Girin
Saikia
Gauhati
University, India
E-mail: girin.stats@gmail.com
Amit
Choudhury
Gauhati
University, India
E-mail: achoudhury@rediffmail.com
Submission: 10/22/2020
Accept: 11/6/2020
ABSTRACT
The phenomena are balking can be said to have been observed when a customer who has arrived into queuing system decides not to join it. Reverse balking is a particular type of balking wherein the probability that a customer will balk goes down as the system size goes up and vice versa. Such behavior can be observed in investment firms (insurance company, Mutual Fund Company, banks etc.). As the number of customers in the firm goes up, it creates trust among potential investors. Fewer customers would like to balk as the number of customers goes up. In this paper, we develop an M/M/1/k queuing system with reverse balking. The steady-state probabilities of the model are obtained and closed forms of expression of a number of performance measures are derived.
Keywords:
Finite capacity; Performance Measure; Queuing Model; Reverse Balking;
Steady-State Solution
1.
INTRODUCTION
In our day-to-day life, we commonly experience waiting lines at bus stops, bank counters, hospitals and the like. A queue or waiting line forms when we need some kind of service and arrive at a service channel that offers such facilities (Sharma, 2013). Queuing is essential to manage congestion in traffic of any type in the modern technological world. Queuing theory describes probabilistically and mathematically, the interaction between the arrival process of customers and the service provided to them in order to manage the system in an efficient manner. It started when Danish mathematician (Erlang, 1909) showed for the first time how probability theory can be used to mathematically model the task of managing telephone conversations.
Among the various aspects in the study of any queuing system, there is the phenomenon of customer attrition. Of late, this has gained importance especially in the cutthroat world of business and commerce. Serving a customer is the whole purpose of any business entity. Firms invest substantial portion of their time, effort, energy and money to attract customers. Consequently, should there be any attrition of customers from a queuing system, it is not only an opportunity lost to expand business (with consequent revenue loss) but the fair name of the firm also takes a beating in the marketplace. All this is detrimental to it interests.
One of the reasons behind attrition is long queues. In any real life queuing system, such long queues discourage customers in two ways. A customer may decide not to join the queue (which is defined as balking) or leave the queue after joining it (which is defined as reneging). In both the cases, the customer does not receive service. We can find such queuing systems almost all around us. In fact, attrition is very common in queues. Examples are firms operating in the domains of health, service systems, call centers just to name a few.
Balking is the subject matter of this paper. We can observe balking when a customer arrives into a queuing system but refuses of join it. Traditionally, balking occurred with longer queues - the higher the size of the queue, the higher was the probability of balking. However, a particular type of balking which has gained acceptability in recent years is reverse balking. It aims to model a phenomena where the probability of balking (of an arriving customer) is low when the queue size is large and vice-versa.
Clearly, this is the reverse of the traditional concept of balking – an antithesis perhaps but there are a number of real like service delivery systems where this phenomenon can be observed. For example, consider a financial entity say an insurance company, Mutual Fund Company, commercial bank etc. If the number of investors is already large, it provides some sort of comfort to potential customers. In such cases, the (large) number of customers does not discourage potential (arriving) investors (customers) arriving into the queuing system.
On the contrary, the large size instills a sense of confidence among them with regard to the financial soundness of the firm (as a large number of customers are already patronizing it, the firm must be financially sound). As a result, the probability of an arriving customer joining the queue goes up (consequently, the probability of balking comes down).
In this paper, we therefore analyze a queuing system with reverse balking. Among the various models available in literature, we choose the single server Markovian queuing system with finite buffer. Symbolically, we denote it by M/M/1/k. We describe the model in a later section.
We arrange the subsequent sections of the paper as follows. After this introduction, section 2 reviews the literature. Section 3 and 4 contain model description and derivation of steady state probabilities. In section 4, we discuss a numerical example. We conclude in section 5. To unclutter the paper, we place the detailed mathematical derivations in Appendix.
2.
