Analysis of M1 , M2 /G1 , G2 /1 retrial queueing system with

non-pre-emptive priority services, modified Bernoulli vacation,

working breakdown, repair and reneging
Abstract
This paper considers M1 , M2 /G1 , G2 /1 general retrial queueing system with non-pre-
emptive priority services. The server serves two type of units subject to modified Bernoulli
vacation. After completing the service, if there are no high priority units present in the system
then the server may go for a vacation. The high priority unit who find the server busy are queued
in the system and the low priority unit finds the server busy, they are routed to a retrial queue
that attempts to get the service. The system may become defective at any point of time when it
is in operation and does not sent for repair immediately. The remaining service will be given to
the unit after that only the server go for repair station. After completing the repair and vacation
if there is no high priority unit then the server becomes idle. The low priority unit renege the
orbit during server’s busy and unavailable periods. Using the supplementary variable technique,
the steady-state distributions of the server state and stability condition are obtained.

Keywords: Priority Queueing Systems, Working Breakdown, modified Bernoulli Vacation,

Retrial, Repair, Reneging.

MSC 2010 No.: 60K25, 68M 30, 90B22

1 Introduction
Retrial phenomenon naturally arises in various systems such as call centers, cellular networks and
random access protocols in local area networks. Phung-Duc, T. (2019) gives a comprehensive
survey on theory and applications of retrial queues. Queues including impatient units have been
dragged the notice of many authors and we see notable participation by numerous researchers in
this area. The important works on impatient behaviour of the units was studied by Cheng-Dar Liou
(2013). Dong-Yuh Yang and Ying-Yi Wu (2017) analyzed customer’s impatience behaviour with
server vacations. In this paper, the low priority units renege the orbit during the server’s busy and
unavailable periods.
Studying queueing models with server breakdown, it is generally assumed that the server stops
service when the server getting breakdown. However, in most of the models considered so far
of queueing systems with server breakdowns, the underlying assumption has been that a server
breakdown disrupts the service completely in the system. Kalidass and Kasturi (2012) introduced
the working breakdown policy, in which the server works at a lower service rate rather than stopping
service during the breakdown period. In this policy the service can decrease complaints from the
units who should wait for the server to be, repaired and reduces the cost of waiting units. Therefore,
the working breakdown service is a more reasonable repair policy for unreliable queueing systems.
Tao Li and Liyuan Zhang (2017) and Zaiming liu and Yang Song (2014) describes more about
working breakdowns. Bo Keun Kim and Doo Ho Lee (2016) studied the M/G/1 queueing system
with disasters and working breakdowns.
 In manufacturing systems based on the needs, the works are categorized and we give the
preference as high and low priority one. Here, the correct assignment of work will reduce the
time consumption and cost management in the service system. Kailash C. Madan (2011) studied a
non-preemptive priority queueing system with a single server serving two queues. Boutarfa L and
Djellab L studied the performance of the M1 , M2 /G1 , G2 /1 retrial queue with pre-emptive resume
policy (2015).
This paper considers M1 , M2 /G1 , G2 /1 general retrial queueing system with priority services.
Two different sort of units arrive at the system in two independent Poisson processes. Under the
non-pre-emptive priority rule, the server providing general service to the arriving unit. Here, we
propose a retrial queueing model with the additional characteristics of server’s working breakdown,
repair and reneging. Arriving high priority unit who find the server busy are queued and then are
served in accordance with FCFS discipline. But the arriving low priority unit on finding the server
busy cannot be queued and join the orbit as a retrial unit. After completing the service, if there are
no high priority units present in the system then the server may go for a vacation. The system may
become defective at any point of time when it is in operation due to normal breakdown. However,
when the system is defective, it does not sent for repair immediately. The remaining service will
be given to the interrupted unit. After that only the server go for repair station. After completing
the repair and vacation if there is no high priority unit then the server becomes idle. We consider
reneging to occur at the low priority unit while the server’s busy and unavailable periods.
The remaining content of the paper is structured as follows: Section 2 provides a mathematical
description of the model and governing equations of the model are formulated. The steady-state
solutions of the system and conclusion are provided in sections 3 and 4 sequentially.

