matrix equation
0 +R1 +2 2 =0,
(29)
it can be proved (Neuts [5]) that satisfies the following. = 1 1
for n 2.
(30)
Using (26),
0 satisfies
0 = 1 2 (1 )1
(31)
So using (27) and (31) and (30) the vector 1 can be calculated up to multiplicative constant since 1 satisfies
the equation
1 [2 1 1 0 + 1 + 2 ] =0.
(32)
Using (31) and (30) it can be seen that 1 [2 (1 )1 e+(I-R)1 ] = 1.
(33)

International organization of Scientific Research

6|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

Replacing the first column of the matrix multiplier of 1 in equation (32), by the column vector multiplier of
1 in (33), a matrix which is invertible may be obtained. The first row of the inverse of that same matrix is 1
and this gives along with (31) and (30) all the stationary probabilities of the system. The matrix R is iterated
1
2
starting with 0 = 0; and finding + 1 = 0 1
1 ()2 1 , for n 0. The iteration may be
terminated to get a solution of R at a norm level where + 1 () < .
2.3. Performance Measures of the System
(i) The probability of the queue length S = r > 0, P(S=r) can be seen as follows. For 1 r M-1, P(S =r) =

=1 1 =1 2 =1 (0, , , 1 , 2 ). For r M, let n and k be non negative integers such that r = n M + k. Then
P(S=r) =

1 =1

=1

2 =1 (, , , 1 , 2 )

, where r = n M + k, n 1 and k 0.

=1

(34)

=1

(ii) The probability that the queue length is zero is P(S =0) =
0, , .
(35)
(iii) The expected queue level E(S), can be calculated. Using (35) and (34), it may be seen that

1
E(S)= 0 ( = )=
=1 =1 0 0, , + =1
=1 1 =1 2 =1 (0, , , 1 , 2 )

1 =1 2 =1 (, , , 1 , 2 )( + )

=1 .(MnMn,Mn+1,Mn+1,Mn+2,Mn+2,Mn+M-1,
2 =1 (0, , , 1 , 2 ) +

1
=1 1 =1 2 =1 (0, , , 1 , 2 ) + M =1 +1 ( ) .

Here = 0, 0,1, ,1,2, ,2, , 1, , 1

is of type [( =1 )M]x1 column vector
in which consecutively (

)
times
0,1,2,3..,M-1
appear. Let it be called when 0 appears
=1
(

)
times
and
others
in
that
order
appear
(
times.
Then
=1
=1 )
1
2
E(S)= 0 + 1 ( ) + 1 ( )
(36)
2
2 2
2
2
2

=1
1
=1

1
=0
=1
1

=
=1 1 =1
Mn+M-1)= 1
=1

(iv)Variance of S can be derived. Let be column vector =[0, . . ,0, 1 , 1 2 , . . , 2 , 1) , , ( 1)

of type [(
=1 ) M]x1in which consecutively ( =1 ) times squares of 0,1,2,3.., M-1 appear. Let it be

called when 0 appears ( =1 ) times and others in the same manner as in appear (
=1 ) times. Then

2
1
2
2
E( )= 0 ( = )= =1 =1 0 0, , + =1 =1 1 =1 2 =1 (0, , , 1 , 2 )

2
2
=1 1 + =1 +
1 =1 2 =1 (, , , 1 , 2 )( + ) =0 +

=1 + 2M =1 .
So, E( 2 )= 0 + 2 [1 ( )3 2 + 1 ( )2 ] + 1 ( )1 + 21 ( )2
2
2

=1

1
=0

=1

(37)

VAR(S)=E( ) [()] may be written from (36) and(37).

III.

