QUEUEING THEORY

Queueing Theory Problems

PROBLEM 1
The counter of a bank branch performs the transactions with a mean time of 2
minutes. The customers arrive at a mean rate of 20 customers/hour. If we assume that
arrivals follow a Poisson process and that the service time is exponential, determine:

a) Percentage of the time the bank teller is idle

b) Mean waiting time of the customers
c) Percentage of customers that wait in a queue

PROBLEM 2
One person attends a grocery shop. The arrival pattern of customers on Saturdays
follows a Poisson process with an arrival rate of 10 persons/hour. Customers are attended
to following a FIFO order, and due to the shop's prestige, they are willing to wait for the
service once they arrive. The service time is distributed exponentially, with a mean time
of 4 minutes. Determine

a) Probability of waiting in line

b) Average length of the waiting line
c) Average waiting time

PROBLEM 3
Customers arrive at a hair salon with a mean rate of 5 customers/hour, and inter-
arrival times are distributed exponentially. There is always one hairstylist in the salon at
any moment, and there are 4 chairs for the customers who arrive when the hairstylist is
busy. The law regarding fire prevention limits the total number of customers in the salon
at any moment to a maximum of 5. Customers who arrive when the salon is complete
can’t enter it. Service time is distributed exponentially with a mean that depends on the
number of customers. Determine:

a) Average number of waiting customers

b) Average waiting time
c) Percentage of time that the hairstylist is idle

Number of customers in the salon 1 2 3 4 5

Mean service time per customer (min.) 9 10 10 13 20

PROBLEM 4
In a factory, there is a Social Security office where the workers can go during working
hours. The Chief of Staff, who has observed the inflow of workers to the office, has asked
to perform a study of this service. A specialist has been appointed to determine the
workers' average waiting time and the average conversation duration with the office

November 2019 1
QUEUEING THEORY

employee. This analyst concludes that the inflow is significantly reduced and fluctuates in
the shift's first and last half hours. Still, during the remaining seven hours, the
phenomenon can be considered in a steady state. From the analysis of 100 periods of 5
minutes, consecutive or not, but in the steady state phase, it can be deduced that the mean
number of workers going to the office was 1.25 per period, and the inter-arrival time
followed an exponential distribution. A similar study about the duration of the
conversations concluded that its distribution was exponential, with a mean duration of
3.33 minutes. Determine:

a) Average number of workers waiting

b) Average waiting time
c) Compare the time spent by the workers and the time spent by the office employee.
Determine the cost for the company if an hour for the office employee is 250€ and
one hour for the worker is 400€. Would it be profitable to open another office?

PROBLEM 5
Let us assume that the Iberia fleet has four planes of the type Airbus A350. Since
2015, the turbine failures of these planes have been observed. Data indicate that the failure
of any turbine in any plane is a random variable with exponential distribution, and the
mean time to failure for any turbine is one year. The mean time to repair the turbine
failure is only 45 days (one-eighth of a year) and follows an exponential distribution. Only
a single expert team can fix them, and they provide the service under the policy “first in
the workshop, first to be repaired.” In the repair period, the plane doesn’t fly. Describe
quantitatively the queueing system.

PROBLEM 6
A bank is considering installing a teller network in one of its branches. Given that
they don’t know the people inflow demanding the service, they only put a single teller for
a month. The data about the arrival time of the customers, as well as the service time, is
collected daily. If the branch is in a district with no such service, the customer who arrives
waits until they can use the teller when it is busy. After the corresponding data analysis,
it is estimated that:

1. The arrivals follow a Poisson process

2. The distribution of the service time is exponential
3. The mean time between consecutive arrivals is 7.5 minutes
4. The mean service time is 5 minutes/customer

Determine:
a) Mean waiting time that each customer must spend in the queue
b) Average number of waiting customers
c) Average queue length when it is not empty and the probability that when going to the
teller, there will be somebody in the queue

November 2019
 QUEUEING THEORY

PROBLEM 7
Two technicians can repair up to four machines in a workshop. Machines arrive at
the workshop for being repaired once every three hours on average and the mean repairing
time is 45 minutes. If both times follow an exponential distribution, what is the average
number of machines broken and waiting to be repaired in a steady state?

PROBLEM 8
A bank branch has two office employees who are equally efficient and capable of
attending 60 operations/hour on average, with the service time distributed exponentially.
Customers arrive at the bank branch following a Poisson process with a mean rate of 100
customers/hour. Determine:

a) Probability of strictly more than 3 customers simultaneously in the bank branch

b) Probability that at least one of the employees will be idle

PROBLEM 9
A self-service car wash has four sections. In each one, the customers can wash and
wax their cars. There is only space for three additional cars when the washing sections
are complete. Customers arrive at the service following a Poisson process with a mean
rate of 15 cars/hour. If there is no room for waiting, cars must leave the service. The
time required to service a car is distributed exponentially, with a mean of 12 minutes.
Determine:

a) Average number of cars in the service at any moment

b) Average number of cars per hour that are rejected because the service is full

PROBLEM 10
The road from B to A is split into five migration and customs inspection booths on
the border between countries A and B. Let’s assume that car arrivals have a Poisson
distribution with 15 arrivals/hour, while the number of services follows an exponential
distribution of 8 services/hour.
By Government decree, there is no treatment priority (is that credible?), so the
inspection booths provide the service as soon as they are free, attending to the cars by
order of arrival to the queue.
Check if 𝜆 < 𝜇, and interpret this inequality in a positive case. Moreover, determine:

a) Compute 𝑝 , 𝐿, 𝐿 , 𝑊 , 𝑊, defining its meaning

b) Study if it would be reasonable to suppress three booths

PROBLEM 11

November 2019 3
QUEUEING THEORY

The factory workers must carry their work to the quality control department before
the product goes to the final production process. There are many employees, and the
arrivals are approximately 20 workers/hour, following a Poisson process. The time to
inspect a part follows an exponential distribution with a mean of 4 minutes. Determine
the average number of workers in the quality control department if there are: a) 2
inspectors, b) 3 inspectors.

PROBLEM 12
Consider a queue with a single server and Poisson arrivals with 10 arrivals/hour on
average. Usually, the server works with an exponential distribution with a mean service
time of 5 minutes. The management has a training course that achieves an improvement
(decrease) in the service time variance at the expense of a small increase in the mean.
After finalizing the study, it is estimated that the mean service time will increase by 0.5
minutes, and the standard deviation will decrease by 5 to 4 minutes.

a) The management would like to know if they should continue with the employee as is
or if they should continue with the new training course. Discuss the results
b) Compute the variance reduction needed to counteract an increase of 0.5 minutes of
the mean

PROBLEM 13
The manufacturer of a specialized electric conductor has a large manufacturing
volume. The conductor is subject to a single mechanical manipulation by a specialized
machine. Given the sale volume, the company has a lot of equal machines working all the
time (let’s assume that the number of machines is infinite). Machines break following a
Poisson process with a mean of 5 machines/hour. The company has a single repair
technician, and the failures are of two possible types due to the characteristics of the
machines. Depending on the failure type, it can be repaired in 9 or 12 minutes.
As the technician is an expert and the machines are identical, the possible variation
in repair times is negligible and can be ignored. The type of cause that produces the failure
is distributed randomly, but one-third of the failures need 12 minutes to be repaired.
Determine the average number of failed machines.

PROBLEM 14
An airplane takes almost 4 minutes to land from when the control tower gives it the
landing signal. If the airplane arrivals are 8/hour on average and follow a Poisson process,
how much time can the pilot expect to fly around the airport until receiving the landing
signal?

PROBLEM 15
In the emergency unit of a hospital, emergency cases arrive with a Poisson distribution
of once every 5 hours on average. Before any surgical operation or other intervention,

November 2019
 QUEUEING THEORY

each patient must follow a medical exam and a disease diagnosis. This process takes 15
minutes (almost exactly). Compute (assuming the system is in a steady state):

a) Average number of patients that have not been operated yet

b) Average time a patient waits until entering the operating room
c) Average time a patient waits until being examined
d) Average number of patients waiting to be examined

November 2019 5
QUEUEING THEORY

Additional Queueing Theory Problems

PROBLEM 16
A computer company has a server, and customers can access this server from
terminals (at different times) that they rent. A customer needs to determine the optimal
terminal speed at which to rent. Customer works are generated as a Poisson process with
a rate of 50 works per day of 8 hours. The average size of the work is 1000 instructions.
It is known that the processing time of the instructions is exponential. The customer
estimates 10€ as the cost of delaying work for one day.
The company estimates that a speed of 100 instructions/minute and any similar
increase will increase the monthly rent price of the terminal by 100 €. Assume that a
month has 22 working days. Determine the optimal terminal speed.

PROBLEM 17
Consider that a workshop can hire one out of two possible technicians. The first
technician earns 3 € an hour and can repair 5 machines/hour. The second one earns 5€
an hour but can repair 8 machines/hour. Both salaries are paid no matter if they are
busy or idle. Customers estimate that a broken machine costs 8€/hour, and the workshop
assumes it is a cost per unit of waiting time per customer. Assume that machines arrive
at the workshop following a 4 machines/hour Poisson process, and the repair time follows
an exponential distribution. Which technician would be more economical for the
company?

PROBLEM 18
A company that has four machines that are frequently broken hires a person devoted
exclusively to repairing them. The repair time follows an exponential distribution of 2
hours on average. Once the machine has been restored, the time to failure again follows
an exponential distribution of mean 10 hours. When the machine is under repair, the
wasted time is 20 €/hour, and the repair service costs 45 euros daily. Assume a working
day of 8 hours.

a) Compute the steady-state distribution

b) Compute the average number of operating machines in a steady state
c) What is the average cost per day of the wasted time?
d) Would it be profitable to have two technicians attending two machines?
e) Would it be profitable to have two technicians, each one attending the four machines?

PROBLEM 19
A small company which manufactures computer devices has 10 machines for this task.
Machines are frequently broken, but are 8 persons in charge of operating them. So, two
of them are in reserve to be used when some of the active ones fail. Therefore, eight will
be used when there are no more than two broken machines.