2 Model Description
We analyze an unreliable server in a retrial queueing model with two sort of units. The basic
operation of the model can be described as follows.
Two class of units arrive at the system in two independent Poisson processes with arrival rate
λ1 (> 0) and λ2 (> 0) respectively. The high priority unit who find the server busy are queued in the
system and the low priority unit find the server busy, they are routed to a retrial queue and follow
constant retrial policy that tries to get the service. The retrial time is generally distributed with
distribution function I(s) and the density function i(s).
After completing the service of all high priority units and every service completion of low
priority unit the server may demand a vacation with probability θ or becomes idle. Vacation time is
generally distributed with distribution function V (s) and the density function v(s) respectively .
During the busy period, the server may become inactive due to normal breakdown with break-
down rate α, which is exponentially distributed. At the moment of breakdown, the unit who is in
service will get the service continuously at a lower service rate which is known as working break-
down. After that the repair process starts so as to regain its functionality.The working breakdown
time and repair time are generally distributed with distribution functions W (s) and R(s) and the
density functions w(s) and r(s) respectively.
If the server is busy or unavailable in the system, the low priority unit may renege the orbit
exponentially with rate ξ. After completing vacation and repair if there is no high priority unit
present in the system then the server becomes idle. Several stochastic processes associated in the

2
 system are independent of each other.
Let µi (x) dx, i =1,2,3, η(x) dx, β(x) dx and γ(x) dx be the conditional probability of comple-
tion rate of the service period of the high and low priority unit, working breakdown service, retrial
period, vacation period, and repair period given that the elapsed time is x.
dBi (x) dI(x) dV (x)
Also, µi (x)dx = ,i= 1, 2, 3, η(x)dx = , β(x)dx = , and
1 − Bi (x) 1 − I(x) 1 − V (x)
dR(x)
γ(x)dx = are the first order differential (hazard rate) functions of Bi (.), I(.), V(.)
1 − R(x)
and R(.) respectively.

2.1 State of the system

We get the probability generating function of the joint distribution of the state of the server including
the number in the system by using I 0 (t), B10 (t), B20 (t) and B30 (t) are the elapsed retrial time and
service time of the high and low priority units at time t respectively. Also V 0 (t) and R0 (t) are the
elapsed vacation time and elapsed repair time of the server as supplementary variables at time t
respectively. Assuming that the system is empty initially. N1 (t) and N2 (t) denotes the number of
units in the queue (high priority units) and the orbit (low priority units) respectively and Y(t) be the
state of the server at time t,


 0, if the server is in idle state;
1, if the server is in retrial state;




 2, if the server is busy with high priority unit;


Y(t) = 3, if the server is busy with low priority unit;
4, if the server is on vacation;




5, if the server is on working breakown state;




6, if the server is on repair process.


For the Markov process {Y (t); t ≥ 0} defined the probability as,

I 0,0 (s, t) = P r{Y (t) = 0, N1 (t) = 0, N2 (t) = 0},

and the probability densities are,

I 0,k2 (x, s, t)dx = P r{Y (t) = 1, N1 (t) = 0, N2 (t) = k2 }, k2 ≥ 0

(1)
P k1 ,k2 (x, s, t)dx = P r{Y (t) = 2, x < B10 (t) ≤ x + dx, N1 (t) = k1 , N2 (t) = k2 }, k1 , k2 ≥ 0,
(2)
P k1 ,k2 (x, s, t)dx = P r{Y (t) = 3, x < B20 (t) ≤ x + dx, N1 (t) = k1 , N2 (t) = k2 }, k1 , k2 ≥ 0,
V k1 ,k2 (x, s, t)dx = P r{Y (t) = 4, x < V 0 (t) ≤ x + dx, N1 (t) = k1 , N2 (t) = k2 }, k1 , k2 ≥ 0,
W k1 ,k2 (x, s, t)dx = P r{Y (t) = 5, x < B30 (t) ≤ x + dx, N1 (t) = k1 , N2 (t) = k2 }, k1 , k2 ≥ 0,
Rk1 ,k2 (x, s, t)dx = P r{Y (t) = 6, x < R0 (t) ≤ x + dx, N1 (t) = k1 , N2 (t) = k2 }, k1 , k2 ≥ 0, .

Further it is assumed that,

(1) (2)
Pk1 ,k2 (0) = Pk1 ,k2 (0) = Vk1 ,k2 (0) = Wk1 ,k2 (0) = Rk1 ,k2 (0) = I0,k2 (0) = 0; k1 , k2 ≥ 0 and I0,0 (0) = 1.

