410 • Operations Research

The replacement of an individual bulb on failure cost Rs. 1.25. The cost of
group replacement is 80 paise per bulb. Determine the better one among the
chapter
individual and group replacement policies.
(MU, MBA, April 96)
(Ans, End of 3,d week group, replacement is preferable) QUEUEING
10. There are 1000 bulbs in the system. survival rate is given below.

Week D 1 2 3 4
THEORY
Bulbs in operation
at the end of the week 1000 850 500 200 00

The group replacement of 100 bulbs costs Re. 1.00 and individual replace-
ment cost Re. 0.50 per bulb. Suggest a suitable replacement policy.
(Ans: Group replacement policy) 6.1 INTRODUCTION

flow of customers from finite/infinite population towards the service

acility forms a queue (waiting line) on account of lack of capability to
rve them all at a time. In the absence of a perfect balance between the
ice facilities and the customers, waiting time is required either for
service facilities or for the customer's arrival.
The arriving unit that requires some service to be performed is called
~j'lmfel:..JJhe customer may be persons, machines, vehicles, etc. Queue
aiting line) sta or the number of customers waiting to be serviced.
is does not include the eustomer being serviced. The process or sys-
that performs services to the customer is termed by service channel
service facility.

6.2 QUEUEING SYSTEM

queueing system can be completely described by

(i) the input (arrival pattern),
(ii) the service mechanism (service pattern),
(iii) the queue discipline, and
(iv) customer's behaviour.

6.2.1 The Input (arrival pattern)

e input describes the way in which the customess-arrive and join the
stem. Generally, customers arrive in a more or.' less random fashion
ich is not worth making the prediction. Thus the arrival pattern can be
. ribed in terms of probabilities, and consequently the probability dis-
bution for inter-arrival times (the time between two successive arrivals)
t be defined. We deal with those-Queueing system in which the cus-
rs arrive in Poisson fashion. The mean arrival rate is denoted by I..
412 • Operations Research Queueing Theory • 413

16.2.2 The Service Mechanism Transientand steady states A system is said to be in transient
state when its operating characteristics are dependent on time.
This means the arrangement of service facility to serve customers. If A steady state system is the one in which the behaviour of the system
there is infinite number of servers then all the customers are served is independent of time. Let Pn(t) denote the probability that there are n
instantaneously on arrival, and there will be no queue. customers in the system at time t. Then in steady state
If the number of servers is finite then the customers are served accord_
ing to a specific order w.it-hservice time a constant or a random variable. lim PII (t) = Pn (independent of t)
I~-
Distribution of service time which is important in practice is the nega_
d PII (t)
tive exponential distribution. The mean service rate is denoted by m. ~---
dt

16.2.3 The Queue Discipline ~ lim p/(t)

I~-
=0

It is a rule according to which the customers are selected for service Traffic intensity (or utilisation factor) An important mea-
when a queue has been formed. The most common disciplines are sure of a simple queue is its traffic intensity
- First come first served (FCFS) . Mean arrival rate It
- First in first out (FIFO) given by traffic intensity p = MS.
ean ervrce rate /1
- Last in first out (LIFO)
The unit of traffic intensity is Erlang.
- Selection for service in random order (SIRO).
There are various other disciplines according to which a customer is
served in preference over the other. Under priority discipline, the service
16.3 KENDALI:S NOTATION FOR REPRESENTING
is of two types, namely pre-emptive and non-pre-emptive. In pre-emptive
QUEUEING MODELS
system, the high priority customers are given service over the low prior-
Generally, queueing model may be completely specified in the follow-
ity customers; in non-pre-emptive system, a customer of low priority is
ing symbol form (albic): (d/e) where
serviced before a customer of high priority is entertained for service. In
a = probability law for the arrival (inter-arrival) time
the case of parallel channels "fastest server rule" is adopted.
b = probability law according to which the customers are being
served
16.2.4 Customer's Behaviour c = number of channels (or service stations)
d=capacity of the system, i.e., the maximum number allowed in the
The customers generally behave in the following four ways: system (in service and waiting)
e = queue discipline
(i) Balking A customer who leaves the queue because the queue is
too long and he has no time to wait or has no sufficient waiting space. 16.4 CLASSIFICATION OF QUEUEING MODELS
(ii) Reneging This occurs when a waiting customer leaves the The queueing models are classified as follows:
queue due to impatience. Model I: (MIMI!): (00 IFCFS)
This denotes Possion arrival (exponential inter arrival), Poisson depar-
(iii) Priorities. In certain applications some customers are s~ry.ed
ture (exponential service time), Single server, Infinite capacity, and First
before others regardless of their arrival. These customers have prtOnty
come first served service discipline. The letter M is used due to
over others.
Markovian property of exponential process.
Model II: Multiservice Model (MIM/S):(oo/FCFS)
(iv) Jockeying Customers may jockey from one waiting line to
This model takes the number of service channel as S.
another. This is most common in a supermarket.
414 • Operations Research Queueing Theory • 415

