You are on page 1of 39

Chapter 13:

Capacity Planning
and Queuing Models
Learning Objectives

• Discuss the strategic role of capacity planning.


• Describe a queuing model using A/B/C notation.
• Use queuing models to calculate system performance
measures.
• Describe the relationships between queuing system
characteristics.
• Use queuing models and various decision criteria for
capacity planning.
Capacity Planning
• Capacity is the ability to deliver service over a
particular time period.
• Capacity is determined by the resources available to
the organization, e.g. facilities, equipment, and labor.
• Capacity planning is the process of determining the
type and amounts of resources that are required to
implement an organization’s strategic business plan.
• The capacity planning decision involves a trade-off
between the cost of providing a service and the cost
of inconvenience of customer waiting.
Economic Trade-off in Capacity
Planning
Queuing System Cost Tradeoff
Let: Cw = Cost of one customer waiting in queue for an
hour
Cs = Hourly cost per server
c = Number of servers

Total Cost/hour = Hourly Service Cost +


Hourly Customer Waiting Cost
= Cs c + Cw Lq

Note: Only consider systems where c . 


Capacity Planning Challenges
• Inability to create a steady flow of demand to fully
utilize capacity.
• Idle capacity always a reality for services.
• Customer arrivals fluctuate and service demands also
vary.
• Customers are participants in the service and the
level of congestion impacts on perceived quality.
• Inability to control demand results in capacity
measured in terms of inputs (e.g. number of hotel
rooms rather than guest nights).
Strategic Role of Capacity Decisions
• Using long range capacity as a preemptive strike
where market is too small for two competitors (e.g.
building a luxury hotel in a mid-sized city).
• Lack of short-term capacity planning can generate
customers for competition (e.g. restaurant staffing).
• Capacity decisions balance costs of lost sales if
capacity is inadequate against operating losses if
demand does not reach expectations.
• Strategy of building ahead of demand is often taken to
avoid losing customers.
Capacity Planning for Cookies and
Cream - Naive Approach
• An enterprising student is considering opening a “Cookies and
Ice Cream” shop in space that has become available in a food
court. Observations of traffic during lunch hour suggest a
potential peak demand of 50 customers, each ordering an
average one ice cream sundae, 6 baked-to-order cookies, and
a soft drink and spending 20 minutes at a table.
• A cookie sheet can accommodate a dozen (12) cookies and
baking time 10 minutes. One server requires on average 6
minutes to take an order, mix a batch of cookies, etc.
• Capacity requirements are determined by calculating the units
of facility, equipment, and labor needed to accommodate the
anticipated peak demand.
Capacity Planning for Cookies and
Cream - Naive Approach
• Facility requirements include the seats necessary to
accommodate diners. We will use a relationship
called “Little’s Law”.
• “Little’s Law” states that the average number of
customers in a system (L) is equal to the arrival rate
(λ) times the average waiting time (W); L= λW.
• With 50 customers arriving during the peak and each
staying approximately 20 minutes, we need (50)(20)/
(60)=16.7 chairs.
Capacity Planning for Cookies and
Cream - Naive Approach
• Equipment Requirements include the calculations of
the number of cookie sheets needed.
• This is determined by dividing the total number of
cookies ordered for the hour by the capability of one
cookie sheet that is used for only 10 minutes per
batch (i.e., reused 6 times per hour).
• Assume that orders can be combined to fill a sheet.
(50 cust/hr)(6 cookies/order)
Sheets needed= =4.17
(12 cookies/sheet)(6 cycles/hour)
Capacity Planning for Cookies and
Cream - Naive Approach
• Labor Requirement is focused on calculating
the number of servers required.
(50 cust/hr)(6 min each)
Servers needed = = 5.0

60 min/hr
Capacity Planning for Cookies and
Cream - Naive Approach
• Caution needs to be exercised in
implementing such an approach.
• Excess capacity is required in a service system.
• Capacity to serve must exceed the arrival rate
to avoid out-of-control waiting lines.
• Our capacity planning analysis of Cookies and
Cream will be revisited with a more
sophisticated queuing analysis after our
discussion of queuing models.
Analytical Queuing Models
• A popular system classifies parallel-server
queuing models using the following notations
in which three features are identified: A/B/C
– A: the distribution of times between arrivals
– B: the distribution of service times
– C: the number of parallel servers
Elements of Queuing Systems
Processing
order

Arrivals Waiting Service Exit


Line (Queue)

System
Classifications of Queuing Models
Service Performance Measures
• Average number of customers waiting (in the queue
or in the system)
• Average waiting time of customers (in the queue or
in the system)
• Utilization (percent capacity used)
• Probability of delay: probability that an arrival
customer will have to wait
• Service level: fraction of customers that wait (in the
queue) less than a specified threshold time
Notation
•  = arrival rate
•  = service rate
• c = number of servers
•  = utilization (mean number of customers in service)
• Lq = average number of customers in the queue
• Ls = average number of customers in the system
• Wq = average Waiting time in the queue
• Ws = average Waiting time in the system
• Pn= probability of n customers in the system
Basic (Steady State) Relationships

• Utilization:   (should be < 1)
c

• Average Number in Service: r 

• Average Number in the System: Ls Lq  r

• Average Time Customers


1
Waiting in the System: Ws Wq 

Basic (Steady State) Relationships
• Little's Law
Lq Wq
Ls Ws
1
r 

Standard M/M/1 Model
• Assumptions
– Calling population: an infinite or very large population of
callers arriving. The callers are independent of each other
and not influenced by the queuing system.
– Arrival process: Poisson distribution of arrival rate.
– Queue configuration: single waiting line with no restriction
on length and no bulking or reneging.
– Queue discipline: FCFS
– Service process: one server with exponential distribution
of service times.
Standard M/M/c Model
• Assumptions for the standard M/M/c model
are the same as those for the standard M/M/1
model with the stipulation that service rates
across channels be independent and equal ρ
now must be less than c, the number of
servers, for steady state results to occur.
Two Standard Models
• M/M/1