3
2.2 Equations Governing the System
The Kolmogorov differential-difference equations for the preceding model:
∂ ∂ (1) (1)
( + + λ1 + λ2 + α + ξ + µ1 (x))Pk1 ,k2 (x, t) = (1 − δk1 0 )λ1 Pk1 −1,k2 (x, t)
∂t ∂x
(1) (1)
+ (1 − δ0k2 )λ2 Pk1 ,k2 −1 (x, t) + ξPk1 ,k2 +1 (x, t); k1 , k2 ≥ 0, (1)
∂ ∂ (2) (2)
( + + λ1 + λ2 + α + ξ + µ2 (x))Pk1 ,k2 (x, t) = (1 − δk1 0 )λ1 Pk1 −1,k2 (x, t)
∂t ∂x
(2) (2)
+ (1 − δ0k2 )λ2 Pk1 ,k2 −1 (x, t) + ξPk1 ,k2 +1 (x, t); k1 , k2 ≥ 0, (2)
∂ ∂
( + + λ1 + λ2 + ξ + β(x))Vk1 ,k2 (x, t) = (1 − δk1 0 )λ1 Vk1 −1,k2 (x, t)
∂t ∂x
+ (1 − δ0k2 )λ2 Vk1 ,k2 −1 (x, t) + ξVk1 ,k2 +1 (x, t); k1 , k2 ≥ 0, (3)
∂ ∂
( + + λ1 + λ2 + ξ + γ(x))Rk1 ,k2 (x, t) = (1 − δk1 0 )λ1 Rk1 −1,k2 (x, t)
∂t ∂x
+ (1 − δ0k2 )λ2 Rk1 ,k2 −1 (x, t) + ξRk1 ,k2 +1 (x, t); k1 , k2 ≥ 0, (4)
∂ ∂
( + + λ1 + λ2 + ξ + µ3 (x))Wk1 ,k2 (x, t) = (1 − δk1 0 )λ1 Wk1 −1,k2 (x, t)
∂t ∂x
+ (1 − δ0k2 )λ2 Wk1 ,k2 −1 (x, t) + ξWk1 ,k2 +1 (x, t); k1 , k2 ≥ 0, (5)
Z ∞ Z ∞
d
( + λ1 + λ2 )I0,0 (t) = R0,0 (x, t) γ(x)dx + V0,0 (x, t) β(x)dx
dt 0 0
Z ∞ Z ∞
1 2
+ (1 − θ){ P0,0 (x, t) µ1 (x)dx + P0,0 (x, t) µ2 (x)dx}, (6)
0 0
∂ ∂
( + + λ1 + λ2 + ξ + η(x) )I0,k2 (x, t) = 0; k2 ≥ 1. (7)
∂t ∂x
with the boundary conditions
Z ∞ Z ∞
I0,k2 (0, t) = R0,k2 (x, t) γ(x)dx + V0,k2 (x, t) β(x)dx
0 0
Z ∞ Z ∞
1 2
+ (1 − θ){ P0,k2 (x, t) µ1 (x)dx + P0,k 2
(x, t) µ2 (x)dx}, (8)
0 0
Z ∞ Z ∞
(1) (1) (2)
Pk1 ,k2 (0, t) = Pk1 +1,k2 (x, t) µ1 (x)dx + Pk1 +1,k2 (x, t) µ2 (x)dx
Z ∞0 Z ∞0
+ Rk1 +1,k2 (x, t) γ(x)dx + Vk1 +1,k2 (x, t) β(x)dx; k1 , k2 ≥ 0, (9)
0 0
Z ∞
(2)
P0,k2 (0, t) = I0,k2 +1 (x, t) η(x)dx + λ2 I0,k2 (x, t); k2 ≥ 0, (10)
0
 ∞ (1)
Z Z ∞
(2)
V0,k2 (0, t) = θ P0,k2 (x, t)µ1 (x)dx + P0,k2 (x, t)µ2 (x)dx ; k2 ≥ 0, (11)
0
Z0 ∞
 ∞ (1)
Z
(2)
Wk1 ,k2 (0, t) = α Pk1 ,k2 (x, t)(x)dx + Pk1 ,k2 (x, t)dx ; k1 , k2 ≥ 0, (12)
Z ∞0 0

Rk1 ,k2 (0, t) = Wk1 ,k2 (x, t) µ3 (x)dx; k1 , k2 ≥ 0. (13)

