On a Single Server Queueing Inventory

System

K. A. K. AL Maqbali(B) , V. C. Joshua, and A. Krishnamoorthy

Department of Mathematics, CMS College Kottayam, Kerala, India

{khamis,vcjoshua,krishnamoorthy}@cmscollege.ac.in

Abstract. In this paper, we consider a queueing inventory model. Cus-

tomers arriving to a single server queueing system in which, server uses
some commodities from the inventory to fulfill the service. Service of a
customer is initiated only when the server is free and the commodities
for the service are available in the inventory. Otherwise, the arriving cus-
tomer joins the queue. The inventory level is managed with (s,S) policy.
Lead time is assumed to be positive. Steady state analysis of the model
is performed. Various performance measures are estimated. A numerical
example is provided.

Keywords: Queueing inventory · Lead time · Phase type

distribution · Matrix analytic method

1 Introduction

In many real life phenomena, customers waiting in the queue need inventoried
items to complete the service. The study of such models fall under the category
of Queueing Inventory Models. M/M/1 Queues with attached inventories were
studied by Krishnamoorthy et al. [2,3], Neethu and Dhanya [4] and Saffari et al.
[5]. Sometimes, customer may go through different phases of service to get the
items. Krishnamoorthy et al. [1] studied a P H/P H/1 inventory system under
(s, S) policy when the lead time is zero. Another paper by Sigman and Simchi-
Levi [6] studied an M/G/1 queue with limited inventory.
In this paper, we consider an M/P H/1 queueing system with inventory level
under (s, S) policy. When the inventory level reaches s, an order is placed. We
assume that the customer immediately takes the item when he starts to go
through different phases of service. Also, we assume that the lead time is positive
and follows exponential distribution with rate θ. The shortage is permitted.
Customers arrive to a single server according to Poisson Process with arrival
rate λ. An arriving customer can immediately obtain the service if the server is
idle and at least one item is available in the inventory. Otherwise, the customer

* Supported by the Indian Council for Cultural Relations (ICCR) and Ministry of
Higher Education in Sultanate of Oman.
c Springer Nature Switzerland AG 2020
V. M. Vishnevskiy et al. (Eds.): DCCN 2020, LNCS 12563, pp. 579–588, 2020.
https://doi.org/10.1007/978-3-030-66471-8_44
580 K. A. K. A. Maqbali et al.

must wait in the queue. We assume that the service time follows a phase-type dis-
tribution (PH-distribution) with representation (β, T ) of order m. The customer
takes one item of inventory at the beginning of the service.
In other words, this model can be described as follows: customers occur
according to a Poisson Process with arrival rate λ and the service time, which is
attached with inventory level, follows PH-distribution with representation (β, T )
of order m. Under (s, S) policy, the lead time follows exponential distribution
with rate θ. The order is placed when the inventory level reaches s. Hence,
replenishment of s, s − 1, s − 2, ..., 0 to become S, occurs according to the rate
of lead time. The arriving customer can not go through different phases of the
service when inventory level is zero and he must wait in queue until inventory
level becomes S.
The first arriving customer can immediately go through different phases of
the service when at least one item is available. At the same time, he directly
takes one item of inventory at the beginning of his service. Instantaneously,
if inventory level reaches s, an order decision is made based on the reorder
point s and replenishment depends on the lead time, which follows exponential
distribution with rate θ. The next arriving customer can directly go through
different phases of service when the previous customer completed his service.
Otherwise, he has to wait in the queue until the previous customer completes
his service. This keeps going.
The motivation for the model comes from the hospital inventory manage-
ment. For example, heart patients are waiting for angioplasty stent placement
for cardio vascular treatment. In general, hospitals keep coronary artery stents
in store. When coronary artery stents are available in store and the operation
theatre (service station) is available, the heart patient may go through different
phases in hospital and the coronary artery stent is immediately taken from store
to the patient at the beginning of his service. If the coronary artery stent is
not available in store, the heart patient has to wait until the availability of the
stents.
Another motivating example is the vehicle service repair station, which deals
only with one type of parts. When parts are available in store and service station
is free, the vehicle may be taken to service station and one inventory is directly
taken from store at the beginning of its service. If these parts are not available
in store of the service station, the vehicle has to wait until the availability of the
parts.

2 Mathematical Description of the Model

For the analysis of the model, we introduce the following notations.

