Queuing Theory

Presented by: Dr. Pratik Ranjan Nayak

Mentored by: Dr. Smaranita Sabat

1
 Outline
• Introduction
• Customer and server
• Customer characteristics
• Types of queue
• Queue discipline
• Other considerations
• Little’s law
• Models
• Application
• Summary

2
 Introduction
• Definition: It is mathematical study of formation, function and congestion
of queues.
• It was developed by A.K. Erlang in 1904 for telecom industry.
• It was to solve the congestion in telecom industry.
• Customer = One requiring a particular service (Demand)
• Server = One performing a service (Supply)
• A queue is formed when customers (demand) exceeds capacity of servers
(supply)

3
 Customer and server

Category Living Non-living

Customer Patient X-Rays

Server Physician Patient Record retrieval

software

4
Queueing process

5
 Customer characteristics
• Taking the scenario of a walk in clinic:
 On arrival, customers can balk/leave when long queue is present.
 If wait time is long, then customers can renege/abandon queue

• Queue psychology is how people feel while waiting.

• Some ways to make wait time seem short:
 Distractions – TV screen in waiting area
 Token system – Known, finite wait time
• Any delay due to technical error should be communicated promptly.

6
 Types of queue
• Parallel: Each server – Own queue
Eg. A Physician has own patients
• Network: Customers - Different servers in a sequential manner
Eg. Patient comes for surgery

Operating room

Recovering unit bed

Surgical ICU bed

7
 Queue discipline
• Order in which queued customers are served
 First come, first serve (FCFS) – Default option, Eg. OPD patients
 Priority – Done as part of triage system (time-sensitive customers)
1. Pre-emptive (Eg. Stroke patients) If service can be interrupted to serve
2. Non-pre emptive higher priority customer

 Last in, first serve (LCFS) – Used in data structures like Stack.
 Service in random order (SIRO)

8
 Other considerations
• Infinite waiting room is only possible in digital scenario.
• In health sector, mainly waiting area is limited.
• Telemedicine is an exception.
• Kendal notation is used to characterize Queueing systems.
• A/S/c/K/N/D (Arrival process/Service time/number of servers/capacity of
queue/Number of possible customers/Discipline of queue)

9
 Continued
• Steady state:
 System operates with same arrival rate
 Average service time
 All characteristics are there for a long time, so
Probabilistic behaviour of performance measures (queue length and
customer delay) independent of system observed.

• Health care systems:

 Time varying characteristics
1. Time of day
2. Day of week
3. Seasonal effects

10
 Little’s law
• It is formula that operationalizes Queueing theory.
• L = λW
• L is average number of customers in system.
• λ is average arrival rate into system.
• W is average amount of time spent in system.

11
 Continued
• Let us say, there are 10 people in a queue to get appointment for OPD,
1 server, 1 person is served every 10 minutes. Estimate how long will it
take for your turn:
• W = L/λ
• W = 10/0.1
• W is 100 minutes
• So you will have to wait 100 minutes or 1 hour 40 minutes to get your turn

12
 Models
• Poisson process assumptions:
 Customers arrive one at a time.
 Probability of one customer’s arrival independent of other customers’
arrival.
 Probability customer arrives at a given time is independent of time.
 Emergency department is a place where this model is implemented.

13
 Continued
• M/M/S model – Erlang delay model
• It has Markovian (statistical process to describe randomness) assumptions:
 Single queue – unlimited waiting room Feeds identical servers

 Customers arrive with constant rate, service duration is exponential in

distribution.

14
 Continued

• ATM queue uses Markovian model.

• One customer at a time.
• First come, first out order.
• Randomly distributed arrival process and service distribution time.
• Unlimited queue capacity and number of possible customers.

15
 Advantages and disadvantages
• Advantages:
 Very little data required and simple formulas used.
 Performance measures can be predicted:
1. Mean delay
2. Probability of waiting more than given amount of time before being
served.

• Disadvantage: Operational data is required for input (often unavailable)

16
 Application
• Healthcare is a major service sector.
• Queue is formed when waiting for appropriate care (OPD appointment or
diagnostic test to be carried out).
• Most of time, disparity between demand (service) and supply (capacity)
exist.
• Timely access is a key element of quality healthcare.
• Queueing theory is used to utilize resources in most cost-effective way to
reduce delays.
• Manager is always presented with demands from competing sources and
different types of patients.

17
 Continued
• There might be difference between administrators, Physicians, nurses and
patients.
• What-if analyses could be performed.
• Tradeoffs identified (Bed flexibility and quality of care)
• Find attractive solutions and not only estimate performance.

• Models are used for:

 Identify appropriate staff levels, equipment and beds.
 Resource allocation (Decision making)
 Designing new services

18
 Summary
• Queueing theory is mathematical study of queues – its formation, function
and congestion.
• There are customers (demand) and servers (supply) for any queue.
• Both can be living or non-living.
• Queueing psychology is how people feel waiting in queue.
• There are two main types of queues – parallel and network
• Queueing discipline is the order in which customers are served.
• Little’s law operationalizes this theory.
• There models like Poisson process that can be used in healthcare setting to
maximize utilization of available resources to met patient demands.

19
THANK
YOU

20