0

4
The Probability Generating Function (PGF) of this model:
∞
X ∞ X
X ∞
(2)
I0 (x, z2 , t) = z2k2 I0,k2 (x, t), P (2) (x, z1 , z2 , t) = z1k1 z2k2 Pk1 ,k2 (x, t),
k2 =1 k1 =0 k2 =0
∞ X
X ∞ ∞ X
X ∞
(1)
R(x, z1 , z2 , t) = z1k1 z2k2 Rk1 ,k2 (x, t), P (1) (x, z1 , z2 , t) = z1k1 z2k2 Pk1 ,k2 (x, t),
k1 =0 k2 =0 k1 =0 k2 =0
X ∞ X ∞ ∞ X
X ∞
W (x, z1 , z2 , t) = z1k1 z2k2 Wk1 ,k2 (x, t) V (x, z1 , z2 , t) = z1k1 z2k2 Vk1 ,k2 (x, t),
k1 =0 k2 =0 k1 =0 k2 =0
and I0,0 = 1.
Taking Laplace transform for the equations (1) - (13) and applying probability generating function
for the equations we get,
∂ 1 (1)
+ (s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + α + ξ(1 − ) + µ1 (x)) P (x, s, z1 , z2 ) = 0, (14)
∂x z2
∂ 1 (2)
+ (s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + α + ξ(1 − ) + µ2 (x)) P (x, s, z1 , z2 ) = 0, (15)
∂x z2
∂ 1
+ (s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + ξ(1 − ) + β(x)) V (x, s, z1 , z2 ) = 0, (16)
∂x z2
∂ 1
+ (s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + ξ(1 − ) + µ3 (x)) W (x, s, z1 , z2 ) = 0, (17)
∂x z2
∂ 1
+ (s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + ξ(1 − ) + γ(x)) V (x, s, z1 , z2 ) = 0, (18)
∂x z2
∂
+ (s + λ1 + λ2 + η(x)) I 0 (x, s, z2 ) = 0. (19)
∂x
Solving equations from (14) to (19),
Z x

(1) (1)
{−ϕ1 (s, z1 , z2 )x − µ1 (t)dt}
P (x, s, z1 , z2 ) = P (0, s, z1 , z2 )e 0 , (20)
Z x
(2) (2)
{−ϕ1 (s, z1 , z2 )x − µ2 (t)dt}
P (x, s, z1 , z2 ) = P 0 (0, s, z2 )e 0 , (21)
Z x
{−ϕ3 (s, z1 , z2 )x − β(t)dt}
V (x, s, z1 , z2 ) = V (0, s, z2 )e 0 , (22)
Z x
{−ϕ2 (s, z1 , z2 )x − µ3 (t)dt}
W (x, s, z1 , z2 ) = W (0, s, z1 , z2 )e 0 , (23)
Z x
{−ϕ2 (s, z1 , z2 )x − β(t)dt}
R(x, s, z1 , z2 ) = R(0, s, z1 , z2 )e 0 , (24)
Z x
{−ϕ(a, s)x − η(t)dt}
I 0 (x, s, z2 ) = I 0 (0, s, z2 )e 0 . (25)

5
Integrating equations (20) to (25), by parts with respect to x yields,

(1) (1) 1 − B 1 [ϕ1 (s, z1 , z2 )]

P (s, z1 , z2 ) = P (0, s, z1 , z2 )[ ], (26)
ϕ1 (s, z1 , z2
(2) (2) 1 − B 2 [ϕ1 (s, z1 , z2 )]
P (s, z1 , z2 ) = P 0 (0, s, z2 )[ ], (27)
ϕ1 (s, z1 , z2
1 − V [ϕ2 (s, z1 , z2 )]
V (s, z1 , z2 ) = V 0 (0, s, z2 ) [ ], (28)
ϕ2 (s, z1 , z2 )
1 − B 3 [ϕ2 (s, z1 , z2 )]
W (s, z1 , z2 ) = W (0, s, z1 , z2 ) [ ], (29)
ϕ2 (s, z1 , z2 )
1 − R[ϕ2 (s, z1 , z2 )]
R(s, z1 , z2 ) = R(0, s, z1 , z2 ) [ ], (30)
ϕ5 (s, z1 , z2 )
1 − I[ϕ(a, s)]
I 0 (s, z2 ) = I 0 (0, s, z2 ) [ ]. (31)
ϕ(a, s)