Let
N (t) be the number of customers in the system at time t.
I(t) be the number of items, which are the same type, in the inventory at
time t.
M (t) be the phase of service.
 On a Single Server Queueing Inventory System 581

X(t) = {(N (t), I(t), M (t)); t ≥ 0} is a continuous time Markov Chain on the
state space.
Therefore, this model can be studied as a level Independent Qusi-Birth-Death
(LIQBD) process with state space is given by
Ω = {(0, i); 0 ≤ i ≤ S} ∪ {(n, 0); n ≥ 1} ∪ {(n, i, m); n ≥ 1; 0 ≤ i ≤ S; m =
1, 2, . . . }.
The terms of transitions of the states are given in the Table 1.

Table 1. Intensities of Transitions

From To Transition Rate

(0, i) (1, i − 1, m) 1≤i≤S λβm
(n, i, m) (n + 1, i, m) n ≥ 1; 0 ≤ i ≤ S λ
(n, 0) (n + 1, 0) n≥0 λ
(0, i) (0, S) 0≤i≤s θ
(n, 0) (n, S − 1, m) n≥1 θβm
(n, i, m) (n, S, m) i ≤ s; n ≥ 1 θ
0
(1, i, m) (0, i) 0≤i≤S τm
′
0
(n, i, m) (n − 1, i − 1, m ) 0 < i ≤ S ; n > 1 τm βm′
′ ′
(n, i, m) (n, i, m ) 1 ≤ n; 0 ≤ i ≤ S; m = m τmm′
0
(n, 0, m) (n − 1, 0) n≥2 τm

The infinitesimal generator Q of the level Independent Qusi-Birth-Death

(LIQBD) process with state space is of the form
⎛ ⎞
B00 B01
⎜ B10 A1 A0 ⎟
⎜ ⎟
Q=⎜
⎜ A 2 A 1 A 0 ⎟;
⎟
⎜
⎝ A 2 A 1 A0
⎟
⎠
.. .. ..
. . .
where 
−(λ + θ)I(s+1) Γ
B00 = ;
O(S−(s+1))×(s+1) −λI(S−s) (S+1)×(S+1)
where Γ = O(s+1)×(S−(s+1)) θe(s+1)×1 ;
⎛ ⎞
λ O
⎜ λβ1×m ⎟
B01 = ⎜ ;
⎜ ⎟
.. ⎟
⎝ . ⎠
λβ1×m O (S+1)×(1+m×(S+1))
582 K. A. K. A. Maqbali et al.
⎛ ⎞
0
0
⎜ Tm×1 ⎟
0
⎜ ⎟
B10 =⎜
⎜ Tm×1 ⎟
⎟ ;
⎜ .. ⎟
⎝ . ⎠
0
Tm×1 (1+m×(S+1))×(S+1)
⎛ ⎞
0 ... 0
0
⎜ Tm×1 ... O⎟
.. ⎟
⎜ ⎟
⎜
⎜
⎜ ∧ .⎟⎟
A2 = ⎜
⎜ ∧ ⎟
⎟ ;
⎜ .. ⎟
⎜
⎜ . ⎟
⎟
⎝ ∧ O⎠
∧ O (1+m×(S+1))×(1+m×(S+1))
0
where ∧ = Tm×1 ⊗ β1×m ;
⎛ ⎞
λ
⎜ λIm ⎟
⎜ ⎟
A0 = ⎜
⎜ λIm ⎟
⎟ ;
⎜ .. ⎟
⎝ . ⎠
λIm (1+m×(S+1))×(1+m×(S+1))

⎛ ⎞
−(λ + θ) O φ O
⎜ ϕO O θIm ⎟
.. .. ⎟
⎜ ⎟
⎜
⎜ Oϕ O . . ⎟
.. ⎟
⎜ ⎟
.
O ..
⎜
⎜ O . ⎟
A1 = ⎜
⎜ ⎟
⎜ O ϕ O . . . O θIm ⎟ ⎟
⎜
⎜ Oω O O ⎟ ⎟
⎜ .. .. ⎟
⎜
⎜ O .O . ⎟ ⎟
⎝ O ω O ⎠
O ω (1+m×(S+1))×(1+m×(S+1))

where ϕ = Tm×m − (λ + θ)Im ; ω = Tm×m − λIm and φ = θ ⊗ βm .

3 Steady-State Analysis
3.1 Stability Condition
Theorem 1. The Markov chain with the infinitesimal generator Q of the level
Independent Qusi-Birth-Death (LIQBD) is stable if and only if
λ<µ
0 S+1 0
Where µ = π 1 Tm×1 + [( i=2 π i ) (Tm×1 ⊗ β1×m )]em×1 ; where π 1 and π i are
sub vectors of order 1 × m.
 On a Single Server Queueing Inventory System 583

Proof. Let A = A2 + A1 + A0 . We can notice that A is an irreducible matrix.