November 2019
 QUEUEING THEORY

The time to failure is exponential, with a mean of 20 days. The repair time is
distributed exponentially, with a mean of 2 days.
Until now, the factory had had only a technician who repaired the broken machines
with the corresponding effect on production when fewer than 8 available machines were
available. Therefore, they are thinking about hiring another technician to be able to repair
two machines at the same time.
Each technician costs 70 €/day, but the profit loss estimated for productivity reduction
per each of the 8 machines not working is 100 €/day.

PROBLEM 20
As needed, a railway company paints its wagons in its workshops where each wagon
is manually and individually painted with a speed exponentially distributed with a mean
of one every 4 hours and an annual cost of 4 million €. Wagons arrive with a Poisson
process of mean one every 5 hours. Besides, the wagon costs 500 euros an hour when it
is inactive.
They are thinking about two possibilities. One is to subcontract the process to a
painting company that would use aerosol spray, which would save time. However, the
budget for this alternative is 10 million € per year. In this case, the process can be
approximated by a Poisson process with a rate of one every 3 hours.
The other option is to install a new workshop precisely like the existing one, with an
equal service rate and annual cost that allows painting two wagons simultaneously.
In all the cases, work is considered uninterrupted, i.e., they work 8760 hours a year.
Which of the three procedures is preferable?

PROBLEM 21
A radio repair company receives an average of 10 daily repair demands, distributed
according to a Poisson process. Assume that µ is the repair rate of the repairing machine
in radios/day and the repair time is exponential. Each unit of repair rate has a cost of
100 €/week.
Moreover, it has been estimated that the cost of having unrepaired radios is 200 € per
radio per week, with this cost proportional to time. Assuming a week has 5 working days,
determine:

a) Optimal repair rate

b) If it would be more economical to have two machines, each one with half the rate
determined in the previous question

PROBLEM 22
A dentist has a consultation room where patients arrive according to a Poisson
process with a rate of 5 patients/hour. The waiting room has three chairs, and no patients
are allowed if no seats are available. The average time the dentist pays attention to each
patient is exponentially distributed, but its value depends on the number of patients in the
consultation room:

November 2019 7
QUEUEING THEORY

Number of patients 1 2 3 4
Mean service time (minutes) 9 10 12 15

Give the steady-state probabilities of the number of patients in the consultation room.

PROBLEM 23
A company's accounting department has a computer system with several PCs in the
same domain; a PC acts as a printing server (which processes the works by first in-first
out criterion without any priority) and two equal printers (with an unlimited buffer).
Department members have applied (complaint) to management to improve the printing
service, claiming that the printers are saturated at the end of the month (when they close
the balance sheet).
After making the corresponding observations at the end of the month and analyzing
the data obtained, they have extracted the following conclusions:

 Time between printing orders from the server follows an exponential distribution of a
mean 5 minutes
 The processing time of the server follows an exponential distribution of a mean 1
minute
 Printing time follows an exponential distribution of a mean of 7.5 minutes

Answer the following questions:

1. If a steady state is reached, what would be the arrival rate of the printing works?
2. Let’s eliminate the printing server. Does the system have a steady state distribution?
Why?
3. Continue without the server, which is the average number of works pending at the end
of the day?
4. In the same conditions as the previous question, if they remain at work doing extra
hours until the printers end (without sending additional work), how many hours will
need to be paid on average?
5. In the same conditions as the previous question, if we consider that while work is
being printed, the employee is idle and, therefore, it is wasted time and the employee
hour costs 20 €, which cost per hour has this system?
6. In the same conditions that the previous question, if the employee wasted time while
waiting for the work to be printed, it is considered an added cost, which cost will be
added to each work if the printed work is invoiced to a customer if on average each
work requires to be printed 10 times before being ready?

PROBLEM 24
Consider a repair workshop with a robot. Consider the possibility of changing it to
another one. Both robots are different, with different costs. The following table shows the

November 2019
 QUEUEING THEORY

data of each robot, understanding that the operation costs correspond to the costs per
unit of time when the robot is doing some repair, and amortization costs are costs per
unit of time, whether the robot is repairing or not.

Amortization costs (€/h) Operation costs (€/h)

Robot 1 100 200
Robot 2 150 400

The first robot can repair 6 machines/hour, and the second one can repair 7
machines/hour.
When a machine is in the repair workshop, consider a cost per unit of time due to
productivity loss estimated at 80 €/h, and the workshop assumes it as cost per unit or
time and customer. Assume that the machines arrive at the workshop with a Poisson
process with a rate of 5 machines/h, and the repair time required by the robots follows
an exponential distribution.
Decide analytically which of the two robots is preferable, comparing the cost per hour
with both robots and giving that cost.