Multiplying equations from (20) to (25) by, µ1 (x), µ2 (x), β(x), µ3 (x), γ(x) and η(x) respectively
and integrating with respect to x, we get,
Z ∞
(1) (1)
P (x, s, z1 , z2 ) µ1 (x)dx = P (0, s, z1 , z2 )B 1 [ϕ1 (s, z1 , z2 )], (32)
0
Z ∞
(2) (2)
P (x, s, z1 , , z2 ) µ2 (x)dx = P 0 (0, s, z2 )B 2 [ϕ1 (s, z1 , z2 )], (33)
0
Z ∞
V (x, s, z1 , z2 ) β(x)dx = V 0 (0, s, z2 )V [ϕ2 (s, z1 , z2 )], (34)
0
Z ∞
W (x, s, z1 , z2 ) µ3 (x)dx = W (0, s, z1 , z2 )B 3 [ϕ2 (s, z1 , z2 )], (35)
Z0 ∞
R(x, s, z1 , z2 ) γ(x)dx = R(0, s, z1 , z2 )R[ϕ2 (s, z1 , z2 )], (36)
Z0 ∞
I 0 (x, s, z2 ) = I 0 (0, s, z2 ) I[ϕ(a, s)]. (37)
0

where,
ϕ(a, s) = s + λ1 + λ2 , ϕ1 (s, z1 , z2 ) = s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + α + ξ[1 − 1
z2
],
ϕ2 (s, z1 , z2 ) = s + λ1 [1 − z1 ] + λ2 [1 − z2 ] + ξ[1 − z12 ],

Simplifying the boundary conditions we can get,

(1) (1) (1)
z1 P (0, s, z1 , z2 ) = B 1 [ϕ1 (s, z1 , z2 )]P (0, s, z1 , z2 ) + λ1 z1 I 0 (x, s, z2 ) − P 0 (0, s, z2 )
(2) (1)
× B 1 [ϕ1 (s, z2 )] + P 0 (0, s, z2 )B 2 [ϕ1 (s, z1 , z2 )] − P 0 (0, s, z2 )B 2 [ϕ1 (s, z2 )]
+ R(0, s, z1 , z2 )R[ϕ2 (s, z1 , z2 )] − R0 (0, s, z2 )R[ϕ2 (s, z2 )]+
× V (0, s, z1 , z2 )V [ϕ2 (s, z1 , z2 )] − V 0 (0, s, z2 )V [ϕ2 (s, z2 )], (38)

6
 Z ∞ Z ∞
I 0 (0, s, z2 ) = 1 − (s + λ1 + λ2 )I 0,0 (s) + V 0 (x, s, z2 )β(x)dx + R0 (x, s, z2 )γ(x)dx
0 0
Z ∞ Z ∞
 1 2
+ (1 − θ) P 0 (x, s, z2 )µ1 (x)dx + P 0 (x, s, z2 )µ2 (x)dx ; k2 ≥ 1, (39)
0 0
(2) 1 − I[ϕ(a, s)]
z2 P 0 (0, s, z2 ) = I 0 (0, s, z2 )[I[ϕ(a, s)] + λ2 z2 [ ]], (40)
ϕ(a, s)
 (1) (2)
V (0, s, z1 , z2 ) = θ P (0, s, z2 )B 1 [ϕ1 (s, z2 )] + P 0 (0, s, z2 )B 2 [ϕ2 (s, z2 )] , (41)
 (1) [1 − B 1 [ϕ1 (s, z1 , z2 )]] (2)
W 0 (0, s, z1 , z2 ) = α P (0, s, z1 , z2 ) + P (0, s, z1 , z2 )B 2 [ϕ2 (s, z1 , z2 )] ,
ϕ1 (s, z1 , z2 )
(42)
 (1) [1 − B 1 [ϕ1 (s, z1 , z2 )]] (2)
R(0, s, z1 , z2 ) = α P (0, s, z1 , z2 ) + P (0, s, z1 , z2 )B 2 [ϕ2 (s, z1 , z2 )] B 3 [ϕ2 (s, z1 , z2 )]
ϕ1 (s, z1 , z2 )
(43)
(1)
By applying Rouche’s theorem on (38), we get, P 0 (0, s, z2 )B 1 [ϕ1 (s, z2 )] =
(2)
  

 λ1 g(z2 )I 0 (x, s, z2 ) + P 0 (0, s, z2 ) B 2 [ϕ1 (s, g(z2 ))] − B 2 [ϕ1 (s, z2 )] 

 


 1 − B 2 [ϕ 1 (s, g(z 2 ))] 

+ α B [ϕ (s, g(z ))]R[ϕ (s, g(z ))]
 
3 1 2 2 2
 
ϕ1 (s, g(z2 ))
 