MODEL (B): MAXIMUM ARRIVAL SIZE M < MAXIMUM SERVICE SIZE N

The dual case of Model (A), namely the case, M < N is treated here. (When M =N both models are applicable
and one can use any one of them.) The assumption (vii) of Model (A) is changed and all its other assumptions
are retained.
3.1.Assumption
(vii) The maximum arrival size M= ma1 is less than the maximum service size
N=ma1 .
3.2.Analysis
Since this model is dual, the analysis is similar to that of Model (A). The differences are noted below. The state
space of the chain is as follows presented in a similar way.
The state of the system of the continuous time Markov chain X (t) under consideration is presented as follows.
X(t) = {(0, i, j) : for 1 i k* 1 j )} U {(0, k, i, j, j) ; for 1 k N-1; 1 i k*; 1 j ; 1 j }
U {(n, k, i, j, j): for 0 k N-1; 1 i k*; 1 j ; 1 j and n 1}. (38)
The chain is in the state (0, i, j) when the number of customers in the queue is 0, the environment state is i for
1 i k*and the arrival phase is j for 1 j . The chain is in the state (0, k, i, j, j) when the number of
customers is k for 1 k N-1, the environment state is i for 1 i k*, the arrival phase is j for 1 j and
the service phase is j for 1 j . The chain is in the state (n, k, i, j, j) when the number of customers in the
queue is n N + k, for 0 k N-1 and 1 n < , the environment state is i for 1 i k*, the arrival phase is j for
1 j and the service phase is j for 1 j . When the number of customers waiting in the system is r,
then r is identified with (n, k) where r on division by N gives n as the quotient and k as the remainder. The
infinitesimal generator of the model has the same block partitioned structure given in (4) for Model (A) but
the inner matrices are of different orders.

International organization of Scientific Research

7|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

1 0 0
0
.
.
.

2 1 0 0
.
.
.

0
.
.
2
1
0
=
(39)
0
0 2 1 0 0
.
0
0
0 2 1 0 0

In (39) the states of the matrices are listed lexicographically as 0, 1, 2, 3, . For partition purpose the zero states
in the first two sets of equation (38) are combined. The vector 0 is of type

1x[
=1 + ( 1) =1 ]with
0=((0,1,1),(0,1,2),(0,1,3)(0,1, 1 ),(0,2,1),(0,2,2),(0,2,3)(0,2, 2 )(0,k*,1),(0,k*,2),(0,k*,3)(0,k*, ),
(0,1,1,1,1),(0,1,1,1,2)(0,1,1,1, 1 ),(0,1,1,2,1),(0,1,1,2,2)...(0,1,1,2,1 ),(0,1,1,3,1)...(0,1,1,3,1 )(0,1,1,1 ,1)
(0,1,1,1 , 1 ),(0,1,2,1,1),(0,1,2,1,2)(0,1,2,1, 2 ),(0,1,2,2,1),(0,1,2,2,2).(0,1,2,2,2 ),(0,1,2,3,1)....(0,1,2,3,
2 )(0,1,2,2 ,1)(0,1,2,2 , 2 ),(0,1,3,1,1)(0,1,3,3 , 3 )(0,1,k*,1,1),,(0,1,k*, , ),(0,2,1,1,1),(0,2,
1,1,2)(0,2,k*, , ),(0,3,1,1,1)(0,3,k*, ),(0,4,1,1,1)(0,4,k*, )(0,N-1,1,1,1)
(0,N-1,k*, , )) and the vector is of type 1x[
=1 ] and is given in a similar manner as follows
=(n,0,1,1,1),(n,0,1,1,2)...(n,0,1,1, 1 ),(n,0,1,2,1)(n,0,1,2,1 )(n,0,k*,1,1)(n,0,k*, , ),(n,1,1,1,1)
(n,1,k*, , ),(n,2,1,1,1)...(n,2,k*, , )(n,N-1,1,1,1),(n,N-1,1,1,2)(n,N-1,k*, , )).

The matrices1 1 have negative diagonal elements, they are of orders [

=1 + ( 1) =1 ] and

[ =1 ]
respectively
and
their
off
diagonal
elements
are
non-negative.

The matrices 0 2 have nonnegative elements and are of order [ =1 ] . The matrices

0 2 have non-negative elements and are of types [

=1 + ( 1) =1 ] x [ =1 ]

and [ =1 ]
x [ =1 + ( 1) =1 ] and they are given below. Using Model (A) for
definitions of , , , and U and letting and as in Model (A), the partitioning matrices are defined
as follows. The matrix 0 is same as that of 0 with first zero block row is of order

[
=1 ] x[ =1 ]. The matrix 2 is same as that of 2 except the first column block is of type

[ =1 ] x [
=1 ] and is ( , 0, .0) where
1 1 e

0
=

0 =
1

2
1

0
0

3
2

0
0
0
2 =

0
0
0

0
0

0
0
0

0
2

0
0
0

0
0

0
0
0

0
0
0

0
0
0

0
0

3
4
5
6

0
0

2
3
4
5

2
1

0
0
0

1
2
3
4

2
1

International organization of Scientific Research

(40)

