COMMERCE 341

OPERATIONS MANAGEMENT
Queuing Theory
Fall 2015
Geoff Pond

Agenda

Introduction
Single Channel, Single Server w/ Exp. Service Time
Single Channel, Single Server w/ Constant Service Time
Single Channel, Multi-Server w/ Exp. Service Time
Psychology of Queues
Strategies

Introduction
Like it or not, it happens everyday:
Vehicles arriving at a traffic light
Customers arriving at Tim Hortons
Container ships arriving at a port and waiting for a tug
boat
Lineups at customs/immigration/passport control
Aircraft waiting on a taxiway for runway clearance
Students waiting in line outside Stages on a Friday
night at 11pm
Students waiting at Smokies for some 2am poutine
Patients arriving at KGH ER to have their stomach
pumped at 3am

Introduction
Why is it important to study?
Customers waiting for extended periods may balk
There may be insufficient space to accommodate the
queue length (causing safety problems)
Efficiency/capacity implications (bottlenecks)
Effectiveness implications (e.g., patient conditions
may deteriorate)
Just imagine the monetary
value of the fuel being burned
as these aircraft wait for takeoff (Atlanta International)

Introduction

Our goal is to find

the sweet spot
that minimizes
total cost.

Different Queue Models

Naming Convention

A / S / k
where:
A represents the distribution of arrival times
S represents the distribution of service times
k represents the number of servers
A and B are normally either:
M meaning Markov or memoryless
D meaning deterministic or fixed service
time

Naming Conventions
Most models deal with Markovian or memoryless
processes.
Arrival rates are typically described using the Poisson
distribution.
Service times are typically described using the
Exponential distribution.

Naming Conventions
M/M/1
One server (e.g., a single
cashier)
Markovian service, i.e., exponentially distributed
service times
Random arrivals, i.e., Poisson arrival rate

M/M/3

Queuing Source
Infinite Source
There is a large calling population
Finite Source
The calling population is small enough that as the
number of units in the system increases, a change
is observed in the arrival rate of new units
In the extreme case, if a calling population where 5
units and all 5 units were in the system, the arrival
rate would become 0
But obviously it was initially higher than that (or
the five units wouldnt have arrived in the first
place)

Typical Waitline Parameters

Note that problems dont necessarily give you the data in a rate in
these cases it will be up to you to make the conversion.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

M/M/1

Example

Consider a bank where customers arrive at a rate of 20 customers

per hour. On average, a customer is served every 2 minutes.

WARNING
The equations we just discussed are specific to
M / M / 1 queuing systems. They are NOT general
equations, meaning that they cannot necessarily
be applied to M / D / 1 or
M / M / >1 systems that we will look at next.

M/D/1

This is the only equation that is

different from M / M / 1

Example

A vending machine dispenses hot chocolate or coffee. Service

duration is 30 seconds per cup and is constant. Customers arrive at a
mean rate of 80 per hour (following a Poisson distribution). Also
assume that each customer buys only one cup. What is the average
number of customers waiting in line?

M/D/1
Example

A vending machine dispenses hot chocolate or coffee. Service

duration is 30 seconds per cup and is constant. Customers arrive at a
mean rate of 80 per hour (following a Poisson distribution). Also
assume that each customer buys only one cup. How long do
customers spend in line?

M/D/1

Example

A vending machine dispenses hot chocolate or coffee. Service

duration is 30 seconds per cup and is constant. Customers arrive at a
mean rate of 80 per hour (following a Poisson distribution). Also
assume that each customer buys only one cup. What is the server
utilization?

Littles Flow Equations

Regardless of the queuing model, the following
equations are general. Theyre also easy so they can
be quite handy, especially when dealing with
otherwise complicated problems.

You Try One!

Many of a banks customers use its automated teller
machine (ATM). During the early evening hours in the
summer months, customers arrive at the ATM at the
rate of one every other minute s(assume Poisson).
Each customer spends an average of 90 seconds
completing his or her transaction. Transaction times
are exponentially distributed. Determine:
a) The average time customers spend at the
machine, including waiting in line and completing
transactions.
b) The probability that a customer will not have to
wait upon arrival the ATM.
c) Utilization of the ATM.
d) The probability of three customers waiting in line.