 1 − B 2 [ϕ1 (s, z2 )] 
− α B [ϕ (s, z )]R[ϕ (s, z )] + θB [ϕ (s, z )]

 

 3 1 2 2 2 2 1 2 



 ϕ 1 (s, z 2 ) 




× [V [ϕ2 (s, g(z2 ))] − V [ϕ2 (s, z2 )]]
 
 . (44)

 B [ϕ (s, z )]{1 − θV [ϕ (s, g(z ))] + θV [ϕ (s, z )]} + α 1 − B 2 [ϕ 1 (s, z2 )] 

1 1 2 2 2 2 2
ϕ1 (s, z2 )

 × B [ϕ (s, z )]R[ϕ (s, z )] 

3 1 2 2 2

By substituting (??) in required equations we get,

 
1 − I 0 (0, s, z 2 )  
λ1 [ ] [ζ 4 (s, z1 , z2 ) − ζ 1 (s, z1 , z2 )](1 − (s + λ1 + λ2 )I 0,0 (s)) 


 


 φ(a, s) 


 
 1 − I 0 (0, s, z2 )
 
(2)
 
+ P 0 (0, s, z2 ) λ1 [ ][ζ 3 (s, z1 , z2 )ζ 5 (s, z1 , z2 ) + ζ 4 (s, z1 , z2 )

 φ(a, s) 

 
× ζ (s, z , z ) − ζ (s, z , z )ζ (s, z , z ) − ζ (s, z , z )ζ (s, z , z )]

 


 6 1 2 1 1 2 6 1 2 2 1 2 5 1 2 


 

(1) + ζ 1 (s, z1 , z2 )ζ 3 (s, z1 , z2 ) + ζ 2 (s, z1 , z2 )ζ 4 (s, z1 , z2 )
P (0, s, z1 , z2 ) =   ,
 1 − I 0 (0, s, z2 ) 
 ζ 4 (s, z1 , z2 ) − λ1 [ ] z1 − B 1 [ϕ1 (s, z1 , z2 )] 

 


φ(a, s)

 1 − B 1 [ϕ1 (s, z1 , z2 )]  
 − α[ ]B 3 [ϕ1 (s, z1 , z2 )]R[ϕ2 (s, z1 , z2 )] 

 
ϕ1 (s, z1 , z2 )
(45)

7
   
 (1 − (s + λ1 + λ2 )I 0,0 (s)) ζ 4 (s, , z1 , z2 )
 
 1 − I[ϕ(a, s)] 
 × I[ϕ(a, s)] + λ2 z2 [
 ] 
(2) [ϕ(a, s)]
P 0 (0, s, z2 ) =  , (46)
 1 − I[ϕ(a, s)] 
z2 ζ 4 (s, z1 , z2 ) − λ1 g(z2 )[ ]ζ 5 (s, z1 , z2 ) 


ϕ(a, s)

 


  

− ζ 3 (s, z1 , z2 ))ζ 5 (s, z1 , z2 ) + ζ 4 (s, z1 , z2 )ζ 6 (s, z1 , z2 )
 
1 − I[ϕ(a, s)]

 

  
 × I[ϕ(a, s)] + λ2 z2 [
 ] 

ϕ(a, s)

( (2)
)
(1 − (s + λ1 + λ2 )I 0,0 (s))ζ 4 (s, z1 , z2 ) + P 0 (0, s, z2 )

× ζ 4 (s, z1 , z2 )ζ 6 (s, z1 , z2 ) + ζ 3 (s, z1 , z2 )ζ 5 (s, z1 , z2 )
I 0 (0, s, z2 ) =   . (47)
1 − I[ϕ(a, s)]
ζ 4 (s, z1 , z2 ) − λ1 (g(z2 ))[ ]ζ 5 (s, z1 , z2 )
ϕ(a, s)

where,
 1 − B 1 [ϕ1 (s, z2 )]
ζ 1 (s, z1 , z2 ) = B 1 [ϕ1 (s, z2 )] 1 − θV [ϕ2 (s, z1 , z2 )] + θV [ϕ2 (s, z2 )] + α[ ]
ϕ1 (s, z2 )
× B 3 [ϕ1 (s, z2 )]R[ϕ2 (s, z2 )],