LITERATURE SURVEY
There is a considerable body of literature on customer attrition. Among them Haight (1957) can be considered to be the earliest. Haghighi, Medhi & Mohanty (1986) analyzed a multi-server queuing model with balking and reneging. They obtained the steady state distribution of the number of customers in the system. They also obtained an expression for average customer lost during a fixed duration of time.
Choudhury
and Medhi (2011A) considered an M/M/k model with the restriction that customers
may balk from a non-empty queue as well as may renege after they join the
queue. Derivations of closed form expression of a number of performance
measures are given. Choudhury and Medhi (2011B)
analyzed a multi-server Markovian queuing system under the assumption that
customers may balk as well as renege.
Explicit closed form expressions are
given. Demonstration of results derived is through a numerical example with
design connotations. Choudhury and Medhi (2011C) analyzed a single server finite buffer Markovian
queuing model M/M/1/N with the additional restriction that customers may balk
as well as renege.
Choudhury and
Medhi (2012) considered a finite buffer multi
server queuing system with balking along with position dependent reneging.
Derivations of explicit closed form expressions of a number of performance
measures are given. Demonstration of usefulness of results derived is through a
practical problem.
Som and Kumar (2017) considered a single server finite capacity queuing system with customer retention and balking in which the inter-arrival and service times follow negative-exponential distribution. The assumption is that the reneging time is exponentially distributed. The steady state solution of the model is given. Derivation of some performance measures is given. A discussion on sensitivity analysis of the model is also given.
Jain, Kumar and Som (2014) perhaps developed and introduced the concept of reverse balking in a single server Markovian queuing system. Derivation of the steady-state solution of the model and different measures of effectiveness is given. Kumar, Som and Jain (2015) considered a single server finite capacity feedback queuing system with reverse balking. Feedback customer in queuing literature refers to a customer who is unsatisfied with incomplete, partial or unsatisfactory service.
They derived the steady-state solution of the model and obtained some important performance measures. Sasikala and Thaigaranjan (2016) considered the M/M/1/N interdependent queuing model with controllable arrival rates and reverse balking. They derived the steady state solutions along with the system characteristics of the model. Som and Kumar (2017) also considered a finite capacity Markovian queuing system with two heterogeneous servers, reverse balking and reneging.
3.
ASSUMPTIONS OF THE MODEL
In this paper, we shall deal with the M/M/1/k model with reverse balking. Specifically, the assumptions of the model are as follows:
a) Arrivals
follow Poisson probability distribution and the inter-arrival times follow
exponentially distribution with parameter λ.
b) There is only one server and the service times follow exponentially distribution with parameter µ.
c) The system capacity is restricted to k i.e. the capacity of the system is finite. In other words, the maximum number of customers that can be present in the system at any point of time is k.
d) The queue discipline is FCFS (first come first served) i.e. the customer who arrives first into the queuing system will receive service before others.
e) The
probability that an arriving customer will balk is and consequently, the probability that an
arriving customer will join the queuing system (i.e. not balk) is
. Here n is the state of the
system (i.e. the number of customers present in the system) at the instant of
time the test customer arrives into the system. Further, ‘a’ and ‘b’
are constants to be suitably chosen. Since
is a probability, the constants have to be
chosen in such a manner that (
) > 0 (for all n = 1, 2,…
k). Essentially this means that ‘k’ should be less than
. Thus the buffer size should
not exceed the ratio of the constants
and
. At this point, an obvious question is how to
choose
and
. We shall demonstrate the
same through an example in section 4. The probability of balking will increase
as
increases and probability of balking will
decrease as
increases. In other words,
is directly proportion to
probability of balking where as
is inversely proportional to the balking
probability.
4.
THE STEADY STATE PROBABILITIES
In this section, we derive
the steady state probabilities by using the Markov process method. Let denote
the probability that there are ‘n’ customers in the system. The steady
state equations are
µ =
(1)
; n = 1, 2… k-1 (2)
= µ
(3)
Solving recursively, we get
; n = 0, 1, 2… k
(4)
Where
is obtained from the normalizing
condition
and is given as
=
An important performance measure is the average number of customers in the system. It is traditionally denoted by L. We shall derive an expression for the same from the p.g.f. (probability generating function) of the steady state distribution.
