Design of Services

Capacity Planning under uncertainty

Use of waiting line models
• Often demand placed on resources is uncertain making
capacity requirement estimation difficult
• In such cases, waiting line models
– make use of queueing theory fundamentals
– to analyse the impact of alternative capacity choices
– on important operational measures such as queue length, waiting time
and utilisation of resources
• In service systems, waiting time is an important operational
measure that determines the service quality
– Computerised passenger reservation facility of Indian Railways
– Banking system or BSNL’s bill payment counters
Components of Queuing System

Calling
Population
Arrivals
Waiting Line Server Served
customers
Components of Queuing Systems
Infinite
Calling Population
Finite
Rate Markovian, General Distbn., Deterministic
Arrival Parameters
Pattern Single, Bulk, Special group
Queue
FCFS, LCFS, Random, Balk, Renege, Jog
Parameters
Servers Single, Multiple

System Structure Stages Single, Multiple

& Parameters Routing Single, Serial, Network
Capacity Finite, Infinite
Service
Markovian, General Distbn., Deterministic
Parameters
Performance Queue length, Waiting time, Utilisation,
Metrics Cost based
 Single-Channel Structures

Single-server, single-stage

Waiting line Server

Single-server, multiple stages

Waiting line Servers

 Multi-Channel Structures

Multiple-servers, single stage

Servers

Multiple-servers, multiple-stages

Waiting line
Servers
 Single Server Queue
Formulae for Lq
Ls Average number of customers in the system
(waiting to be served)
Lq Average number of customers in the waiting line
Ws Average time a customer spends in the system
(waiting and being served)
Wq Average time a customer spends waiting in line
 mean arrival rate
 mean service rate
S Number of servers in a multi-server queue
Single server Queue 
Lq =
(Exponential service time) 
 Performance Metrics
Relationships
Server utilisation

In the case of single server: 


In the case of multiple servers:  
S
Little’s Formula
Average time customer Ls
Ws =
spends in system 
Lq
Average time customer Wq =
spends in queue 
In the case of a Single Server
Average number of 
Ls = Lq +
customers in system 
 Problem
• The teller facility of a bank has a one-man operation at present.
Customers arrive at the bank at the rate of one every 4 minutes to
use the teller facility. The service time varies randomly across
customers on account of some parameters. However, based on
the observations in the past, it has been found that the teller takes
on an average 3 minutes to serve an arriving customer. The
arrivals follow Poisson distribution and the service times follow
exponential distribution.
– What is the probability that there are at most three customers in front
of the teller counter?
– Assess the various operational performance measures for the teller
facility.
– Of late the bank officials notice that the arrival rate has increased to
one every three and a half minutes. What is the impact of this change
in the arrival rate? Do you have any observation to make?
 Solution to Problem

• Arrival rate at the bank: =  = 15 per hour

• Service rate at the teller: = = 20 per hour

• Utilisation of teller facility:  15

   0.75
 20 n 3
• Probability of at most three customers in the system = P
n 0
n

• Using equation, we compute Pn for values of n = 0 to 3

P0 = (1-
 ) = 0.25; P1 = 0.25*0.751 = 0.1875;
P2 = 0.25*0.752 = 0.1406; P3 = 0.25*0.753 = 0.1055.

• Probability of at most 3 customers =

0.25 + 0.1875 + 0.1406 + 0.1055 = 0.6836
Operational Performance Measures
2
Avg. No. of customers in the waiting line: Lq 
 (   )

Avg. No. of customers in the system: Ls  Lq 

Lq
Avg. time a customer spends waiting in line: Wq 


Ls
Avg. time a customer spends in the system: Ws 

Operational Performance Measures

2 15 2
Avg. No. of customers in the waiting line: Lq    2.25
 (    ) 20(20  15)
 15
Avg. No. of customers in the system:
Ls  Lq   2.25   3.00
 20
Lq2.25
Wq    0.15 Hr  9 min
Avg. time a customer spends waiting in line:  15
L 3.00
Ws  s   0.20 Hr 12 min
Avg. time a customer spends in the system:  15
 Impact of Arrival Rate

Arrival rate = 15 Arrival rate =

per hour 17.143 per hour

Utilisation of the teller facility 75% 85.7%

Avg. number of customers in 2.25 5.14
waiting line

Avg. number of customers in the 3.00 6.00

system

Average time a customer spends 9 minutes 18 minutes

waiting in line

Average time a customer spends 12 minutes 21 minutes

in the system
 Capacity Design issue
Flexibility/Utilization Trade-off
High utilization
Low cost of operation
Operational Performance

Poor service
Measures

Low utilization
High cost of operation
Good service

0 100%
Utilization ()
 Cost Relationship in Queuing
Expected costs

Total cost

Service
cost

Waiting Costs

Level of service
 Formulae For Lq
Three types of Queuing systems


Single server Queue Lq =

Exponential service time

Single server Queue 

Lq =
General service time 


Single server Queue Lq =

Deterministic service time
 
= =
 
 Multi-Server Queues
Approximation for Lq based on
_ data
_
X
_a Mean Inter-Arrival Time (IAT)
Xs Mean Service Time (ST)
Sa Standard deviation of inter-arrival time
Ss Standard deviation of service time _
Ca Coefficient of variation of IAT = Sa/X_a
Cs Coefficient of variation _of ST = Ss/Xs
 Mean arrival rate = 1/ X _a
 Mean service rate = 1/ Xs
 Utilisation of the ‘s’ servers = S
 2( S 1)  C a2  C s2 
Lq  * 
(1   )  2 
 