ζ 2 (s, z1 , z2 ) = B 2 [ϕ1 (s, z1 , z2 )] − B 2 [ϕ1 (s, z2 )]θB 2 [ϕ1 (s, z2 )] V [ϕ2 (s, z1 , z2 )]
1 − B 2 [ϕ1 (s, z1 , z2 )]
− V [ϕ2 (s, z2 )] + α[ ]B 3 [ϕ1 (s, z1 , z2 )]R[ϕ2 (s, z1 , z2 )]
ϕ1 (s, z1 , z2 )
1 − B 2 [ϕ1 (s, z2 )]
− α[ ]B 3 [ϕ1 (s, z2 )]R[ϕ2 (s, z2 )],
ϕ1 (s, z2 )
1 − B 2 [ϕ1 (s, g(z2 ))]
ζ 3 (s, z1 , z2 ) = B 2 [ϕ1 (s, g(z2 ))] − B 2 [ϕ1 (s, z2 )] + α B 3 [ϕ1 (s, g(z2 ))]
ϕ1 (s, g(z2 ))
1 − B 2 [ϕ1 (s, z2 )]
× R[ϕ2 (s, g(z2 ))] − α B 3 [ϕ1 (s, z2 )]R[ϕ2 (s, z2 )]
ϕ1 (s, z2 )
+ θB 2 [ϕ1 (s, z2 )][V [ϕ2 (s, g(z2 ))] − V [ϕ2 (s, z2 )]],
1 − B 2 [ϕ1 (s, z2 )]
ζ 4 (s, z1 , z2 ) = B 1 [ϕ1 (s, z2 )]{1 − θV [ϕ2 (s, g(z2 ))] + θV [ϕ2 (s, z2 )]} + α
ϕ1 (s, z2 )
× B 3 [ϕ2 (s, z2 )]R[ϕ2 (s, z2 )],
 1 − B 1 [ϕ1 (s, z2 )]
ζ 5 (s, z2 ) = B 1 [ϕ1 (s, z2 )] 1 − θ + θV [ϕ2 (s, z2 )] + α B 3 [ϕ1 (s, z2 )]R[ϕ2 (s, z2 )],
ϕ1 (s, z2 )
 1 − B 2 [ϕ1 (s, z2 )]
ζ 6 (s, z2 ) = B 2 [ϕ1 (s, z2 )] 1 − θ + θV [ϕ2 (s, z2 )] + α
ϕ1 (s, z2 )
× B 3 [ϕ1 (s, z2 )]R[ϕ2 (s, z2 )].

8
3 Steady-state Analysis: Limiting Behaviour
By using the property,
lim sf (s) = lim f (t).
s→0 t→∞
to the equations from (26) to (31), (??) and (??), we obtain the steady-state solutions of this model.
The steady-state probability for the high priority and low priority units with vacation, working
breakdown, repair and retrial time are given by
1 − I[ϕ(a)]
I0 (z2 ) = I0 (0, z2 )[ ], (48)
ϕ(a)
1 − B 1 [ϕ1 (z1 , z2 )]
P (1) (z1 , z2 ) = P (1) (0, z1 , z2 )[ ], (49)
ϕ1 (z1 , z2 )
(2) 1 − V [ϕ2 (z1 , z2 )]
V (z1 , z2 ) = θ P (1) (0, z2 )B 1 [ϕ1 (z2 )] + P0 (0, z2 )B 2 [ϕ2 (z2 )] [

], (50)
ϕ2 (z1 , z2 )
(2) 1 − B 2 [ϕ1 (z1 , z2 )]
P (2) (z1 , z2 ) = P0 (0, z2 )[ ], (51)
ϕ1 (z1 , z2 )
1 − B 1 (ϕ1 (z1 , z2 )) 1 − B 2 [ϕ1 (z1 , z2 )]
W (z1 , z2 ) = α P (1) (z1 , z2 )[ ] + P (2) (z1 , z2 )[

]
ϕ1 (z1 , z2 ) ϕ1 (z1 , z2 )
1 − B 3 [ϕ2 (z1 , z2 )]
×[ ], (52)
ϕ2 (z1 , z2 )
1 − B 1 (ϕ1 (z1 , z2 )) 1 − B 2 [ϕ1 (z1 , z2 )]
R(z1 , z2 ) = α P (1) (z1 , z2 )[ ] + P (2) (z1 , z2 )[