PROBLEM 25
Juan Pérez is studying the possibility of opening a car wash workshop. He has received
an offer to buy a site priced at 750 €/m2 for that purpose. With his experience in this
type of business, he knows he will need a plot of 30 m2 to install the washing machine
and, besides, 12 m2 for each vehicle to be admitted in the workshop and waiting in the
queue to be washed. Cars that don’t enter the workshop will be forced to go to another
service center.
Moreover, given the knowledge about another workshop of the same type he had in the
same area and a small statistical study, he has obtained that interarrival times are
distributed with an exponential distribution of a mean of 10 minutes.
Determine:

1. If the time to wash a car is distributed with an exponential mean of 6 minutes and the
price is 4 €, determine the additional site is needed to optimize the design of the
workshop for an operation period of a year (365x12x60=262800 min), assuming a
steady state.
2. Assume there is no site limit, so no reserved site is required because the cars can wait
in front of the workshop door as much as they want, but it is estimated that 0.5
€/minute the vehicle is in the car workshop waiting for the access. Juan Pérez intends
to buy a slow machine that lasts exactly 6 minutes to wash a car and costs 36000 €; a
quick machine costs 60000 € and only requires 4 minutes to wash a car. Moreover,
vehicles to be washed can be cars or trucks. The washing time for a truck is double
that for a vehicle, and the percentage of vehicles is 65 %. What is the average queue
length for each machine? Which machine will be bought thinking on a month of
operation (262800 minutes)?

November 2019 9
QUEUEING THEORY

PROBLEM 26
A manufacturing plant has two spots, A and B. Products must go through both spots,
being processed first in A and then in B. The process in spot A takes 3 minutes on average
and in spot B 4 minutes.
Determine:

1. What is the maximum production that can be achieved per hour in the plant? If the
products arrive with an interarrival time distributed exponentially with a mean of 5
minutes, which is the hourly production assuming a steady state?
2. If, besides being exponential, the interarrival times are also the service times, which
will be the estimated time from the arrival of the product to the plant until it reaches
the queue of spot B? And if the service times were deterministic and not exponential?
3. Assuming that the interarrival time is exponential, the products are of two types, so
in spot A, 20% takes 3 minutes to process. The remaining ones are 4 minutes, and
the cost of a product's time in the system is 0.2 €, which is the total hourly cost for
the company corresponds to spot A. Which cost attributable to spot A should be
assigned to each product?

PROBLEM 27
At the beginning of the year, a toy manufacturer faces production planning for the
third trimester. From the business point of view, the third trimester has been good 75%
of the year, understanding it as good when the average daily demand is 12 units and as
bad when this demand is only 8 units. Daily demands are random, independent, and
distributed according to Poisson laws in both cases.
Historically, it is known that when the second trimester is good, the third one is also
good in 80 % of the cases, while if the second one is bad, the third one is good in 30% of
the cases.
Building a toy requires two machines, A and B, and the manufacturer has only one
machine of each type. The process in machine A consists of assembling 6 identical
components, and the machine takes a random time following an exponential distribution
of a mean of 30 minutes. The process in machine B, done after machine A, consists of
assembling 9 identical components (although of a different type than those assembled in
A), and the processing time follows an exponential distribution of a mean of 48 minutes.
The manufacturer wants to attend to all the orders for this and, if needed, rents
additional machines. If the third trimester is good, they must rent an additional machine
of type B to satisfy the demand, given that works 8 hours a day. The rent cost for the
third trimester for machine B is the following:

 If the machine is rented at the beginning of the year, the cost is 1200 €
 If the machine is rented before the end of June (when it is known if the second
trimester has been good or bad), 1500 €

November 2019
 QUEUEING THEORY

 If the machine is rented at the beginning of the third one and it is known if the third
trimester is good or bad, 2000 €

1. What is the optimal policy for the toy manufacturer regarding the possible renting of
the machines?
2. What is the average time it takes to be sent to their requestor since their request until
both machines process the toy if the trimester is bad and there is only one machine?
What if the trimester is good?

Note: if arrivals and services are Poisson-type, departs also follow a Poisson process.

PROBLEM 28
Some parts arrive at a workshop to be processed following a Poisson law with a mean
of 15 parts an hour. Parts can go through two different processes, A and B, being 2/3 of
the processes of type A. The process sequence is shown in the figure, where in spot C2,
there are 2 machines for parts type A.

Artículos
C2 Proceso A
2/3
2 máquinas
15art/h
C1

1/3
Artículos
C3 Proceso B

The process in spot C1 is the same for both parts and follows an exponential law of
mean 3 minutes. In spot C2, process time follows an exponential distribution of a mean
of 10 minutes; in C3, the processing time follows a normal distribution of a mean of 7.5
minutes and a standard deviation of 2 minutes.

a) Estimate the time required by a part type A and a part type B from its arrival until
its departure1.
b) It is estimated as 0.1 € each minute the time spent by any part in the system; which
hourly cost will the system cost? Which cost must be assigned to each part by the
process?

PROBLEM 29

1
Consider that if the arrivals follow a Poisson process and service times are exponential, then
departures also follow a Poisson process.

