See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267176376

M/M/C queueing model for waiting time of customers in bank sectors

Article  in  International Journal of Mathematical Sciences & Applications · January 2011

CITATIONS READS
6 8,856

2 authors, including:

Santhanagopalan Srinivasan
B. S. Abdur Rahman University
40 PUBLICATIONS   130 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Santhanagopalan Srinivasan on 23 January 2015.

The user has requested enhancement of the downloaded file.

Int. J. of Mathematical Sciences and Applications,
Vol. 1, No. 3, September 2011
Copyright  Mind Reader Publications
www.journalshub.com

M/M/C Queueing Model For Waiting Time of Customers In Bank Sectors

Mrs. S.Maragatha Sundari Dr. S.Srinivasan

Asst Prof. ,Dept. of Mathematics Professor& HOD,
DMI College of Engineering,Chennai Dept. of Mathematics
maragatham01@gmail.com BSA University, Chennai
Srinivasan@bsauniv.ac.in

Abstract

In this paper, the highly suitable modeling tool for MMC queueing model, the stochastic
Birth-death Markov process is used. The Model M/M/C is a multi channel queueing
system with poisson arrival and exponential distribution. Queue discipline is FCFS. This
model is applied to HDFC Bank Chennai. All the characters of the model (ie) Number of
Customers in the system, queue, waiting time, service time are calculated. The traffic
intensity 0.17 indicates that the utilization factor is less than the probability of Queuing.

Keywords : - Poisson arrival rate, experimental service rate, utilization factor, Birth-death Markov
process.

Introduction
A perfect banking system motivates and invites the investors within and outside the country. It generates
more external reserves to boost a nation’s economy. There are some avoidable problems that are militating
against the work of the Bank sector. The frequent problem is the problem of waiting lines. Queue causes
unnecessary delay and reduce the service effectiveness. Application of the theory of queue as the
development of mathematical models to study and analyse waiting lines with the hope of reducing this
social factor in our Banking systems will lead to improve the services. The end result would be a increase
in number of customers, more profits, investment which lead directly or indirectly to improve the country’s
economy.

Queueing theory
In 1908, Copenhagen Telephone Company requested Agner K. Erlang to work on the holding times in a
telephone switch. He identified that the number of telephone conversations and telephone holding time fit
into poisson distribution and exponentially distributed. This was the beginning of the study of queueing
theory.
Objectives
M/M/C queuing model to waiting line in HDFC Bank, Chennai in applied. The following are determined.
1. Number of Customers in the system and queue.
2. Waiting time of the Customers spend in the system and the queue.
3. Arrival and service rate per unit time of customers.
4. Traffic intensity.

Datas for the Research.

Datas on arrivals, waiting time, service patterns, departures in HDFC Bank, Chennai was collected for 15
days and used as input data into the M/M/C model to obtain the above objectives. M/M/C queueing system
is involved in this research through Birth – death Marker process. M/M/C Queuing model always exhibits
Markov behaviour which makes the M/M/C system mathematically tractable.

Birth-death Markov processes

Stochastic birth-death Markov processes turns out to be a highly suitable modeling tool for many queueing
processes[5]. Examples of these are any M/M/n queueing models which are our preoccupation in this study with much
emphasis on M/M/C model. Full examination of these models will be considered later in this section. Let

1569
 Mrs. S.Maragatha Sundari & Dr. S.Srinivasan

 N (t ), t  0 , with parameter set T, be an integer-valued continuous-time discrete stochastic process. The

discrete state space of the process comprises (or can assume) non-negative integer values 0,1, 2,......,  . Here
N (t ) is interpreted as the random number of members in some population as a function of time.