(41)

(42)

8|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

1 =

1 =

1
2

2
1

+1

2
1

2
1

3
2

3
2
1

4
3

2
1

2
1

3
2

2
1

3
2

3
2
1

4
3

1 + 1
+

3
2
1

3 + 3
2 + 2

1 + +1
2 + +2`

+
1

2
1
+

2 + 2
1 + 1

1
2

1
2

2
1

3
2

1
2

1
2

2
1

3
2

2
1

+
1 + +1

1 + 1
+
1

2
1

1
+

2 + 2
1 + 1
+

3
2

0
0

2
1

2
1
0
0

2
1

0
0
0

1
2
3

1
0
0
0

1
2
3

0
0
0

(43)
1
2

0
0
0

1
2

+
1 + +1
2 + +2

(44)

1
2

1
+ (45)
1 + +1

1 + 1
+

The basic generator which is concerned with only the arrival and service is = 0 + 1 + 2 . This is also
block circulant. Using similar arguments given for Model (A) it can be seen that its probability vector is w =

, , ,..,
where w is given by (23) and the stability condition remains the same. Following the

arguments given for Model (A), one can find the stationary probability vector for Model (B) also in matrix
geometric form. All performance measures including expectation of customers waiting for service and its
variance for Model (B) have the form as in Model (A) except M is replaced by N.

IV.

NUMERICAL ILLUSTRATION

Numerical examples are presented for five cases, namely, (i) M=N=4; (ii) M=4,N=3 ; (iii) M=4, N=2; (iv) M=3,
1 1
N=4and (v) M=2, N=4. There are two environments with generator
. For the case (i) M=N=4 the batch
1 1
Markovian arrival and service processes (BMAP and BMSP) have representations given by { : 0 n 4 and
i=1, 2 } and { : 0 n 4 and i=1, 2}respectively where the BMAP arrival matrices are for environment 1,
3 1
1
. 15 . 15 . 2
. 09 . 09 . 12
. 03 . 03 . 04
. 03 . 03 . 04
01 = 1 4 2 , 11 = . 18 . 18 . 24 , 21 = . 09 . 09 . 12 , 31 = . 03 . 03 . 04 , 41 = 0
0
0
3
1 5
. 21 . 21 . 28
. 06 . 06 . 08
. 03 . 03 . 04
0
0
0
3 1
0
. 48 . 48 . 24
. 24 . 24 . 12
and for environment 2, 02 = 0 4 2 , 12 = . 4 . 4
. 2 , 22 = . 32 . 32 . 16 ,
3
0 5
.4 .4
.2
. 24 . 24 . 12
. 08 . 08 . 04
0 0 0
4 1
32 = . 08 . 08 . 04 , and 42 = 0 0 0 . The service BMSP matrices are for environment 1, 01 =
,
1 5
. 16 . 16 . 08
0 0 0
.6
.9
. 36 . 54 1 . 12 . 18
. 12 . 18
4 1
11 =
, 1=
, =
, and 41 =
and for environment 2, 02 =
,
. 96 1.44 2 . 48 . 72 3 . 16 . 24
0
0
2 3
1.08 . 72 2 . 18 . 12
0 0
12 =
, 3 =
and 42 =
The starting probability vectors of service processes for the two
.3
.2
. 06 . 24
0 0
environments on arrival of customers when the queue is empty are 1 =(.4, .6) and 2 =(.6, .4). For the case (ii)