November 2019 11
QUEUEING THEORY

An entrepreneur wants to install a motorbike repair workshop in a summer zone from

June 15 until September 15. The rented premises have a capacity of 4 motorbikes, either
being repaired or waiting to be repaired, so any biker who arrives at the workshop and
finds it full must go to another workshop.
From June 15 until July 15, the arrivals of failed motorbikes follow a Poisson law with
a mean of 12 motorbikes daily. Since July 16, this mean can increase to 24 motorbikes
with a 0.75 probability or reduce to 6 with a 0.25 probability.
Repairing a motorbike takes a random time governed by an exponential law of mean
2 hours. The workshop Works 24 hours a day in three shifts of 8 hours each.
Regarding the number of technicians to hire for repairing, the entrepreneurship has
the following options:

1. Hire on June 15 for the season a set of three technicians (one for each shift) with a
monthly cost of 15000 € or two sets of three technicians (two for each shift) with a
monthly cost of 30000 €.
2. Hire on June 15 for all the season a set of three technicians (one for each shift) with
a monthly cost of 15000 €. For the last two months of the season, but signing the
contract on July 16, before knowing the evolution of the arrival of failed motorbikes,
we can hire a second set of three technicians that will cost 25000 € a month.

If the mean profit for each motorbike is 65 € and using the expected value criterion,
which hiring policy must be recommended to the entrepreneur? Assume months of 30
days.
How much would the entrepreneur be willing to pay to know before July 15 the average
failed motorbikes until the end of the season?

PROBLEM 31
A section of the local Administration has two offices, A and B. Users go to office A,
where there is one clerk, following an arrival pattern governed by a Poisson law of 12
people/hour, and the attention time follows an exponential law of mean of 4 minutes.
Users then go to office B, where the attention time also follows an exponential mean
of 7.5 minutes for each clerk.
How many clerks will be needed in office B if you want the probability of waiting more
than 10 minutes to be lower than 0.05?
Being 𝑆 the number of servers in an M/M/S system, the probability of not having
people in the system is:

= ,𝑝 = ( ) ( )
)
∑
!( !

The probability of waiting for 𝑡 time units greater than 𝑥 is given by:
(𝑆 )
𝑃(𝑡 > 0) = 𝑝
𝑆! (1 − )

November 2019
 QUEUEING THEORY

P tc  x   e S x 1P tc  0

November 2019 13
QUEUEING THEORY

Solutions to Queueing Problems

SOLUTION OF PROBLEM 1
This is an M/M/1 system with an infinite population, with lambda = 20 customers/hour
and 𝜇 = 30 customers per hour.
a) P (idle b ank teller)  p0  1 -   1 / 3. 33.33 % of the time is idle.
b) 𝑊 = 1/15 = 4 minutes
c) 𝐿 = 2, 𝐿 = 4/3. Fraction in queue = 2/3. 66.67 % of the system customers are
waiting in the queue.

SOLUTION OF PROBLEM 2
This is an M/M/1 system with an infinite population, with lambda = 10 people/hour and
𝜇 = 15 people per hour.
a) P (waiting line)  1  p0  p1  4 / 9
b) 𝐿 = 4/3 people in the queue.
c) 𝑊 = 2/15 hours = 8 minutes on average in the queue.

SOLUTION OF PROBLEM 3
It is an M/M/1/5 system where the arrival rate is constant 𝜆 = 5 customers/hour,
and the service rate depends on the number of customers in the hairdresser, where 𝜇 =
6. 6̑ customers/hour or 𝜇 = 3 customers/hour.
𝐶 = . ̑
= 0.75, 𝐶 = . ̑⋅
= 0.625, 𝐶 = . ̑⋅ ⋅
= , 𝐶 = . ̑⋅ ⋅ ⋅ .
= , 𝐶 =

. ̑⋅ ⋅ ⋅ . ⋅
=

𝑝 +𝑝 +𝑝 +𝑝 +𝑝 +𝑝 = 1

𝑝 +𝐶 𝑝 +𝐶 𝑝 +𝐶 𝑝 +𝐶 𝑝 +𝐶 𝑝 =1

𝑝 (1 + 𝐶 + 𝐶 + 𝐶 + 𝐶 + 𝐶 ) = 1
1 1
𝑝 = = = 0.2272
∑∞ 𝐶 1 + 0.75 + 0.625 + 25/48 + 325/576 + 1625/1728

𝑝 = 𝐶 𝑝 = 0.1704, 𝑝 = 𝐶 𝑝 = 0.1420, 𝑝 = 𝐶 𝑝 = 0.1184, 𝑝 = 𝐶 𝑝 = 0.1282,

𝑝 = 𝐶 𝑝 = 0.2137
a) Average number of people in the system 𝐿 = ∑∞ 𝑛𝑝 = 2.3909 customers. (In this
case here: L=0*p0+1*p1+2*p2+3*p3+4*p4+5*p5).
b) Average waiting time 𝑊 = = 0.4116 hours being 𝜆 = 𝜆(1 − 𝑝 ) =
.
= .
5(1 − 0.2137) = 3.9315 customers/hour and 𝐿 = ∑ (𝑛 − 1)𝑝 = 1 ⋅ 0.1420 + 2 ⋅
0.1184 + 3 ⋅ 0.1282 + 4 ⋅ 0.2137 = 1.6182 customers.
c) Hairstylist is idle with a probability 𝑃 = 0.2272.

SOLUTION OF PROBLEM 4

November 2019
 QUEUEING THEORY

This is an M/M/1 system with infinite population, with 𝜆 = 0.25 workers per minute
(1.25/5) and 𝜇 = 0.3 workers per minute.
a) 𝐿 = 4.16̑ workers.
b) 𝑊 = 16. 6̑ minutes.
c) The worker attention cost is approximately 14000 €/day (7 hours/day*5 workers*400
€/hour). The inflow is reduced during the first hour of the shift. The L (number of
workers in the system) is 5. Office employee cost is 2000 €/day. If another employee
is working (model M/M/2), the worker cost is 2823.53 €/day (L=1.0084), and of the
employees, 4000 €/day. It is profitable to hire two employees for the system.