• M/M/c
Congestion
Queuing ToolPak in Excel
http://queueingtoolpak.org/
Use template.xls with qtp.xll!
•Average server utilization
•Average number of customers in the system
•Average number of customers in the queue
•Average time a customer spends in the system
•Average time a customer spends in the queue
•Service level: fraction of customers that wait (in the queue) less
than a specified threshold time
•Minimum number of servers needed to provide a specified level
of service
Example
• Quick Lube Inc. operates a fast lube and oil change garage. On a
typical day, customers arrive at the rate of three per hour, and
lube jobs are performed at an average rate of one every 15
minutes. The mechanics operate as a team on one car at a time.
Assuming Poisson arrivals and exponential service, find:
• Utilization of the lube team.
• The average number of cars in the queue.
• The average time a car waits before it is lubed.
• The average total time in system.
• The fraction of customers that wait less than half an hour
Example
•  =3 / hour,  = 4 / hour, c = 1

• Utilization  =  /  = 75%
• From the ToolPak in Excel: Lq = 2.25 cars
• From Little's Law: Wq = Lq /  = 0.75 hour
• From the basic relationship: Ws = Wq + 1 /  = 1 hour
• From the ToolPak in Excel: service level = 3%
Example
• Suppose Quick Lube Inc. would like to ensure 90% of customers
that wait less than half an hour.
• What is the minimum number of lube teams needed?

• From the ToolPak in Excel, at least 2 lube teams.


Example
• Suppose Quick Lube Inc. believes that there is a loss-of-goodwill
cost associated with the customers waiting in the queue. This
cost is estimated as $50/hour/customer. The cost for each lube
team is $40/hour.
• What is the optimal number of lube teams to minimize the total
cost?
Example
• One team (C = 1)
Total waiting cost = 50 * Lq = $112.5 /hour
Total labor cost = 40 * C = $40 /hour
$152.5 /hour
• Two teams (C = 2)
Total waiting cost = 50 * Lq = $6.2 /hour → Optimal!
Total labor cost = 40 * C = $80 /hour
$86.2 /hour
• Three teams (C = 3)
Total waiting cost = 50 * Lq = $0.7 /hour
Total labor cost = 40 * C = $120 /hour
$120.7 /hour
Capacity Pooling
• Waiting time in the system is less for M/M/2
with  and  than for M/M/1 with  and (/2).
Limitation
• Mathematical analysis becomes very complicated or
intractable for more complex waiting lines. Some
examples:
– Non-Poisson arrivals
– Non-exponential service times
– Complex customer behavior (e.g. customers switching
between lines, or leaving after some time etc.)
– Multiple phase systems
• Simulation can be useful analysis tool!
M/G/1 Model
• For this model, any general service time
distribution with mean E(t) and variance V(t)
may be used.
• The condition that ρ be less than1 still applies
for the steady state, where ρ now equals λE(t)
• Except for the generality of the service time
distributions, all the assumptions for the
standard M/M/1 model apply.
General Self-Service M/G/∞
• If a multiple-server system has an infinite
number of servers or arrivals serve
themselves, then no arriving customer must
wait for service.
• This also describes exactly the concept that
has made the modern supermarket so
popular.
Finite-Queue M/M/1 Model
• A modification of the standard M/M/1 model can be
made by introducing a restriction on the allowable
number of customers in the system.
• Suppose that N represents the maximum number of
customers in allowed, or, in a single sever model,
that N-1 indicates the maximum number of
customers in the queue.
• This particular model is very useful in estimating lost
sales to be expected from an inadequate waiting
area or excessive queue length.
Finite-Queue M/M/c Model
• This model is similar to the finite-queue M/M/1
model, with the exception that N, the maximum
number in the system, must be greater than or equal
to c.
• An arriving customer is rejected if the number in the
system equals N or the length of queue is N-c.
• All other assumptions for the standard M/M/c model
hold, except that ρ now can exceed c. Because excess
customers are rejected, the system can reach a
steady state even when the capacity is inadequate to
meet the total demand (i.e., λ=cμ).
Finite-Queue M/M/c Model
• An interesting variation on this model is the
no queue situation, which occurs when no
possibility exists for a customer to wait.
• This situation can be modeled as a finite
queue system with N=c.
• A parking lot is an illustration of this no-queue
situation.
Capacity Planning Criteria
• Average Customer Waiting Time
• Probability of Excessive Waiting
• Minimizing the Sum of Customer Waiting
Costs and Service Cost
• Probability of Sales Lost Because of
Inadequate Waiting Area
Average Customer Waiting Time
• For example, a restaurant owner may wish to
promote liquor sales in the bar, and therefore,
stipulate that customer be kept waiting 5 minutes on
the average for a table.
• Another example may be drive-in bank facility design
• In these cases , use of the M/M/c model would be
appropriate to identify the service capacity in terms
of the number of server.
Topics for Discussion
• Discuss how one could determine the economic cost of
keeping customers waiting.
• Example 16.1 presented a naïve capacity planning exercise
criticized for using averages. Suggest other reservations
about this planning exercise.
• For a queuing system with a finite queue, the arrival rate can
exceed the capacity. Explain with an example how this is
possible.
• What are some disadvantages associated with the concept of
pooling service resources?
• Discuss how one could determine the economic cost of
keeping customers waiting.

You might also like