]
ϕ1 (z1 , z2 ) ϕ1 (z1 , z2 )
1 − R[ϕ2 (z1 , z2 )]
× B 3 [ϕ2 (z1 , z2 )][ ]. (53)
ϕ2 (z1 , z2 )
Let Pq (z1 , z2 ) be the probability generating function of the queue size irrespective of the state of the
system and combining equations from (48) to (53), we obtain,
Pq (z1 , z2 ) = P (1) (z1 , z2 ) + V (z1 , z2 ) + P (2) (z1 , z2 ) + W (z1 , z2 ) + R(z1 , z2 ) + I0 (z2 ) (54)
Using the normalization condition,
Pq (1, 1) + I0,0 = 1.
we obtain I0,0 as,

α(λ1 + λ2 )(λ1 + λ2 − ξ)ζ4 (1, 1)
I0,0 = . (55)
Dr
And the utilization factor is,
 (2)   

 P 0 (0, 1) (B 2 (α)) (λ1 + λ2 − ξ) + α(E(W ) + E(R)) ζ 4 (1, 1)(λ1 + λ )
2 
 
 + αθE(V )(λ + λ )B (λ + α)ζ (1, 1) + B (λ + α)ζ (1, 1) − 4αθξ 
 
1 2 2 1 4 1 1 3


 × I0 (0, 1)[1 − I[ϕ(a)]]B 1 (λ1 + α)E(V 2 ) + P (1) (0, 1, 1)(1 − B 1 (α)) 


  
× ζ4 (1, 1)(λ1 + λ2 ) (λ1 + λ2 − ξ) + α(E(W ) + E(R))
 
ρ= , (56)
Dr

9
where
(2)  
Dr = α(λ1 + λ2 )(λ1 + λ2 − ξ)ζ4 (1, 1) + P0 (0, 1) (B 2 (α)) (λ1 + λ2 − ξ)

+ α(E(W ) + E(R)) ζ4 (1, 1)(λ1 + λ2 ) + αθE(V )(λ1 + λ2 ) B 2 (λ1 + α)

× ζ4 (1, 1) + B 1 (λ1 + α)ζ3 (1, 1) − 4αθξI0 (0, 1)[1 − I[ϕ(a)]]
× B 1 (λ1 + α)E(V 2 ) + P (1) (0, 1, 1)(1 − B 1 (α))ζ4 (1, 1)(λ1 + λ2 )

× (λ1 + λ2 − ξ) + α(E(W ) + E(R)) .

and ρ < 1 is the stability condition under which steady state exists, for the model studied.

4 Conclusion
We have analysed an M1 , M2 /G1 , G2 /1 retrial queue with constant retrial policy and non-pre-
emptive priority services under modified Bernoulli vacation subject to working breakdown and
repair process are also investigated. In addition, the effect of impatient behaviour of the unit on a
service system is studied. The joint steady-state probability generating functions of the server state
and the stability condition are derived.

References
[1] Bo Keun Kim, Doo Ho Lee (2016), The M/G/1 queue with disasters and working breakdowns,
Applied Mathematical Modelling, Vol. 4, 437 − 459.

[2] Cheng-Dar Liou (2013), Markovian queue optimisation analysis with an unreliable server
subject to working breakdowns and impatient customers, International Journal of Systems
Science ,Vol. 46(12), 2165 − 2182.

[3] Dong-Yuh Yang, Ying-Yi Wu (2017), Analysis of a finite-capacity system with working
breakdowns and retention of impatient customers, Journal of Manufacturing Systems, Vol. 44
207 − 216.

[4] Kailash C. Madan (2011), A Non-Preemptive Priority Queueing System with a Single Server
Serving Two Queues M/G/1 and M/D/1 with Optional Server Vacations Based on Exhaustive
Service of the Priority Units, Applied Mathematics, Vol. 2, 791-799.

[5] Kalidass K, Kasturi R (2012), A queue with working breakdowns, Computers and Industrial
Engineering, Vol. 63, 779-783.

[6] Boutarfa L, Djellab N (2015), On the performance of theM1 , M2 /G1 , G2 /1 retrial queue with
pre-emptive resume policy, Yugoslav Journal of Operations Research, Vol. 25(1), 153-164.

[7] Phung-Duc, T. (2019), Retrial Queueing Models: A Survey on Theory and Applications,
arXiv 2019, arXiv:1906.09560v1.

10
[8] Tao Li, Liyuan Zhang (2017), An M/G/1 Retrial G-Queue with General Retrial Times and
Working Breakdowns, Math. Comput. Appl. Vol. 22, 15.

[9] Zaiming liu, Yang Song (2014), The M X /M/1 queue with working breakdown, Rairo
operations research, Vol. 48, 339 − 413.

11