Model III: (MlMll): (NIFCFS) In a similar fashion the probability that there will be n unit (ie 17 =0)
In this model, the capacity of the system is limited (finite), say N. in the system at time (1+ £o.t) will be the sum of the fo\1owing two inde-
Obviously, the number of arrivals wi\1 not exceed the number N in any pendent probabilities.
case. (i) Probability (that there is no unit in the system at time t and no arrival
Model IV: (M/M/S): (NIFCFS) in time £o.t)
This model is essentially the same as model II except the mu'ximum =P)t) (I-A£o.t)
number of customers in the system is limited to N, where tN> S). (ii) Probability (that there is one unit in the system at time t, one unit ser-
Modell: (MIMII) (00 IFCFS) (Birth and death model) viced in £o.tand no arrival in £o.t)
To obtain the steady state equations = Pj(t)p £o.t (I - A £o.t)
The probability that there wi\1 be It units (n > 0) in the system at time = P;Ct)p £o.t+ O(£o.t)
(t+ £o.t) may be expressed as the sum of three independent compound Adding these two probabilities,
probabilities by using the fundamental properties of probability, Poisson
P,,(t+ £o.t) =P)t)(l - A£o.t) + PI(t)p£o.t + 0 (£o.t)
arrivals and exponential service times.
The following are the three cases:
poet + £o.t)- PaCt) = _ A P (t) + II P (t)+ 0 (M)
£o.t ovr s r: I /j[
Time t Arrival Service Tillie (t +A. t)
No. of units No. of units
n 0 0 11

n-I 1 0 11

n+1 0 1 dP (t)
II
-'-'
d(t) - = -A P ",....
(t)+ If P I (t)

Now, by adding above three independent compound probabilities, we Under steady state, we have
obtain the probability of n units in the system at time (t+£o.t) O=-AP,,+pPI (2) I
PII(t+£o.t)= PII(t) (J-(A+p) M)+PII_I(t) A£o.t Equations (I) and (2) are called steady state difference equations for
this model.
+ P" + I (t)p £o.t + O(£o.t) A
From (2), PI =- P
PII(t +£o.t) - P" (1) = _ (A+ If) P (t) + AP (t) + IfP (t) + O(£o.t)
£o.t fA'" II-I ,...,,+1 /j[
From (l), P?=-PI
: o
= -
(A)2 P
_ P p"
~,(I+I1I)- ~,(t) [ () () (011(1)
Lt
tl.1 ....• C 11/
Lt
tl.r ....• O
-(?.+p)P" 1 +?.~'-I 1 +pP,,+1 1)+-
11/ Generally P
II
= (~)Tlp
P "
d~,(t) =-(?"+p)P,,(I)+?.P,,_I(I)+PP'I+I(I) Since
d/

where 1/ >0 (.: Lt 011(1) =

tl., ....•O 111
oj
In the steady state
,",P,,+~p"+(~)'p" + ~I
PII(t)-:'o, P,,(t) = P"

~ O
__=_-__(A_+
__P_)_P_"_+_AP
__ "_-_I+
__P_P_"_+_I ~
P" [ I + ~ + (~ )2 + ]=I
416 • Operations Research Queueing Theory • 417

expected (average) queue length L"

i.e., PO[~) 1--

=1
L =L_~=L
" S f.l. I-p

f.l.
IL " - P
1- P
Since ~ < I, sum of infinite G.P is valid.
f.l.
A expected wlait:g~ine i~the que;:=l
p()= 1--;;= I-p
" f.l.(f.l.-A) I-Pj
Also p,,~(~),p,,~(~),(1-~) Expected waiting line in the system

I
P"=p"(l_p) I W,.=W,,+--
f.l.
Measures of Model I _.:.:..A_ + _~=_~
f.l.(f.l.-A) f.l. f.l.-A
I. Expected (average) number of units in the system Ls

W=-~
s f.l.-A

Expected waiting time of a customer who has to wait (w/w > 0)

rw/w>O) = _1-
f.l.-A
= 1
f.l. (I-p)
I
6. Expected length of non-empty queue
f.l. 1
(UL>O)=--=--
f.l.-A I-p

~
= f p(f.l. - A)e -(J.I-A)Wdca

bability (waiting time in the system ~ t)

A
~
=~=~
A
p=-<1.
f.l.
= f p(f.l.-A)e(J.I-A)Wdw
1-- l-p'
f.l.
obability of waiting time in the queue ~ t
raffj'c mtensity
. A
CPI p =-
f.l.
~
 Queueing Theory • 419
418 • Operations Research
. L L W W (Littles formula) Expected number of jobs are
Inter-relationship between s- q' s, q

We know A
A J.l A
L,.=--A =--
L _ )1 =_A_ 1__ ).J.-A
s- A )1-A ).J.
1--
)1
1
1 = 48 = 1 2/3 jobs.
W=-- l/30-1/48
, )1-A