International organization of Scientific Research

9|Page

BMAP/BMSP/1 Queue with Randomly Varying Environment

M=4, N=3 the above batch arrival and batch service rates (matrices) of case (i) are assumed except two service
. 24 . 36
0 0
rates matrices which are replaced as 31 =
, and 41 =
. For the case (iii) M=4, N=2 the above
. 16 . 24
0 0
batch arrival and batch service rates (matrices) of case (i) are assumed except the six service rates matrices
.6
.9
. 72 . 48 1 1
which are assumed as 21 =
, 2=
, 3 =4 =0 matrices and 32 =42 =0 matrices. For the case
. 64 . 96 2 . 3
.2
(iv) M=3, N=4 the above batch arrival and batch service rates matrices of case (i) are assumed except two
. 06 . 06 . 08
0 0 0
arrival rates matrices which are replaced as 31 = . 03 . 03 . 04 and 41 = 0 0 0 . For the case (v) M=2, N=4
. 03 . 03 . 04
0 0 0
the above batch arrival and batch service rates (matrices) of case (i) are assumed except the six arrival rates
. 15 . 15 . 2
. 32 . 32 . 16
matrices which are assumed as 21 = . 12 . 12 . 16 , 22 = . 4 . 4
. 2 , 31 =41 =0 matrices and 32 =42 =0
. 09 . 09 . 12
.4 .4
.2
matrices. The partitioned matrices are of order 48, the rate matrix R is of order 48 and fifteen iterations are
performed to evaluate R matrix. The results obtained for various performance measures are tabulated in table 1.
Queue length probabilities and the probabilities of batch sizes are estimated. These probabilities and expected
queue lengths show variations depending on arrival, service rates and batches of arrival and service sizes. The
figures (1) and (2) present the probabilities of various levels and blocks.
Table 1. Results Obtained for the Five Cases.
P(S=0)
P(S=1)
P(S=2)
P(S=3)
0e
1e
2e
3e
( n 4)e
Arr rate
Ser rate
Norm
E(S)
Std(S)

M=4=N
0.380758062
0.129820258
0.110187529
0.086386188
0.707152037
0.185713263
0.067545585
0.024943425
0.01464569
0.59475
1.110243056
4.82513E-05
2.621582695
3.863770414

M=4,N=3
0.377246667
0.129530535
0.110124761
0.086540194
0.703442157
0.187070777
0.068669501
0.025583927
0.015233637
0.59475
1.0921875
5.40723E-05
2.656880644
3.904750478

M=4,N=2
0.348733946
0.125595408
0.108434643
0.086908008
0.669672005
0.197490543
0.078987097
0.032001528
0.021848827
0.59475
1.004108796
0.000113132
2.992491652
4.318346577

M=3,N=4
0.3849806
0.131199748
0.111159883
0.088387612
0.715727843
0.181491843
0.06535554
0.023784268
0.013640506
0.588958333
1.110243056
4.49819E-05
2.559407301
3.787733035

M=2,N=4
0.421313996
0.143065373
0.131171064
0.081273082
0.776823515
0.159704069
0.045203898
0.013006974
0.005261545
0.540481481
1.110243056
1.49854E-05
2.030074169
3.021600968

0.45
0.4
0.35
0.3
P(S=0)

0.25

P(S=1)

0.2

P(S=2)

0.15

P(S=3)

0.1
0.05
0
M=4=N

M=4,N=3

M=4,N=2

M=3,N=4

M=2,N=4

Figure 1. Probabilities of Queue Lengths

International organization of Scientific Research

10 | P a g e

BMAP/BMSP/1 Queue with Randomly Varying Environment

0.9
0.8
0.7
0.6

0e

0.5

1e

0.4

2e

0.3

3e

0.2

( n 4)e

0.1
0
M=4=N

M=4,N=3

M=4,N=2

M=3,N=4

M=2,N=4

Figure.2. The probabilities of customer blocks

V.

CONCLUSION

Two BMAP/BMSP/1 bulk arrival and bulk service queues with random environment have been studied
by identifying the maximum of the arrival and the service sizes and by grouping the customers as members of
blocks of such maximum sizes. Matrix geometric results have been obtained by partitioning the infinitesimal
generator by grouping of customers, environment state BMAP and BMSP phases together. The basic system
generators of the queues are block circulant matrices which are explicitly presenting the stability condition in
standard forms. Numerical results for bulk queue models are presented and discussed. Effects of variation of
rates on expected queue length and on probabilities of queue lengths are exhibited. The decrease in arrival rates
(so also increase in service rates) makes the convergence of R matrix faster which can be seen in the decrease of
norm values. The standard deviations also decrease. The BMAP/BMSP/1 queue with random environment has
number of applications. The process includes Exponential, Erlang, Hyper Exponential, Coxian distributions and
PH distributions as special cases and the PH distribution is also a best approximation for a general distribution.
Further the BMAP/BMSP/1 queue is a most general form almost equivalent to G/G/1 queue. The bulk arrival
models because they have non-zero elements or blocks above the super diagonals in infinitesimal generators,
they require for studies the decomposition methods with which queue length probabilities of the system are
written in a recursive manner. Their applications are much limited compared to matrix geometric results. From
the results obtained here, provided the maximum arrival and service sizes are not infinity, the most general
model of the BMAP/BMSP/1 queue with random environment admits matrix geometric solution. Further
studies with block circulant basic generator system may produce interesting and useful results in inventory
theory and finite storage models like dam theory. It is also noticed here that once the maximum arrival or
service size increases, the order of the rate matrix increases proportionally. However the matrix geometric
structure is retained and rates of convergence is not much affected. Inventory models with BMAP and BMSP
assumptions may be focused for further study which may produce more general results.

