QUEUING AND RELIABILITY THEORY

[MATH712]

Dr. Ritu Gupta

Department of Mathematics, Amity University, Uttar Pradesh

[email protected]

MODULE 2: Advanced Queuing Models

Embedded Markov Chain (EMC):
 A stochastic process is said to be embedded in the
continuous time process only when the service of the
customer is considered just after the departure or prior
the arrival. In this process, state parameter is discrete.

2 Lecture-19
M/G/1 Model:
 Consider a queuing system immediately after customer has
departed and service is about to commence on the next
customer in queue with service times assumed to be
independently and identical distributed random variables with
an arbitrary probability distribution.

3 Lecture-19
Notations:
𝐵 𝑡 − CDF of service time
b 𝑡 – pdf of service time
𝜆 - arrival rate of customers ( according to Poisson
process)
X 𝑡𝑖 – the embedded stochastic process, where X denotes
the no. of customers in the system and 𝑡1, 𝑡2, 𝑡3 , … are the
successive times of completion of service.
X 𝑡𝑖 = 𝑋𝑖 ; since the state space is discrete
X 𝑡𝑖 is a Markov process where 𝑋𝑖 denotes the no. of
customers remaining in the system as the 𝑖 𝑡𝑡 customer
departs.
4 Lecture-19
We have
𝑋𝑛 − 1 + 𝐴𝑛+1 ; 𝑋𝑛 ≥ 1
X 𝑡𝑛+1 = 𝑋𝑛+1 =�
𝐴𝑛+1 ; 𝑋𝑛 = 0
where
𝑋𝑛 - no. of customer in the system at the 𝑛𝑡𝑡 departure
point
&
𝐴𝑛+1 - no. of customers who arrived during the service
time 𝑆 (𝑛+1) 𝑜𝑜 𝑡𝑡𝑡 𝑛 + 1 𝑡𝑡 customer.

5 Lecture-19
 We denote 𝑆 (𝑛+1) 𝑎𝑎𝑎 𝐴𝑛+1 by S and A respectively
since 𝑆 (𝑛+1) is independent of previous service times and
𝐴𝑛+1 depends only on S not on the queue or on the time
of service initiation.
∞
𝑃𝑃 𝐴 = 𝑎 = ∫0 𝑃𝑃 𝐴 = 𝑎/ 𝑆 = 𝑡 . 𝑑 𝐵(𝑡) (1)
𝑒 −𝜆𝑡 . (𝜆𝑡)𝑎
and 𝑃𝑃 𝐴 = 𝑎/ 𝑆 = 𝑡 = , 𝑎 = 0,1,2 … (2)
𝑎!
 ∴ 𝑃𝑃 𝑋𝑛+1 = 𝑗/𝑋𝑛 = 𝑖 = 𝑃𝑃 𝐴 = 𝑗 − 𝑖 + 1
∞
𝑒 −𝜆𝑡 . (𝜆𝑡)𝑗−𝑖+1
� 𝑑𝐵 𝑡 , 𝑗 ≥ 𝑖 − 1, 𝑖 ≥ 1
=� (𝑗 − 𝑖 + 1)!
0
0, 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑖. 𝑒. 𝑗 < 𝑖 − 1
6 Lecture-19
EXPECTED VALUE OF MEASURES OF EFFECTIVENESS :-
(POLLACZEK – KHINTCHINE FORMULA )

We have
𝑋 − 1 + 𝐴𝑛+1 ; 𝑋𝑛 ≥ 1
𝑋𝑛+1 = � 𝑛 (3)
𝐴𝑛+1 ; 𝑋𝑛 = 0

We can also write,

𝑋𝑛+1 = 𝑋𝑛 − 𝑈 𝑋𝑛 + 𝐴 4
where U is the unit step function,
1; 𝑋𝑛 > 0
U 𝑋𝑛 = �
0; 𝑋𝑛 = 0
 Since, we are assuming a steady state solution; we can take expected
value of their random variables.
Noting,
E 𝑋𝑛+1 = 𝐸 𝑋𝑛 = 𝐿 𝐷
where 𝐿 𝐷 represents the expected steady state system size at
departure points.

7 Lecture-19
DEPARTURE POINT: - A point of time just after a customer has
completed service.

Taking expectation of eq. (4)

E 𝑋𝑛+1 = 𝐸 𝑋𝑛 − 𝐸 𝑈 𝑋𝑛 + 𝐸 𝐴
⇒ 𝐿 𝐷 = 𝐿 𝐷 − 𝐸 𝑈 𝑋𝑛 + 𝐸 𝐴
⇒ E 𝑈 𝑋𝑛 =𝐸 𝐴
But from eqs (1) and (2),
∞
E 𝑈 𝑋𝑛 = 𝐸 𝐴 = ∫0 𝐸 𝐴 / 𝑆 = 𝑡 . 𝑑𝑑 𝑡
∞ ∞
= ∫0 𝜆𝜆. 𝑑𝑑 𝑡 = 𝜆 ∫0 𝑡. 𝑑𝑑 𝑡
1 1
= 𝜆. 𝐸 𝑆 = 𝜆. = 𝜌 ∵ 𝐸𝑆 =
𝜇 𝜇
⇒ E 𝑈 𝑋𝑛 =𝐸 𝐴 =𝜌 5

8 Lecture-19
 Squaring equation (4) both sides,

(6)
2
Taking expected values of eq. (6) and noting that E 𝑋𝑛+1 = 𝐸 𝑋𝑛2 ,
we have

(7)

9 Lecture-19
Now, E 𝐴2 = 𝑣𝑣𝑣 𝐴 + 𝐸 𝐴 2 = 𝑣𝑣𝑣 𝐴 + 𝜌2 8
where,
var 𝐴 = 𝐸 𝑣𝑣𝑣 𝐴/𝑆 + 𝑣𝑣𝑣 𝐸 𝐴/𝑆
=E 𝜆𝜆 + 𝑣𝑣𝑣 𝜆𝜆 = 𝜆𝜆 𝑆 + 𝜆2 𝑣𝑣𝑣 𝑆
1
= 𝜆. + 𝜆2 . 𝜎𝑠 2
𝜇
1
𝑤𝑤𝑤𝑤𝑤 𝐸 𝑆 = (first moment of
𝜇
service time distribution)
∴ var 𝐴 = 𝜌 + 𝜆2 . 𝜎𝑠 2 (9)
where 𝜎𝑠 2 is the variance of service time distribution using (8)
and (9) in (7), we get

10 Lecture-19
 (10)

Pollaczek Khintchine
Formula

𝜌 2 + 𝜆2 . 𝜎 2
(𝐷) 𝑠
𝐿𝑞 = 𝐿 − 𝜌 = Remark: For M/M/1,
2(1 − 𝜌) 𝟏
put 𝝈𝒔 𝟐 = 𝟐
𝝁
𝐿𝐿 𝝆
𝐿(𝐷)
𝑊𝑊 = Then 𝑳 =
𝑊= 𝜆
𝟏−𝝆
𝜆

11 Lecture-19