SOLUTION OF PROBLEM 5
This is an M/M/1 system with finite population m=4, with 𝜆 = 1 plane per year and 𝜇
= 8 planes per year.
It is a close system type M/M/1 with population size 𝑚 = 4 planes with individual
arrival rate of each plane 𝜆 = 1 plane/year and departure rate 𝜇 = 8 planes/year. The
utilization factor is
𝜆 1
𝜌= =
𝜇 8
The probabilities of each state can be computed by means of these expressions

𝑚! 𝜌 4! 4! 4! 4!
𝑝 = 1+ = 1+ + + + =
(𝑚 − 𝑛)! 3! 8 2! 8 1! 8 0! 8
1 3 3 3
= 1+ + + + = 0.5746
2 16 64 512

𝑝 = (𝑚 − 𝑛 + 1)𝜌𝑝

p1  4 p0  0.2873 ; p2  3 p1  0.1077 ; p3  2p2  0.0269 ; p4   p3  0.0034

The average number of planes in the system is
1  p0
L m  4  (1  0.5746)8  0.5971 aviones

The average rate of aircraft arriving in the system EF  (m  L)  3.4029 .
The average number of queued aircraft waiting to be repaired is

1  9
Lq  m  (1  p0 )  4  8 (1  0.5746)  0.1717 aviones
 1
8
Which could also have been calculated as
EF 3.4029
Lq  L   0.5971  =0.1717 aviones
 8
The average time of aircraft in the system is

November 2019 15
QUEUEING THEORY

L 0.5971
W    0.1678 años  61.2 días
(m  L) (4  0.5971)1
The average time of aircraft queuing is
Lq 0.1717
Wq    0.0501 años  18.3 días
(m  L ) (4  0.5971)1

SOLUTION OF PROBLEM 6
This is an M/M/1 system with an infinite population, with lambda = 8 customers per
hour and 𝜇 = 12 customers per hour.
a) 𝑊 = 0.16̑ hours = 10 min.
b) 𝐿 = 1. 3̑ customers.
c) When the queue is not empty, the average queue length is calculated by 𝐿 /prob(queue
not empty). This probability = 1- p0 – p1. Finally, the queue length (when it is not
empty) is 3 people. And the probability that there are at least 2 people in the system
is 0. 4̑.

SOLUTION OF PROBLEM 7
This is an M/M/2 system with finite population m=4, with 𝜆 = 1/3 machine per hour
and 𝜇 = 4/3 machines per hour.
L=0.2537 machines in the workshop.

SOLUTION OF PROBLEM 8
This is an M/M/2 system with infinite population, with 𝜆 = 100 customers per hour, and
𝜇 = 60 customers per hour
a) 1 − 𝑝 − 𝑝 − 𝑝 − 𝑝 = 0.5261.
b) P (al menos un cajero ocioso)  0.24º .

SOLUTION OF PROBLEM 9
a) L=3.285.
b) 0.96 is the average number of vehicles that cannot enter the service.

SOLUTION OF PROBLEM 10
a) 𝑝 =0.1526, 𝐿 =0.0283, L=1.9033, 𝑊 =0.0019 hours=7 s, W=0.1269 hours =
7min.36 s
c) 𝑝 =0.03226, 𝐿 =13.61, L=15.49, 𝑊 = 0.90733 hours=54.44min., W=1.03 hours =
61.8min.

SOLUTION OF PROBLEM 11
a) L=2.4 employees in the Quality Control Department.
b) L=1.4778 employees in the Quality Control Department.

SOLUTION OF PROBLEM 12

November 2019
 QUEUEING THEORY

The first system is an M/M/1 system with infinite population, with 𝜆 = 10 customers per
hour and 𝜇 = 12 customers per hour. The second system (with the course) corresponds
to an M/G/1 system.
a) Current employee: L = 5 customers, W = 30 minutes
Employee with the course: L = 8.625 customers, W = 51.750 minutes. The results
are more sensitive to the average service time than to the variance.
b) It is impossible to compensate for the 0.5-minute increase in service time by reducing
the variance. Even if the variance were 0, the resulting L would still be higher than 5,
so only with a negative variance (which does not exist) could the same L=5 be
obtained).

SOLUTION OF PROBLEM 13
L=2.9583 machines.

SOLUTION OF PROBLEM 14
𝑊 = 2.28 minutes.

SOLUTION OF PROBLEM 15
a) L= 39/760 patients.
b) W= 15 min. and 23 s
c) 𝑊 = 23 s
d) 𝐿 = 1/760 patients.

SOLUTION OF PROBLEM 16
The cost per job per unit increment of (100 statements per minute) per unit of time
(day) 𝜇 = 0.05 m.u.
The cost of delaying a job by one day is 10 m.u.
𝜆 = 50 jobs per day
From this, it can be deduced that the optimum miss is 123 jobs per day, that is, 256
sentences per minute.