VI.

ACKNOWLEDGEMENT

The second author thanks ANSYS Inc., USA, for providing facilities. The contents of the article published are
the responsibilities of the authors.

REFERENCES
[1]
[2]
[3]
[4]

Bini.D,Latouche.G,and-Meini.B, Numerical methods for structured Markov chains, Oxford Univ.

Press, 2005.
Chakravarthy.S.R and Neuts. M.F, Analysis of a multi-server queue model with MAP arrivals of special
customers,SMPT,-Vol.43,2014,pp.79-95,
Gaver, D., Jacobs, P., Latouche, G, Finite birth-and-death models in randomly changing environments.
AAP.16,1984,pp.715731
Latouche.G, and Ramaswami.V, Introduction to Matrix Analytic Methods in Stochastic Modeling,
SIAM. Philadelphia.1998

International organization of Scientific Research

11 | P a g e

BMAP/BMSP/1 Queue with Randomly Varying Environment

[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]

[15]
[16]

Neuts.M.F,Matrix-Geometric Solutions in Stochastic Models: An algorithmic Approach, The Johns

Hopkins Press, Baltimore,1981.
Rama Ganesan, Ramshankar.R, and Ramanarayanan.R, M/M/1 Bulk Arrival and Bulk Service Queue
with Randomly Varying Environment,IOSR-JM,Vol.10,Issue6,Ver.III,2014,pp58-66.
Sandhya.R, Sundar.V, Rama.G, Ramshankar.R and Ramanarayanan.R, M/M//C Bulk Arrival And Bulk
Service Queue With Randomly Varying Environment,IOSR-JEN,Vol.05,Issue02,||V1||,2015,pp.13-26.
Ramshankar.R, Rama Ganesan, and Ramanarayanan.R PH/PH/1 Bulk Arrival and Bulk Service Queue,
IJCA,Vol.109,No.3,2015,pp.27-33.
Ramshankar.R, Rama.G, Sandhya.R, Sundar.V and Ramanarayanan.R,(2015),PH/PH/1 Bulk Arrival and
Bulk Service Queue with Randomly Varying-Environment,-IOSRJEN,Vol.05.Issue02-||V4||.pp.01-12.
Rama.G, Ramshankar.R, Sandhya.R, Sundar.V and Ramanarayanan.R, (2015),BMAP/M/C Bulk Service
Queue with Randomly Varying Environment, IOSRJEN, Vol.05, Issue 03,||V1||,pp.33-47.
Lucantoni.D.M, (1993), Models and Techniques for Performance Evaluation of Computer and
Communication Systems,L. Donatiello and R. Nelson-Editors-Springer-verlag,,-pp330-358.
Cordeiro and Kharoufch, J.P, Batch Markovian Arrival Processes (BMAP),www.dtic.mil/get-trdoc/pdf/AD=ADA536697&origin=publication-Detail.
Qi-Ming-He,Fundamentals-of-Matrix-Analytic-Methods,Springer,2014.
Rama Ganesan, and Ramanarayanan R, 2014, Stochastic Analysis of Project Issues and Fixing its Cause
by Project Team and Funding System Using-Matrix-Analytic-Method,IJCA,VOL.107,No.7,2014,pp.2228.
Usha.K, Contribution to The Study of Stochastic Models in Reliability and Queues, Ph.D Thesis, 1981,
Annamlai University,India.
Usha.K, The PH/M/C Queue with Varying Environment,Zastosowania Matematyki Applications
Mathematicae,XVIII,2 1984, pp.169-175.

International organization of Scientific Research

12 | P a g e