:.L,=A W, Since the fraction of the time the repairman is busy equals to A,
number of hours for which the repairman remains busy in 8-hour d~y
Similarly, L" = A W" hold in general.
W"
A
= -)1-(J.l---A-) ~ 8 (~ )~ 8x !~
~ 5 hours

1 Therefore, the time for which the repairman remains idle in an 8-hour
W,=--,
J.l-I\. =(8-5) hours =3 hours.
J.l- (J.l-A) A
1 J J =--~
J.l ()1-A) mple 16.2 At what average rate must a cl~rk at a supermarket
:. W, - )1 = )1-A - J.l = J.l(J.l-A}
in order to insure a probability of 0.90 that the\customers will not
I e to wait longer than 12 minutes? It is assumed that the~is only one
. W - W--
.. ,,- " J.l ter to which customers arrive in a Poisson fashion at an average rate
S per hour. The length of service by the clerk has an exponential dis-
Multiplying both sides by A, we have lion.
AW =A (w.- 1-) 'On
!
" ')1

A A , A.= ) 5/16 = customer/minute J.l= ?

L
"
= A Ws - -J.l = Ls /I
r-
b. (waiting time > 12)= 1-0.9=0.10
\ L,,=Ls -~ \ . :.j A(1- ~} -(p-).)w doi = 0.1
. t on hiS 12
Example 16.1 A r.v repa~rman. finds that the. time spe;e repairs
jobs has an exponential distributIOn with mean 3~ minutes. I~ a pro)(i-
A. ( 1--
).J.)(e -(Jl-)')W)~ =01
sets in the order in which they come in. If the arnval of sets ISh! is the A -(J.l-A) .
mately Poisson with an average rate of 10 per 8-hour. day, w ahead of 12

repairman's expected idle time each day? How many Jobs are
. ?
the average set just broug ht m.

Solution e(3-12/1)=0.4J.l

1
u = 2.48 minutes per service.
420 • Operations Research
Queueing Theory • 421
Example 16.3Arrivals at a telephone booth are considered to b, (ution
Poisson with an average time of 10 minutes between. on~ arrival and th: Given nA.=O.1 arrival/minute
next. The length of a phone call is assumed to be distributed exponen_
11 = 0.33 service/minute
tially with mean 3 minutes.
(a) What is the probability that a person arriving at the booth wiIJ
(a) Prob (waiting time ~ 10)= ](1- A.11 e -(I1-J..){J) dto
have to wait?
(b) What is the average length of the queue that forms from time to
10 J1. r
time? = -~(e -(I1-J..) rvf
(c) The telephone department will install a second booth when con- J1. 10

vinced that an arrival would except to have to wait at least three = 0.3 e -2.3 = 0.03
minutes for the phone. By how much must the flow of arrivals be (b) Prob (waiting time in the system ~ 10)
increased in order to justify second booth?

Solution = J(I1-A.)e-(P-A){J)dm
Given A= 1/10, f.1= 1/3 10

(a) Prob. (w>O)= I-Po= ~ = -Tax ~ =3/10=0.3 = e- 10 (I1-A) = e: 2.~ = 10

(c) The fraction of a day that the phone will be busy = traffic
(b) (L/L>O)=f.1(f.1-A) . . A. .
mtensuy p=-=O.3.
11
= ~ (~ -10) = 1.43 persons (d) Average number of units in the system
3 3 .
1
A
(c) WI/= f.1(f.1-A)
A. 10
L =--=--
J1. _ il.
.1 :1 I' --3/7-043c
-. us t orner
Since W =3, f.1=~, A=A' for second booth, 3 10
q 3

A' , ~ple 16.5 Customers arrive at a one window drive in bank

3= ( ) ~ A = 0.16 109 to Poisson distribution with mean 10 per hour. Service time
~ !..-A'
~stome~ is ex~nential with mean 5 minutes. The space in front of
3 3
indow mcludmg that for the serviced car can accommodate a max-
In of 3 cars. Others can wait outside this space.
Hence, increase In the arrival rate=O.I6-0.1O=0.06 arrival per
minute. (i) ~hat is the probability that an arriving customer can drive
.. dIrectly to the space in front of the ~dow?
Example 16.4 As in example 3, a telephone booth with poiSSO~ (II) W~at is the probability that an arriving customer will have to
arrivals spaced 10 minutes apart on the average and exponential ca '" wall outside the indicated space?
length averaging 3 minutes. (III) ~ow long is an arriving customer expected to wait before start-
(a) What is the probability that an arrival will have to wait more than 109service' (MU, BE, Nov. 93)
10 minutes before the phone becomes free? s
(b) What is the probability that it will take him more than JO minute 'Ulion
all together to wait for phone and complete his call? A. === 10 per hour
(c) Estimate the fraction of a day that the phone will be in use. _I
(d) Find the average number of units in the system. -'5 x 60= 12 per hour
422 • Operations Research
Queueing Theory • 423
A 10
(i) The probability of queue size> 6= p6
p=-=-
u 12
= (0.8333)6 = 0.3348.
(i) The probability that an arriving customer can drive directly to the
.e:
'P';:~n::n:~::: :'t:: +[~ )\
(1·1·) W - L,. _ 1-P