Thus, the stationary vector π of A exists such that

πA = 0
πe = 1.
The Markov chain with the infinitesimal generator Q of the level Independent
Qusi-Birth-Death (LIQBD) is stable if and only if

πA0 e < πA2 e.

Recall, A2 is a square matrix of order (1 + m × (S + 1)) and

⎛ ⎞
0 ... 0
0
⎜ Tm×1 ... O⎟
.. ⎟
⎜ ⎟
0
⎜
A2 = ⎜⎜ ∧ . ⎟ ; where ∧ = Tm×1
⎟ ⊗ β1×m .
⎜
⎝ . ..
⎟
⎠
∧ O

πA2 e = (π0 , π 1 , π 2 , . . . , π S +1 )A2 e;

where π0 is a single value and π 1 , π 2 , . . . , π S +1 are sub vectors of order
1 × m.
0 0 0
πA2 e = [π 1 Tm×1 + π 2 (Tm×1 ⊗ β1×m ) + · · · + π S +1 (Tm×1 ⊗ β1×m )]e;
0 0 0
= π 1 Tm×1 + [π 2 (Tm×1 ⊗ β1×m ) + · · · + π S +1 (Tm×1 ⊗ β1×m )]em×1 ;
0 0
= π 1 Tm×1 + [π 2 + · · · + π S +1 ](Tm×1 ⊗ β1×m )em×1 ;
S+1
0 0
= π 1 Tm×1 +( π i )(Tm×1 ⊗ β1×m )em×1 .
i=2

0 S+1 0
Then, πA2 e = π 1 Tm×1 + [( i=2 π i )(Tm×1 ⊗ β1×m )]em×1 = µ.
Since πA0 e = λ and πA2 e = µ, then the queueing inventory system is stable
if and only if
λ<µ

3.2 Stationary Distribution

The stationary distribution of the Markov chain under consideration can

obtained by solving the set of Eqs. 1 and 2 [7].

XQ = 0; (1)

Xe = 1. (2)
Let X be decomposed with Q as following :
584 K. A. K. A. Maqbali et al.

X = (X0 , X1 , . . . ) where X0 = (x00 , x01 , x02 , . . . , x0S );

Xi = (Xi0 , Xi1 , . . . , XiS ); Xi0 = (xi0 , xi01 , xi02 , . . . , xi0m ) ;
Xij = (xij1 , xij2 , . . . xijm ) for j = 1, . . . , S and i = 1, 2, . . . .
From Eq. 1, we obtain the following equations.

X0 B00 + X1 B10 = 0; (3)

X0 B01 + X1 A1 + X2 A2 = 0; (4)
..
.
Xi−1 A0 + Xi A1 + Xi+1 A2 = 0 for i = 2, 3, . . . . (5)
It may be shown that there exists a constant matrix R such that

Xi = Xi−1 R for i = 2, 3, . . . .

The sub-vectors Xi are geometrically related by the equation

Xi = X1 Ri−1 for i = 2, 3, . . . . (6)

We can obtain the constant matrix R by using the matrix quadratic Eq. 7

R2 A2 + RA1 + A0 = 0. (7)

By successive substitution procedure [7], we can compute the constant matrix

R from
Rk+1 = −V − Rk2 W and R0 = 0;
where V = A0 A−11 and W = A2 A1 .
−1

Knowing the matrix R, X0 and X1 can be obtained by solving the Eqs. 3

and 4 by substituting X2 = X1 R into Eq. 4.
Since there is no unique solution, X0 and X1 must be normalized by using
the normalizing condition 8

X0 + X1 (I − R)−1 e = 1. (8)

Then, we use the formula 6 to compute remainder probabilities X2 , X3 , . . . .

4 Performance Measures

Then, we obtain some performance measures of the system as following:

1. Expected number of customers in the system

∞
E[N ] = iXi e.
i=0
 On a Single Server Queueing Inventory System 585

2. Expected number of items in inventory.

∞ S
E[I] = jXij e.
i=0 j=0

3. Probability that the server is idle

S ∞
b0 = x0j + xi0 .
j=0 i=1

4. Probability that a customer waits for service due to shortage of inventory

∞ ∞
b1 = xi0 + Xi0 e.
i=1 i=2

5. Expected number of customers waiting in the system due to lack of inventory

∞
E[N0 ] = (i − 1)Xi0 e.
i=2

5 Numerical Example

In order to illustrate the performance measures of the system numerically and

the effect of S on the performance measures of this system, we fix arrival rate
λ = 2, the rate of lead time θ = 0.6 and s = 3.
For PH-representation(β,
⎛ ⎞ T ) of order m,
⎛ we ⎞ fix m = 3, β = 0.5, 0.4, 0.1 ,
−6 3 2 1
T = ⎝ 1 −8 2 ⎠ and T 0 = −T e = ⎝ 5 ⎠.
2.5 1.5 −8 4
We find the performance measures of the system with respect to S corre-
sponding to above parameters in the Table 2.