Let P(s) be the p.g.f. We recall that
L
= P/(1) = P(s) |s=1.
Under the assumptions of the model, (see appendix 1 for the derivation)
(5)
Substituting s=1 in equation (5), we obtain the average number of customers in the system.
L
= (6)
5.
NUMERICAL EXAMPLE.
To
illustrate the use of our results, we apply them to a queuing problem. We quote
below an example from Choudhury and Medhi (2011A).
Traffic to a message switching
Centre for Extraterrestrial Communications Corporation arrives in a random
pattern (remember that ‘random pattern’ means exponential inter-arrival time)
at an average rate of 240 messages per minute. The line has a transmission rate
of 800 characters per second. The message length distribution (including
control characters) is approximately exponential with an average length of 176
characters. Calculate the principal statistical measure of system performance
assuming that a very large number message buffers is provide. Suppose, however
that it is desired to provide only the minimum number of messages buffers
required to guarantee that
How many buffers should be provided?
This is a design problem. To be specific, we are required to determine the system capacity k (i.e. size of the buffer) in such a manner that the boundary condition pk < 0.005 is satisfied. It is clear that we would like k to be as small as possible (so as to keep costs down – an additional buffer would mean additional expenditure). We shall therefore first check if k=1 enables us to attain the boundary condition. Then we shall examine for k=2 and so forth. An additional issue is to fix the values of the constants ‘a’ and ‘b’ for which we shall examine various alternatives.
A
reading of the problem statement tells us that the arrival rate is 4 per second
and the service rate is 4.55 per second. Hence here, λ = 4/s and µ =
4.55/s. We analyze the problem for different choices of ‘’ and ‘
’. In particular, we consider the following alternatives –
1.
= 5 &
=1 4.
= 10 &
=3
2.
= 10 &
=1 5.
= 40 &
=1
3.
= 10 &
=2 6.
= 40 &
=19
In what follows, we shall analyze different scenarios for different values of k and different alternatives of ‘a’ and ‘b’.
Table
1: Steady state probabilities for varying system capacity with = 5
&
=1
|
Probabilities |
System Capacity |
|||
|
k=1 |
k=2 |
k=3 |
k=4 |
|
|
|
0.85047 |
0.82341 |
0.81580 |
0.81250 |
|
|
0.14953 |
0.14477 |
0.14344 |
0.14286 |
|
|
undefined |
0.03182 |
0.03152 |
0.03140 |
|
|
undefined |
undefined |
0.00924 |
0.00920 |
|
|
undefined |
undefined |
undefined |
0.00404 |
|
Total |
1.00000 |
1.00000 |
1.00000 |
1.00000 |
Table 1 displays a few
steady state probabilities assuming a=5 & b=1. For k=1, p1 =0.14953
> 0.005, for k=2, and for k=3,
. These do not satisfy the
given design condition. On the other hand k=4
satisfies the given design restriction of
Therefore the number of buffers should
be equal to 4. We need not check beyond k=4 because of our restriction
that
<
(in our case
). Thus if we choose a=5 & b=1, then the optimum choice of k is 4.
Table
2: Steady state probabilities for varying system capacity with = 10
&
=1
|
Probabilities |
System Capacity |
||
|
k=1 |
k=2 |
k=3 |
|
|
|
0.91919 |
0.91199 |
0.91121 |
|
|
0.08081 |
0.08018 |
0.08011 |
|
|
undefined |
0.00783 |
0.00782 |
|
|
undefined |
Undefined |
0.00086 |
|
Total |
1.00000 |
1.00000 |
1.00000 |
Table 2 displays a few
steady state probabilities assuming a=10 & b=1. For k=1, p1 =0.08081
> 0.005, for k=2, p2=0.00783 > 0.005. These do not satisfy the given design condition. On the other hand k=3 satisfies the given design
restriction of Therefore the number of buffers should
be equal to 3. We need not check beyond k=3 because in general, it is better to
have as few buffers as possible. Thus if we choose a=10 & b=1, then optimum
choice of k is 3.
Table
3: Steady state probabilities for varying system capacity with = 10
&