=
s -----=
A A
14.96 minutes
0.833
1- 0.8333
x3
-

~Po [1 + >[~),) 'ample 16.7

s~rvlce of an emergency
On an average 96 patients
clinic. Also on an average
per 24 hour day require
a patient requires

~[1-~)[1+~+[~))...
mmutes of active a~tentlOn. Assume that the facility can handle only

Po~1-~ . emergency at a tl.me. Suppose that it cost the clinic Rs .. loo per
ent treated to obtain an average servicing time of 10 minutes and
5 each minute of decrease in this average time would cost Rs.I 0 per

~[1-:~)[1 )~042 + :~+ :~

ient treated.
e the average
How much would
size of the queue from
have to be budgeted
I+ patients
by the clinic
to + patients?
to
.
(ii) Probability that an arriving customer will have to wait outside the (MU, BE, Mech Apr. 95)
indicated space
= 1-0.42=0.58
(iii) Average waiting time of a customer in a queue 96 I . t/ .
24 x 60 15 patren minute

= ~ f.l~ A = :~. C 2 ~ 10 ) =
5
1 2
I .
f.l=1O patient/minute

= 0.417 hours Average number of patients in the queue

A. A
L=--_
Example 16.6 In a supermarket, the average arrival rate of custom r q f.l Ar-f.l
is 10 every 30 minutes following Poisson process. The average time taken
by a cashier to list and calculate the customer's
following
length exceeds
exponential
6?
distribution.
purchase is 2.5 minute
What is the probability that the queue :: (~r
(
~_~)~
I.
:: 13" patients

What is the expected time spent by a cu tomer in the system? 10 15 10

(MU, MBA, Nov 95)
But L,,=1+ is reduced to L,,'= 1/2
Solution
, 10 .
r, Substituting L,,' = 1/2, A' = A = _1_ in the formula
I\. = - per minute IS
30 L' AI2

I . q f.l'(f.l' -A')

(~rr:,
f.l = - per rmnute
2.5

A
I
"3
r.
p = -=--=0.8333 2- ,( , =2/15 patients/ minute
f.l 1/ 2.5 u ).I -1/ 15
 Queueing Theory • 425
424 • Operations Research

Hence the average rate of treatment required is ~

,
= 7.5 .
minUtes. p= ~ = ~~=0.75
Decrease in time required by each patient u
P 0.75
= 10- Jl=~ minutes (i) L ---- = 3 trains
I ,- I-p - 1-0.75
2 2
:. The budget required for each patient
5 (ii) PC ~ 10) = (0.75)10 = 0.056
= 100+ 2x 10=Rs. 125
(iii) When the input increases to 33 t:ain per day
So inorder to get the required size of the queue the budget should b
increased from Rs .. 100 to Rs .. 125 per patient. e 33 I I. / .
we have A , f.1=--"- trams mill
60 x 24 480 36
Example 17.8 In a public telephone booth the arrivals are on the . A
average 15 per hour. A cal1 on the average takes :I minutes. If there is ==> L =-P-where p=- =0.825
, 1- P J1
just one phone, find (i) the expected number of cal1ers in the booth at
any time (ii) the proportion of the time the booth is expected to be idle?
0.825 5 train (approximately)
L
(MU BE. MBA APR. 96) S I - 0.825

Solution also p (~ 10) = plO = (0.825)10= 0.1460

Given A= 15 per hour
I
u =- x 60 per hour
3
EXERCISES
(i) Expected length of the non-empty queue

=J_= 20 =4 . People arrive at a theatre ticket center in a Poisson distributed arrival rate of
f.1-A 20-15
25 per hour. Service time is constant at 2 minutes. Calculate
··) Th e service
.. b usy =-=-A 15 =- 3
C 11 IS (i) the mean number in the waiting line.
f.1 20 4
(ii) the mean waiting time.
:. the booth is expected to be idle for I - 3 =~ hrs (iii) utilisation factor.
4 4
= 15 minutes. [Ans: (i) 4 people. (ii) 10 minutes. (iii) p=0.833]

Example 17.9 In a railway marshal1ing yard, goods train arrive at At a one-man barber shop, customers arrive according to Poisson distribution
a rate of 30 trains per day. Assuming that inter arrival time and the service with a mean arrival rate of 5 per hour and the hair cutting time was expo-
time distribution fol1ows an exponential distribution with an average of nentially distributed with an average hair cut taking 10 minutes. It is assumed
30 minutes. Calculate the following that because of his excellent reputation. customers were always willing to
(i) the mean queue size wait. Calculate the following -
(ii) the probability that queue size exceeds 10 (a) Average number of customers in the shop and the average numbers wait-
(iii) if the input of the train increases to an average of 33 per day, ing for a hair cut.
what wil1 be the changes in (i) and (ii)? (b) The per cent of time arrival can walk in right without having to wait.
(c) The percentage of customers who have to wait before getting into the bar-
Solution ber's chair. [Ans: (a) 4.8 (b) 83.3% (c) 16.7%]