SOLUTION OF PROBLEM 17
Assuming he was alone, the first employee's cost is 35 m.u./hour. The cost of the
second employee is 13 m.u./hour. The second is cheaper.

SOLUTION OF PROBLEM 18
a) It is a closed system of M/M/1 type queues with font size m  4 máquinas with
individual arrival rate of each machine   0.1 máquinas/hora and output rate
  0.5 máquinas/hora . The utilization factor is
b)
𝜆
𝜌= = 0.2
𝜇
The probabilities of each state can be calculated using these expressions

November 2019 17
QUEUEING THEORY

 m !  n 
1 1
m
 4! 4! 4! 4! 
p0  1  
  1      
 
n 1 (m  n )!   3!5 2!52 1!53 0!54 

 24 
1
4 12 24
 1       0.3983
 5 25 125 625 

𝑝 = (𝑚 − 𝑛 + 1)𝜌𝑝

p1  4 p0  0.3187 ; p2  3p1  0.1912 ; p3  2p2  0.0765 ; p4   p3  0.0153

1. The average number of machines in the system is
1  p0
L m  4  (1  0.3983)5  0.9917 máquinas

Then the number of operating machines is approximately 3.
The average rate of arrival of machines into the system EF  (m  L)  0.3008 .
The average number of machines in the queue waiting to be repaired is
EF 0.3008
Lq  L   0.9917  =0.3900 máquinas
 0.5
The average time of machines in the system is
L 0.5971
W    0.1678 años  61.2 días
(m  L ) (4  0.5971)1

and the average time of machines in the queue is

Lq 0.1717
Wq    0.0501 años  18.3 días
(m  L ) (4  0.5971)1

c) The average total cost per day is 20  8  0.9917  45  203.67 u.m./día

1. Having two mechanics, it is equivalent to two closed systems of M/M/1 fountain

queues with size arrival rate   0.1 máquinas/hora and output rate
  0.5 máquinas/hora . The utilization factor remains 𝜌 = 0.2
The probabilities of each state can be calculated using these expressions
1
 m
m ! n   2! 2! 
1

p0  1    1     0.675
»
 
n 1 (m  n )!   1!5 0!52 


p1  2 p0  0.270
» ; p2   p1  0.135
»

The average number of machines in the system is

November 2019
 QUEUEING THEORY

1  p0
L m  2  (1  0.675)5
»  0.378
» máquinas

The average total cost per day is 2(20  8  0.378
»  45)  211.08 u.m./día

1. In the case of having two mechanics for the four machines, it is equivalent to a
closed system of M/M/2 source queues with size m  4 máquinas and arrival
rate   0.1 máquinas/hora and output rate   0.5 máquinas/hora . The
utilization factor is now 𝜌 = = 0.1. Probabilities can be calculated from the
expression

⎧ 𝑚 𝜆
𝑝 1≤𝑛≤𝑐
𝑛 𝜇
𝑝 =
⎨ 𝑚 𝑛! (𝜆/𝜇)
⎩ 𝑛 𝑐! 𝑐 𝑝 𝑐≤𝑛≤𝑚

𝑝 = 0.8𝑝 ; 𝑝 = 0.24𝑝 ; 𝑝 = 0.048𝑝 ; 𝑝 = 0.0048𝑝

Which results 𝑝 = 0.4778, 𝑝 = 0.3823, 𝑝 = 0.1147, 𝑝 = 0.0229, 𝑝 = 0.0002 and the
average number of machines in the system is

L  0  0.4778  1  0.3823  2  0.1147  3  0.0229  4  0.0002  0.6896 máquinas

The average total cost per day is 20  8  0.6896  2  45  200.34 u.m./día
It turns out to be the most economical system.

SOLUTION OF PROBLEM 19
With one mechanic, you get a cost of 140 m.u., and with two, 152 m.u.

SOLUTION OF PROBLEM 20
Current cost per hour = 2456.62. Cost with aerosol = 1891.55. Cost with two servers
= 1389.43. It is preferable to put two servers.

SOLUTION OF PROBLEM 21
a) The optimal rate is 14.47 radios daily, with a daily cost of €378.43225.
b) In this case, a daily cost of 395,184 is obtained, which is worse.

SOLUTION OF PROBLEM 22
Applying the general model, since the distributions are exponential, but the service
rate varies according to the number of clients in the system, it is obtained:
𝑝 =0.264, 𝑝 =0.1983, 𝑝 =0.1653, 𝑝 =0.1653, 𝑝 =0.2066

SOLUTION OF PROBLEM 23
1. 𝜆 = 1/5 works/minute
2. Yes, because 𝜌 = 3/4 < 1
3. 𝑝 = 1/7 and therefore 𝐿 = 24/7 = 3.4286 works

November 2019 19
QUEUEING THEORY

4. Attention rate 𝑐𝜇 = 4/15 works per minute, then 3.4286 = 12.857 minutes
5. Cost/h = 20𝐿 = 68.572 €/h
6. Total cost per work = 10 𝑊 = = 57.1433 €
.
= .