5.1 The Effect of S on the Performance Measures of the System

In this section, we show the effect of S on the performance measures of the

system as following:

1. From Fig. 1, we can see that the expected number of customers in the system
E[N ] decreases as S increases and the expected number of customers waiting
in the system due to lack of inventory E[N0 ] decreases when S increases, too.
2. Also, we realize from Fig. 2 that the expected number of items in inventory
E[I] increases when S increases.
586 K. A. K. A. Maqbali et al.

3. In addition, from Fig. 3 we note that the probability that a customer waits
for service due to shortage of inventory b1 decreases when S increases and
probability that the server is idle b0 has almost no change as S increases. In
this system, the server is idle when the inventory level is zero or there is no
customer in the system or both.

Table 2. Effect of S on various performance measures

S E[N] E[I] b0 b1 E[N0 ]

7 13.0391 28.0000 0.3570 0.2853 4.3938
8 8.4679 36.0000 0.3570 0.2334 2.5157
9 6.5122 45.0000 0.3570 0.1972 1.7319
10 5.4328 55.0000 0.3570 0.1705 1.3089
11 4.7512 66.0000 0.3570 0.1502 1.0469
12 4.2827 78.0000 0.3570 0.1341 0.8700
13 3.9415 91.0000 0.3570 0.1212 0.7431
14 3.6823 105.0000 0.3570 0.1105 0.6479
15 3.4787 120.0000 0.3570 0.1016 0.5740
16 3.3147 136.0000 0.3570 0.0940 0.5151

Fig. 1. Effect of S on expected number of customers

 On a Single Server Queueing Inventory System 587

Fig. 2. Effect of S on expected number of items in inventory

Fig. 3. Effect of S on probability that the server is idle and on probability that a
customer waits for service due to shortage of inventory
588 K. A. K. A. Maqbali et al.

6 Conclusion
In this paper, we analyse an M/P H/1 system with inventory level under (s, S)
policy. Under Steady state condition, various performance measures are esti-
mated: expected number of customers in the system, expected number of cus-
tomers waiting in the system due to lack of inventory, expected numbers of items
in inventory and probability that a customer waits for service due to shortage
of inventory. We study the effect of S on the performance measures of the sys-
tem numerically. We find that the expected number of customers in the system
E[N ] and the expected number of customers waiting in the system due to lack
of inventory E[N0 ] decrease as S increases. Besides this, the expected number of
items in inventory E[I] increases when S increases. The probability that a cus-
tomer waits for service due to shortage of inventory b1 decreases and probability
that the server is idle b0 has almost no change as S increases.

Acknowledgement. The first author acknowledges the Indian Council for Cultural
Relations (ICCR) (Order No:2019-20/838) and Ministry of Higher Education (MOHE),
Order No:2019/35 in Sultanate of Oman for their supports.

References
1. Krishnamoorthy, A., Jose, K.P., Narayanan, V.C.: Numerical investigation of a
PH/PH/1 inventory system with positive service time and shortage. Neural Par-
allel Sci. Comput. 16, 579–592 (2008)
2. Krishnamoorthy, A., Manikandan, R., Lakshmy, B.: A revisit to queueing-inventory
system with positive service time. Ann. Oper. Res. 233(1), 221–236 (2013). https://
doi.org/10.1007/s10479-013-1437-x
3. Krishnamoorthy, A., Nair, S.S., Narayanan, V.C.: An inventory model with server
interruptions. In: 5th International Conference on Queueing Theory and Network
Applications (QTNA’10), Beijing (2010)
4. Neethu, P.R., Shajin, D.: On a queueing-inventory system with impatience of cus-
tomers. Int. J. Pure Appl. Math. 118(20), 909–916 (2018)
5. Saffari, M., Haji, R., Hassanzadeh, F.: A queueing system with inventory and mixed
exponentially distributed lead times. Int. J. Adv. Manuf. Technol. 53, 1231–1237
(2011)
6. Sigman, K., Simchi-Levi, D.: Light traffic heuristic for an M/G/1 queue with limited
inventory. Ann. Oper. Res. 40, 371–380 (1992)
7. Stewart, W.J.: Probability, Markov Chains, Queues, and Simulation. Princeton Uni-
versity Press, New Jersey (2009)