G.ivcn I\,, 30 I . / .
60 x 24 48 trams minute Cars arrive at a petrol pump with exponential inter arrival time having mean
I . . .!. minute. The attendant take on an average of .!.. minute per car to supply
u =-6 trains/minute ~trol. The service time being exponentiall~ distributed. Determine
3
426 • Operations Research Queueing Theory .·427
(i) the average number of cars waiting to be served.
difference equations of Model I are valid for this model as long as
(ii) the average number of cars in the queue.
<N. Arrivals will not exceed N in any case. The various measures of
(iii) the proportion of time for which the pump attendant is idle.
[Ans: (i) 2 cars, (ii) 4/3 cars, (iii) 0.34] is model are I-p A- (A- )
I. P = 1 where p =- -> 1 is allowed
4. The mean arrival rate to a service centre is 3 per hour. The mean service time o I-p N+ )1 )1
is found to be 10 minutes per service. Assuming Poisson arrival and expo_
l-p
nential service time. find 2. Po= I_pN+1pn for n=0,1,2, .... .N
(i) the utilisation factor for this service facility
(ii) the probability of two units in the system. 3. L,,=AJ)1
(iii) the expected number of units in the system.
N
(iv) the expected time in minutes that a customer has to spend in the system.
= Po I,n p"
+- (ii) i· (iii) I, (iv) 1/3]
4.Ls
[Ans. (i) n=O
A-
5. At a public telephone booth in a post office arrivals are considered to be 5. L,,=Ls--
)1
Poisson with an average inter arrival time of 12 minutes. The length of the
phone call may be assumed to be distributed exponentially with an average 6. W,=L/A-
of 4 minutes. Calculate the following
(a) What is the probability that a fresh arrival will not have to wait for the
phone?
(b) What is the probability that an arrival will have to wait more than 10 min-
mple 16.10 In a railway marshalling yard, goods train arrive at
utes before the phone is free?
rate of 30 trains per day. Assume that the inter arrival time follows
(c) What is the average length of the queue that form from time to time?
exponential distribution and the service time is also to be assumed as
[Ans. a = 0.67, b = 0.063, c = 1.5]
ponential with mean of 36 minutes. Calculate
6. Consider a self service store with one cashier. Assume poisson arrivals and (i) the probability that the yard is empty
exponential service time. Assuming nine customers arrive on the average of (ii) the average queue length assuming that the line capacity of the'
every 5 minutes and the cashier can serve 10 of them in 5 minutes. find yard is 9 trains.
(i) (a) the average number of customers queueing for service.
(b) the average time a customer spends in the system,
(c) the average time a customer waits before being served. 30 '.
(ii) the probability of having more than 10 customers in the system. 60 x 24 48 )1=16 trams per minute
(iii) the probability that a customer has to queue for more than 20 minutes. If A- 36
the service can be speeded upto 12 in every 5 minutes by using a different p=- =-xO.75
cash register what will be the effect of the quantities (i), (ii), (iii).
)1 48
(i) The probability that the queue is empty is given by
[Ans. Case (i) (a) 3 customers
(b)P(~IO)=(0.9)1II l-p
(c) P (W~2)=0.67. P,= •.••, where N=9
r I_p'~'
Case (ii) (i) 3 customers ..••..
I ~O·.75 = 0.25 = 0.28.
(ii) P (~10)=(0.75)1II
1-(0.75)9+1 0.90
(iii) P (W~2)=O.31
(ii) Average queue length is given.by
Model II (MIMII): (NlFCFS)
1 N
This model differs from that of Model I in the sense that the L =
S
-pN+I £..J
~ np"
maximum number of customers in the system is limited to N. Therefore, I-p ,,=0
 Queueing Theory • 429
428 • Operations Research
Example 16.13 Assuming for a period of2 hours in a day (8-10 am)
9
1-0.75 L,n(0.75)O pins arrive at the yard every 20 minutes, then calculate for this period.
10
1-(0.75) ,,;0 (i) the probability that the yard is empty,
(ii) average queue length assuming that the capacity of the yard is 4
= 0.28 X 9.58 = 3 trains
trains only.
Example 16.11 A barber shop has space to acc?mmadatc only 10
customers. He can serve only one person at a time. If a customer comes
to his shop and finds it full he goes to the next shop. Customers ran~omly
. . 36
arrive at an average rate A= 10 per hours and the barber service time is Given p=- = 1.8 N =4
20 '
negative exponential with an average of 1/J1=5 minutes per custome/
p-\
Find Po' P". (a) Po = -5- = 0.04
P -I
Solution 10 \
Given N= \0, A= 60 ,J1=S (b) Average queue size
4
A 5
p=-=- =poIn.p"
J1 6 n=O