In other words N (t ) is viewed as the size of population at time t. By assumption, the classical Markov property
is imposed as a restriction on the process N (t ) , i.e. given the value of N ( s ) , the values of
N ( s  t ) for t  0 are not influenced by the values of N (u ) for u < s. In words, the way in which the
entire past history affects the future of the process is completely summarized in the current state of the
process, In other words; only the present state gives any information of the future behavior of the process,
knowledge of the history of the process does not add new information (The future development of a
continuous-time Markov process depends only on its present state and not on its evolution in the past).
Expressed analytically the Markov property may be defined as thus: A stochastic process
 N (t ), t  0 with set T and discrete state space
is called a continuous-time Markov process (chain) if for any m  1
P[ N (tm 1 )  nm 1 N (tm )  nm ,......N (t1 )  n1 ]  P[ N (tm 1 )  nm 1 N (tm )  nm ] (1)
And it should be valid for all t0  t1  t2  .....  tm  tm 1 and any m
Transition Probabilities
In equation (1), set tm  s, nm  t , tm 1  s  t and nm 1  j then the right-hand side of the equation
expresses the probability that the process makes a transition from state i at time s to state j in time t relative to s. Such a
probability, denoted by Pi , j ( s, t ), is referred to as a state transition probability for the Markov process. In this
research work, we are only concerned with transition probabilities being independent of absolute times, i.e. for all
s  0s>0
j
we have: Pi,j(s,t) = Pi,j(t) = P[N(t) =  i ] = P [N(s+t) =j /N(s) = i] (2)
N (0)
This is called time-homogeneous or stationary transition probabilities. In other words, the intensity of leaving a state is
constant in time.
Definition: Let  N (t ), t  0 be a discrete Markov process. If the conditional probabilities
P  N ( s  t )  j N ( s )  i  for s, t  0 , do not depend on s, the process is said to be time homogeneous. Then
we define the transition probability Pi , j (t )  P  N ( s  t )  j N (0)  i  and the transition matrix P (t ) ,
whose element with index (i, j ) is Pi , j (t )
. Pu (0)  1 and Pij (0)  0 for i  j so that P (t )  1

For a Markov process with time-homogeneous transition probabilities the so called Chapman-Kolmogorov equation
implies.
Pij (t  s )   k 0 Pik (t ) Pkj ( s )

(3)
This equation states that in order to move from state i to j in time (t  s ) , the queue size process N (t ) moves
to some intermediate state k in time t and then from k to j in the remaining time s. It also says how to
compute the long -interval transition probability from a sum of short - interval transition probability components.
An infinitesimal transition probability, denoted by Pij ( t ) , specifies the immediate probabilistic behavior of a
Markov process in that (t )  0 .
GENERALIZED MARKOV BlRTH-DEATH PROCESS

1570
 M/M/C Queueing Model For Waiting Time of Customers In Bank Sectors

A birth-death Markov process is characterized by the fact that the discrete state variable changes by at most one,if it
changes at all, during an infinitely small time interval. The generalized Markov birth-death process thus
satisfies the following criteria:
1. The probability distributions governing the numbers of births and deaths in a specific time interval
depends on the length of the interval but not on its starting point.
2. The probability of exactly one birth in a small time interval, t given that the population size at time t
is n is n t  0(t ) , where n is a constant,
3. The probability of one death in small time interval t given that the population size at time t is n is
μn∆t + 0( t ) where μn is a constant.
4. The probability of more than one birth and the probability of more than one death in a small time
interval t are both 0( t ) .
Reflecting these facts, the following postulations specify the transition kernel of a general birth-death Markov
process:
P[N(t+∆t) = n+1/N(t) = n] = n t  0( t ) , n  0
P[N(t+∆t) = n-1/N(t) = n] = μn∆t + 0( t ) , n  1 (4)
P  N (t  t )  n N (t )  n  1  (n  n )t  0(t ), n  1
P  N (t  t )  k N (t )  n  0(t ), k n  2

Here 0(t ) is a quantity such that lim 0(t )  0 . The first equation handles the case when the state
t 

variable increases by one i.e. N (t )  n  1 . This is referred to as single birth. Here n is proportionality
constant such that the product n t should reflect the probability for a single birth to happen during the
infinitesimal time interval (t , t  t ) . It is customary to interpret n as the instantaneous birth rate. Likewise, the
second equation is for the case when the state variable is reduced by one i.e. N (t )  n  1 . This is referred to as
single death. The product n t signifies the probability that a single death takes place. n denotes the
instantaneous death rate. The third equation handles the case when the state variable does not change i.e.
1  (n  n )t reflects the probability that neither a single birth nor single death occur, i.e. N (t )  n ,
during the infinitely small time interval. Multiple births, multiple deaths and simultaneous births and deaths are
taken care of by the 0( t ) terms in the equations. This should be interpreted such that the probability for
these events to happen is negligible as t  0 , we say that multiple events are prohibited.
We should note that the transition probabilities from (4) are in general state dependent. This is so since the
instantaneous births rate n and also the death rate n may depend on the departing state n . A small
comment also applies to the second and third equations. Since no death can occur if the state variable is already
zero i.e. if n  0, we always define 0  0
By combining equation (3) with the infinitesimal transition probabilities from (4) and using the notation
Pn (t )  P[ N (t )  n], we can write:
Pn (t  t )   k 0 P[ N (t  t )  n N (t )  k ]P[ N (t )k ]

(Bayes'formula) (5)

Pn (t )[1  (n   n )t  0(t )]  Pn 1 (t )[n 1 (t )  0(t )  Pn 1 (t )[ n 1 (t )  0(t )]  0(t )
By rearranging terms and dividing by t , we have
Pn (t  t )  Pn (t )
 (n  n ) Pn (t )  n 1 Pn 1 (t )  n 1 Pn 1 (t ) (6)
t
Taking limit as t  0 , we obtain the differential equation:

1571
 Mrs. S.Maragatha Sundari & Dr. S.Srinivasan

dPn (t )
Pn' (t )   (n  n ) Pn (t )  n 1 Pn 1 (t )  n 1 Pn 1 (t ), n  1, 2,3,....... (7)
dt

While (7) holds for n  1, 2,3.......... we also require an equation for n  0. By following a logical
argument as above, we can write dpn(t)/dt = -λ0P0(t) + μ1P1(t)
Equation (7) is the general model equation for a birth-death Markov process and it essentially captures the
probabilistic dynamics of the process.
Steady State Solution
We now examine the process when it is in equilibrium (steady state). Under proper conditions, such
equilibrium will be reached after the system has been operating for some time. Equilibrium in turn, implies that the
state probabilities Pn (t ) eventually become independent of t and approach a set of constant values (if it
exists) which is denoted by
Pn , n  0,1, 2...... as t  where Pn  lim Pn (t ). This can be interpreted as
t 
steady state probability that there are
n users in the system.
Also under these circumstances in the steady state, lim Pn (t )  0
'
t 
Given the above, (7) and (8) are then transformed to:
0  (n   n ) Pn  n 1 Pn 1n 1 Pn 1 , n  1, 2,3........... (10)
(n   n ) Pn  n 1 Pn 1  n 1 Pn 1
And 0  0 P0  1 P1
or 0 P0  1 P1 (11)
Equations (10) and (11) are called the equilibrium equations or balance equations of birth and death Markov
process.
Analysis of M/M/C Queueing Model
M/M/C is a multi server system with customers arrival follows a poisson process and exponential service
time. When a customer enters an Empty system, he gets the service at once. If the system is non-empty the
incoming customer joins the Queue. M/M/C model is a poisson birth death process[1].
Birth Occurs  Customer arrives
Death Occurs  Customer departs
Both are modeled as Memoryless Markov process.
In M/M/C , M refers to this memoryless/Morkov feature of the arrival and service.

Theorem :
The process  N (t ), t  0 is a birth and death process with birth rate n   for all n  0 and with
death rate n   for all n  1 [7]
Proof:
Because of the exponential distribution of the inter-arrival times and of the service times, it should be clear
that  N (t ), t  0 is a Markov process. On the other hand, since the probability of having two events
(departures, arrivals) in the interval of time (t , t  t ) is 0( t ) , which is characterized by the fact that the
discrete state variable changes by at most one, if it changes at all, during an infinitely small time interval
(criteria for the generalized Markov birth-death process), we have the following transitional probabilities:
P[ N (t  t )  n  1 N (t )  n]  t  0(t ), n  0
P[ N (t  t )  n  1 N (t )  n]  t  0(t ), n  1
P[ N (t  t )  n N (t )  n]  1  (   )t  0(t ), n  1
P[ N (t  t )  n N (t )  n]  1  t  0(t ), n  0
P[ N (t  t )  k N (t )  n]  0(t ), k  n  2

1572
 M/M/C Queueing Model For Waiting Time of Customers In Bank Sectors

Here 0(t ) is a quantity such that lim 0(t )  0. This shows that  N (t ), t  0 is a birth and death
t 

process. From (6), (7), and (8) with the fact that n   and n   for all n in the birth and death
Markov process,[2]
The differential difference equations for M/M/C System are
Pn(t+∆t) = -( λ+nμ)Pn(t) ∆t +(n+1) μPn+1(t) ∆t+ λPn-1(t) ∆t+ Pn(t), for 1  n  c
Pn(t+∆t) = -( λ+cμ)Pn(t) ∆t +c μPn+1(t) ∆t+ λPn-1(t) ∆t+ Pn(t), for n  c (Omitting t 2 )
P0(t+∆t) = - λ P0(t) + μP1(t), n=0
Divide by t & as t  0
Pn’(t) = -( λ+nμ)Pn(t) +(n+1) μPn+1(t) + λPn-1(t) , 1  n  c
Pn’(t) = -( λ+cμ)Pn(t) +c μPn+1(t) + λPn-1(t) , n  c
P01 (t )   P0 (t )   P1 (t ) ; n  0

t   ,[8]
In steady state, as
 P0   P1  0, n  0
(  n ) Pn  (n  1)  Pn 1   Pn 1  0, 1  n  c
(  c ) Pn  c Pn 1   Pn 1  0, n  c
Solving the above, we have
n
Pn = P0 , 1 n  c
(n )(n  1)  ........1
n n
= p  P0 , 1  n  c
n! n
0
n!
n
Pn  .P0 , n  c
[(c )(c )........(c )](c )(c  1)  ........(1 )

( n  c ) times

 n
 .P0
c c !  ncc
nc

n
 n n c .P0
 c c!
n
 .P0 , n  c
c !c n  c
Since the sum of the probabilities must equal unity,

We have P
n 0
n 1
1
 c 1 1    n 1   c   c  
From this we obtain the value for P0 as P0          
 n 0 n !    c !      c    
The mean arrival rate must be less than the mean service rate of the system.
   c  1 says that work brought to the system per unit time is strictly smaller than the processing

rate.