M / M / >1

M / M / >1
Example

A small town with one hospital has two ambulances. Requests for an
ambulance during weekday mornings average .8 per hour and tend
to be Poisson. Travel and loading/unloading time averages one hour
per call and follows an Exponential Distribution. Find the server
utilization.

M / M / >1
Example

A small town with one hospital has two ambulances. Requests for an
ambulance during weekday mornings average .8 per hour and tend
to be Poisson. Travel and loading/unloading time averages one hour
per call and follows an Exponential Distribution. Find the average
number of patients waiting.

M / M / >1
Example

A small town with one hospital has two ambulances. Requests for an
ambulance during weekday mornings average .8 per hour and tend
to be Poisson. Travel and loading/unloading time averages one hour
per call and follows an Exponential Distribution. Find the average
number of patients waiting.

M / M / >1
Example

A small town with one hospital has two ambulances. Requests for an
ambulance during weekday mornings average .8 per hour and tend
to be Poisson. Travel and loading/unloading time averages one hour
per call and follows an Exponential Distribution. Find the average
time patients wait for an ambulance.

Now You Try One!

The manager of a new regional warehouse of a
company must decide on the number of loading
docks to request in order to minimize the sum of
dock-crew and driver-truck costs. The manager has
learned that each driver-truck combination
represents a cost of $300 per day and that each dock
plus crew represents a cost of $1,100 per day. How
many docks should she request if trucks will arrive at
an average rate of four per day, each dock can
handle an average of five trucks per day, and both
rates are Poisson?

Now You Try One!

The manager of a new regional warehouse of a
company must decide on the number of loading
docks to request in order to minimize the sum of
dock-crew and driver-truck costs. The manager has
learned that each driver-truck combination
represents a cost of $300 per day and that each dock
plus crew represents a cost of $1,100 per day. An
employee has proposed adding new equipment that
would speed up the service rate to 5.71 trucks per
day. The equipment would cost $100 per day for each
dock. Should the manager invest in the new
equipment?

In Practice
As you might guess, no one is sitting in corporate
backrooms cranking out these equations. Instead,
sim is often used

Arena
Basic licensing: $2,495 / seat
Student license: FREE!!
https://
www.arenasimulation.com/acad
emic/students

ExtendSim
Basic licensing:

Psychology of Queues

Unoccupied Time Feels Longer than Occupied Time

Pre-Process Waits Feel Longer than In-Process Waits
Anxiety Makes Waits Seem Longer
Uncertain Waits Are Longer than Known, Finite
Waits
Unexplained Waits Are Longer than Explained Waits
Unfair Waits are Longer than Equitable Waits
Maister, D.H. (1984). Psychology of Waiting Lines,
Harvard Business Review.

Psychology of Queues
Keep progress moving. Were often satisfied as long as
were getting closer to the goal, even if its slowly.
Avoid designing queues where the consumers move
away from the service desk.
If necessary, do it as late as possible in the queue
Get consumers in the system ASAP. Consider a tiered
(or multi-phase) service
Social Justice: FIFO is viewed as being most equitable
but dont always happen (think about multiple queues
each associated with an independent server where one
moves faster than the other)
Soman, D. (2013). The Waiting Game: The Psychology of
Time and its Effects on Service Design, Rotman Magazine.

Strategies
Know consumer expectations! How long will
consumers accept waiting in line?
Keep service rates consistent when possible.
Distract them (music, televisions, magazines, etc.)
Keep customers informed
Self-service
Staff not directly involved in serving customers
should be kept out-of-sight
Stay friendly!

Before next class.

Review the supplemental chapter on Queuing
Theory (7S) from McGraw Hill Connect.
Let me know if you have questions!!!
Try the following problems from the end of the
supplemental chapter:
Problem 1
Problem 7
Problem 17
Problem 18
Read Chapter 10 in preparation for next class.