I-p 1-5/6 = 0.04 (p + 2p2 + 3p3 + 4p4) =29 = 3 trains

P. ---=---.,-:-
o-I_pll 1-(5/6)11

0.1667 0.1926 EXERCISES

0.8655
I -p N
P" = ( 1 _ P N+ I) p , . Patients arrive at a clinic according to a Poisson distribution at the rate of 30

Example
~(0.'926)X(~

16.12
l [n~O. 1.2 •... 10]

A car park contains 5 cars. T~e arrival of cars IS

.
patients per hour. The waiting room does not accommodate more than 14
patients. Examination time per patient is exponential with mean rate 20 per
hour.
0) Find the effective arrival rate at the clinic.
Poisson at a mean rate of 10 per hour. The length of time each car spends (ii) What is the probability that an arriving patient will not wait? Will he find
J in the car park is exponential distribution with mean of 5 hours. HoW a vacant seat in the room.
(iii) What is the expected waiting time until a patient is discharged from the
many cars are in the car park on an average?
clinic? -
Solution 10 . 1
. 30 20 2/
Given N=5, A= 60' J1= 2 x 60 Hint: N = 14, A = 60 ' J1= 60 P = 73 .
A.
p=- =20 N
J1
-19
Find P", Po and w, = ; InP"
I-P \-20 11=0
PO=(I_pN+lll-206 - 6399
[Ans: (a) 19.98 (b) 0.67 (c) 0.65]
= 2.9692 x \ 0-3
A stenographer has 5 persons for whom she performs stenographic work.
N 5 II
L, = Po I np lI
=(2.9692 x 10- I n( 2.9692 xl 0-")
3
)
~rrival rate is Poisson and service times are exponential. Average arrival rate
Is 4 per hour with an average service time of 10 minutes. Find
n=O II=()
(i) the average waiting time of an arrival
=4 (approximately) i) the average length of waiting line
430 • Operations Research Queueing Theory • 431

(iii) the average time on arrival spent in the system.

p,,1(t) = -(A + SJ.L)P"(t) + AP,'_I(t) + SJ.LP,,+I(t)
[Ans: (i) 12.4 minUte~
forn=S. S+I. S+2 •....
(Il()..0).7 9
= One (appj
III 22.4 minUtes] Considering the case of steady state i.e. when t-'too P (t)-'tP and
hence P"I(t)-'tO for all n, above equation becomes. '" "
3. Customers arrive at a one-window drive in bank according to a poisson di '_
tribution with ~ean 10 per hour. Service time per ~ustomer is exponenti:1 0= -APO +}.J.PIfor n = 0 (I)
with mean 5 minutes. The car space m front of the window mcluding that fo O=-(A+n,u)p" +}..P,,-I +(n+I),uP,'+1 (2)
the service can accommodate a maximum of 3 cars, Other cars can wait outsid;
for 1;5;n;5; S-I
this space,
(a) What is the probability that an arriving customer can drive directly to the O=-(A+S,u)P" +}..P,,-I +S,uP,,+1 for n z S (3)
space infront of the window?
(b) What is the probability that an arriving customer will have to wait outside Here Po = Po initially
the indicated space? A
(c) How long is an arriving customer expected to wait before starting the ser- PI = - Po from (I)
,u
vice? [Ans: (a) 0.42. (b) 0.48 (c) 0.417]
A A2
Model III: (Multi service model) (MIMIS): (ooIFCFS) P2 = -
2,u
~ = --2
2!,u
po(Put n = I in (2»)
When there are n units in the system. may be obtained in the following
two situations, A A3
P3 = -3,u P"- = --3
3!,u
Po (Put n = 2 in (2»)
(i) if n ~s. all the customers may be served simultaneously, There
will be no queue, (s-n) number of servers may remain idle and then
fin = nil, n = 0, I, 2 ... S;
(ii) if n ~ s, all the servers are busy. and the maximum number of
A
In general P" =-P,,-I
n,u
=-I
n!
(A)"
-
,u
Po
customers waiting in queue will be (n -s). then 11" =sp
l;5;n;5;S
AlsoA.,,';"A.[n=O, 1.2 ... J
The steady state difference equations are

po(t + L\t) = Po (t)(1- A.L\I)+ PI (t)pL\t + 0( L\t)

for n = 0
P" (t + L\t) = P" (t)[ 1- (A. + n,u)L\t] + P,,-I (t )A.L\I
+ Pn+I(/Xn+ 1)I1L\t+O(L\t) for n=l. 2.. ,S-1
A I I (A )S+2
PS+2 = S,u PS+I = STS! -;;. Po
p,,(t + M) = Pn (t)[ 1- (A. + SI1)M] + P,,-I (t)A.L\t +
P,'+I(t)SI1L\t+o(M) for n= S', S+I. S+2, .. In general P" = PS+(II-S) = --- I I -
S"-S S! ,u
(A)" Ro n"?S