1573
 Mrs. S.Maragatha Sundari & Dr. S.Srinivasan

It is to be noted that the queue will be empty if the system is in stable (ie) n = 0
1
 P0  n c
c 1
1    1     c 
    c!     c   
n 0 n !       
M/M/C Model is adopted with poisson arrival and exponential service rate. The discipline rendered here is
FCFS[3]. The main characteristics of thus model are given by
1. The expected number of customers waiting in the queue
 1   c  
Lq      P0
 (c  1)!    (c   ) 
2

2. The expected number of customers in the system is


Ls  Lq 

3. Expected number of customers waiting to be served at any ‘t’ is
c
Lw 
c  
4. The average waiting time of an arrival is
Lq
Wq 

Ls
5. Average time an arrival spends in the system is Ws  .


6. Utilization factor 
c

Results
In HDFC Bank, Chennai, for a duration of 15 days, in the working hours between 9.30 a.m. and 4.30 p.m.
datas regarding arrivals, departures, service patterns are observed. Those datas are collected. Moreover
service time and waiting time are calculated. A total arrival rate of 3384 customers per a total of 7200
minutes and a total service rate of 4317 customers per a total service times of 6214 is found in final
analysis report.

3384
1.   0.47
7200
4317
2.   0.69
6214
3. c  2.76 (c = 4)

4. Utilization factor    0.17  1
c
1
5. P0  n c
c 1
1  1     c 
    c!     c   
n0 n !       

1574
 M/M/C Queueing Model For Waiting Time of Customers In Bank Sectors

1

 1 1 1 3 1 4  2.76 
1  1! (.68)  2! (.68)  3! (.68)   4! (.68)  2.76  0.47 
2

1 1
   0.5067
1.9636  0.0096 1.9732

7. P (Queuing in the system)  1  P0

 1  0.5067
 0.4933
 1  cc
 
8. Lq      P0
 (c  1)!    (c   ) 
2

1 0.3243 
  (.68)4 .5067
 3! 5.2441 
 0.0356  0.0618  0.5067
 0.0011

9. Ls  Lq 

 0.0011  0.68  0.6811
Lq 0.0011
10. Wq    0.0023
 0.47
L 0.6811
11. Ws  s   1.4  1 min
 0.47
Discussion of Results
Utilization factor is greater than probability of Queuing in the system. It clearly indicates that as soon as the
customer, enters the system, he is being served. No need for him to wait in the queue. Multi channel
servicing facility makes the Banking service more efficient .As the serving capacity increases the excessive
waiting of man hours, congestion in banking halls, suffocation due to congestion are eliminated. Moreover
this serving capacity will bring more profit, investments and share holders to the Banks.
More bank branches to be established with multiserver facility across the country. Effective banking
management will lead to a progress.
REFERENCES
1. Churchman, G.W., R.C. Ackoff and F.C. Arnoff. 1957. Introduction to Operations Research. John Willey and
Sons: New York, NY.
2. Ivo Adan and Jacques Resing. 2002. Queueing Theory — Department of Mathematics and
Computing Science, Eindhoven University of Technology, Eindhoven, The Netherlands.
3. Philippe Nain. 1998. Lecture Notes on Basic Elements of Queueing Theory — Application to the
Modelling of Computer System. University of Massachusetts, Abreast, MA.
4. Taha, H.A. 1987. Operations Research: An Introduction, Fourth Edition. Macmillan
Publishing: New York, NY.
5. Bharucha-Reid, A.T. 1988. Elements of the Theory of Markov Processes and their Applications. Dover
Publications, INC Mineola, New York, NY.
6. Famule, Festus Daisi, 2010 Analysis of Queueing model with applications to waiting time, Global journal of
Computer science and technology.
7. Frode B. Nielsen. 1998. Queueing Systems: modelling Analysis and Simulation. Research
Report 259, ISBN 82-7368-185-8, ISS 0806-3036. Department of Informatics, University of Oslo.
8. Richard Bronson and Govindasami Naadimuthu. 1997. Operations Research. Schaum's" Outline Series,
the McGraw-Hill Companies, New York.

1575

View publication stats