Now dividing these equations by M and taking limit as Si-: 0, the

difference equations are- Now find PlIO using the fact

p,,1(t) = -AP" (t) + 11 ~ (t) for n = 0

Pnl(t) = -(A. + n,u) P" (t) + W,,-I (I) + (II + 1),uP,r+ I (t)
for n = 1.2 ... S-1.
432 • Operations Research Queueing Theory • 433

S-I ~ 8 . Probability that there will be some one waiting =~

ie 2/~,+ 2/~,I =
I-p
11=0 n=S
9. Average number of idle servers
This gives the steady state distribution of arrivals (n) as
= S - (average number of customers served)

-1 -(It)n Po, n=O, 1,2 ... S-1

10. Efficiency of MIMIS model.
n! f.1 average number of customers served
P,,=
total number of customers served
...!...~S(~)" P.o' if n = S, S + 1. ..
S!S"- f.1 ample 16.14 A super market has two girls ringing up sales at
counters. If the service time for each customer is exponential with
Measures of Model III:
ean 4-mi-nutes and if people arrive in a Poisson fashion at the counter
I. Length of the queue
the rate of 10 per hour, then calculate
I..)S (a) the probability of having to wait for service.

- P Where P
( - Po
= ~Ji..:.....-_
(b) the expected percentage of idle time for each girl.
L" - Ps 2 s (c) if a customer has to wait find the expected length 01" his waiting
(1- p) S! time.

2. Length of the system (a) Probability of having to wait for service is

It
Ls=-+Lq
f.1
3. Waiting time in the queue
p(W>o)= l.H,
S!{I - p)
Po

L"
~/=T 1..=~,Ji= ~,S=2

4. Waiting time in the system A I

p=-=-
Jis 3
W=~
S A
_ [S-I (Spr + (Sp)s ]-1
5. The mean number of waiting individuals who actually wait is Po - I. --;!
11=0
S!{I- p)
given by
-I
I
L(L>O)"F--
I-p

6. The mean waiting time in the queue for those who actually wait is
given by (WIW>O)
I
Sj.1-A 4

P =1+2/+~=~
7. Prob(W>O) =--'-- 73 2x2/ 2
I-p 73
434 • Operations Research
Queueing Theory • 435

(%)2 .~ _ 1 (pSf + (pS)2

-1

Thus probability (W > 0) = ( ~) =~

2! 1- 13 - [ ~ n! 2! (I-p) ]

(b) The fraction of the time the service remains busy -1

. tra ffiIC mtensity

ityj is gi b y p = SJ1
A. = 3I (pS)1 (pS)2
(ie. IS given
= [ I +-1!-+ 2! (1_p) ]

:. The fraction of the time the service remains idle is

(
I - J..)
3
= ~ = 67% (approximately)
3

I 1 I I .
(c) (W/W > 0) = -_.- = --.-- = 3 minutes.
1- p SI1 1- 7j• 2x ~
4

Example 16.15 A petrol station has two pumps. The service time
follows the exponential distribution with mean 4 minutes and cars arrive P(W>O)=~=_I_(~r Po
for service in a poisson process at the rate of 10 cars per hour. Find I-p I-p S!
the probability that a customer has to wait for service. What proportion
of time the pumps remain idle? (MU, MeA. Nov. 95)

1 I 3
Solution = =-x-=0·167 (app)
I-~ 9 2
Given S = 2, A. = 10 per hour
3
J1 = '4I per .
minute
60
=4 = 15 per hour
ample 16.16 A telephone exchange has two long distance oper-
P=~=~=~ rs. The telephone company finds, that during the peak load long dis-
SJ1 2xl5 3
ce calls arrive in a Poisson fashion at an average rate of 15 per hour.
(ii) The duration of time for which the pumps are busy =~ = 1. e length of service on these calls is approximately distributed with
SI1 3
an length 5 minutes.
:. The duration of time for which the pumps remain idle
(a) What is the probability that a subscriber will have to wait for his
1- 1. = 1. = 67% (app) long distance call during the peak hours of the day.
3 3
(b) If the subscribers will wait and be serviced in turn, what is the
expected waiting time?
Establish the formula used.
(i) Prob (W > 0) lution

1. J1=1. S=2
4 ' 5 '
(Sp)" (Sp)s ]-1
where Po =
[
S-I
I--
,,=0
( )
S! 1- P
n!
•
" P =-=-
A. 5
j1S 8
 Queueing Theory • 437
436 • Operations Research

-1
. S-I (Sp)" (Sp)S
F"" calculate Po = [~ --;;! + S' (I _ p) ]

= [20 +~+ 22 + 23+ 256 X..!.]-I

O! 1! 2! 3! 24 8
l
= [1+2+2+ % +%r = [5+%r
l
=.%3

t, the expected queue length A.q is

(a) p(W > 0) =

(~r
S! I-p
( ) . Po
= -,,--(~_)1),--S
S!
p
l_p2
p. _ ~
0 - 4! (~r
_~_ . 2. _ 4/
23 - /23

EXERCISES

o channel waiting line with Poisson arrival has a mean arrival rate of 50
(b) Wq = LqjA and exponential service with a mean service rate of 75 per hour for each
l. Find
= I p(Sp)S P. probability of an empty system
AS! (1_p)2 0 probability that an arrival in the system will have to wait.
(ADS: (i) 0.83 (ii) 0.167)
=4 Ys.(%)2? .~=_12_5=3.2 minutes
2! (1- Ysf 13 39 are two clerks in a university to receive dues from the students. If the
time for each student is exponential with mean 4 minutes and if the boys
in a Poisson fashion at the counter at the rate 10 per hour, determine
Example 16.17 Four counters are being run on the frontier of a
probability of having to wait for service.
country to check the passports and necessary papers of the tourists: The
expected percentage of idle time for each clerk?
tourists choose any counter at random. If the arrival at the frontier IS
[ADS: (i) 0.167 (ii) 0.67]
Poisson at the rate A and the service time is exponential with parameter
)J2, what is the steady state average queue at each counter?
general insurance company has three claim adjusters in its branch office.
Solution with claims against the company are found to arrive in Poisson fashion
average rate of 20 per 8-hour day. The amount of time that an adjuster
Here S = 4, A = A, )1 = N2 with a claimant is found to have a negative exponential distribution
n service time 40 minutes. Climants are processed in the order of
ceo
many hours a week can an adjuster expect to spend with claimants?
 Queueing Theory • 439
438 • Operations Research
mple 16.18 A barber shop has two barbers and three chairs for
(b) How much time on an average does a claimant spend in the branch offic
. ~ lomers. Assume that the customers arrive in Poisson fashion at a rate
(Ans: (a) 22 hours (b) 49 minUt
es) S per hour and that each barber services customers according to an
4. A railway goods traffic station has 4 claims assistants. Customers with clai nential distribution with mean 15 minutes. Further, if a customer
against the railway are observed to arrive in a Poisson fashion at an average r:S 'ves and there are no empty chairs in the shop, he wi!lleave. What is
of 24 per 8 hour day for 6 days. The amount of time the claims assistant spen~e expected number of customers in the shop?
with the claimant is found to have an exponential distribution with a mean ser~
vice time of 40 min. Service is given in the order of appearance of the Customer
. s.
(i) How many hours/week can a claim Assistant expect to serve the claimanp 5 I
(ii) How much time does a claimant on the average spend in the goods traffi~ Here S = 2, N = 3, A =- = - Customer / min
60 12
office [Ans: (i) 72 hours (ii) 47.2 minutes]
1 .
j1=-/mm
IS
Model IV(MIMIS/NIFCFS)
This model is essentially the same as model III except that the maximum
number of customers in the system is limited to N, where N > S (S ==
No. of channels)

.
.. AII-
_ {A, 0:5 n :5 N
0, n >N
_ {nj1, 0:5 n :5 S
u; - Sj1, S _< n_< N

P,,=

p = {~! (1;;)" x 0.28 0:5n<2

n 211~22! (1;;)" x 0.28

1
= -(1.25)n x 0.28 0:5n<2
n!
c I
L = (Sp) P [1_ pN-S+l _ (1_ p)(N - S + l)pN-S ]po = --n 2 (1.25)" x 0.28
q S!(I_p)2 . 2 - 2!

(A)"
L =L +S_R~(S-n)(Sp)1I
S q o£..t I . S-l (S-n) -
11=0 n.
Ls = L" + S - Po £..t
" n'. j1
11=0
Ws = ~: where A'

1
= A(I- PN)
= ±(n-2)p" +2-P oI (2-n)(1.25)"
n!
W" =Ws-- n=2 n=O
j1
 440 • Operations Research

= P3 + 2 - 3.2 Po

= [_1_(1.25)3 X 0.28] + 2- 3.2 X 0.28 = 1.226 customers

2.2! .

EXERCISES

I. Let there be an automobile inspection site with three inspection stalls. ASSUI
that cars wait in such a way that when a stall becomes vacant, the car at I
head of the line pulls up to it. The station can accommodate at most 4 c
waiting (seven in station) at one time. The arrival pattern is Poisson wid
mean of one car every minute during peak hours. The service time is eX
nential with mean 6 minutes. Find the average number of cars in the site d
. ing peak hours, the average waiting time and the average number of cars
hour that cannot enter the station because of full capacity.
I
[Hint: S = 3, N = 7. ,1, = I car per minute J1 = '6 car I rr

[Ans (i) 3.09 e

(ii) 6.06 (
(iii) 12.3 mim
(iv) 30.3 cars per he