SOLUTION OF PROBLEM 24
Cost/h = Amortization cost + Operation cost (1 − 𝑝 ) + Client cost 𝐿
Cost of robot 1 = 100 + 200 + 80 / = 666.66 €/h
/

Cost of robot 2 = 150 + 400 + 80 = 635.7143 €/h

/
/
The second robot is cheaper.

SOLUTION OF PROBLEM 27
a) It is a problem that mixes decision theory and queuing.
P(3º B) = 0.75; P(3º R) = 0.25; P(3º B/2º B) = 0.8; P(3º B/2º R) = 0.3
From these values it can be deduced P(3º R/2º B) = 0.2; P(3º R/2º R) = 0.7
P(3º B) = 0.75 = P(3º B/2º B) P(2º B) + P(3º B/2º R) P(2º R) =
= 0.8 P(2º B) + 0.3 (1 - P(2º B))
P(2º B) = 0.45/0.5 = 0.9; P(2º R) = 1 - 0.9 = 0.1
The cheapest option is to rent the machine in the first quarter.

b) The process of reaching the machines can have 𝜆 = 8 pieces/day for the regular
demand or 𝜆 = 12 pieces/day for the high demand. The service rates are 𝜇 = 16
pieces/day and 𝜇 = 10 pieces/day. The queuing system is M/M/1 for each machine.
For the regular demand we have 𝜌 = = 0.5 and for regular demand there is 𝜌 =
= 0.8. 𝑊 = ( )
= ( . ) = 0.125 day and 𝑊 = ( )
= ( . ) = 0.5 day.
So, in this system, the total time is 0.625 days.
In the case of good demand, it is necessary to install two B machines in parallel in
order to meet the demand. So, it is a system of M/M/1 queues for machine A and M/M/2
for machine B. Fort the high demand we have 𝜌 = = 0.75 and 𝜌 = ⋅ = 0.6. For
machine A 𝑊 = ( )
= ( . )
= 0.25 day. For machine B. 𝑝 = ( ⋅ . )
=
⋅ .
!( . )
= 0.15625 day.
( . ) . .
0.25, 𝐿 = ( . ) 0.25 + 1.2 = 1.875 𝑊 = =
So, in this system, the total time is 0.40625 days.

SOLUTION OF PROBLEM 28
a) Average time of an article in C1:
1art 60
C 1  15art/h C 1   art/h  20art/h
3min 3
Model M/M/1: 𝜌 = = 0.75 < 1
1 1
WC 1    1/ 5h  12min 𝐿 =𝜆 𝑊 = 15 ⋅ 1/5 = 3
(1  ) 20  0.25
Average time of an article in C2:
2 1art 60
C 2  15  10art/h C 1   art/h  6art/h
3 10min 10

November 2019
 QUEUEING THEORY

Model M/M/2: 𝜌 = = 5/6 < 1

1 1
𝑝 = = = 1/11
(5/3) 25/3 + 3/3 + 5/3
+ 1 + 5/3
2 ⋅ 1/6
( / ) /
𝐿 = ⋅ /
1/11 + 5/3 = = 5.46
LC 2 180 / 33
WC 2    18 / 33h  32.72min
C 2 10
Average time of an article in C3:
1 1art 60
C 3  15  5art/h C 1   art/h  8art/h
3 7.5min 7.5
Model M/G/1: 𝜌 = 5/8 < 1
𝜎 = 4/60
LC 3 511/ 432
WC 3    0.2366h  14.19min
( / ) ⋅ /
𝐿 = 5/8 + =
⋅ / C 3 5
Process A time: 12 min + 32.72 min = 44.72 min
Process B time: 12 min + 14.19 min = 26.19 min

The hourly cost of the system is

0.1€/min  60min/h(LC 1  LC 2  LC 3 )  6(3  180 / 33  511/ 432)  57.82 €/h
The cost attributable to the processed products according to A:
0.1€/min(WC 1  WC 2 )  0.1  44.72  4.47€/art
The cost attributable to the processed products according to B:
0.1€/min(WC 1  WC 3 )  0.1  26.19  2.619€/art

SOLUTION OF PROBLEM 29
At the beginning of June, the businessman is faced with three possible alternatives:

1. Hire a team of three mechanics for the three summer months,

2. Hire two teams of three mechanics for the entire summer or
3. Hiring a team in June and another in July.

The three decisions cost €45000, €90000 and €95000 respectively. Subsequently, in
July, chance plays its role by modifying upwards, or downwards, the daily rate of arrivals
of broken-down motorcycles, with respective probabilities of 0.75 and 0.25𝜆 = 24𝜆 = 6.
All of this is reflected in the following decision tree.

November 2019 21
QUEUEING THEORY

 julio  24 1

0.75

A1 0.25 2
a
r
pa ano
 julio  6
o
p e r
ui l v € 3
q
e oe 0 0  julio  6
1 d
to 4 50
- 1equipo para junio y 2 0.25
para el resto del verano
D1 A2
-9 - 95000 €
2 00 0.75
e
to qu 00
do ip €  julio  24
el o s
ve pa
r a ra  julio  24 4
no A3
0.75 5

106340 €
0.25
 julio  6

Different revenues are obtained depending on the decision made and the evolution of
the rates. To calculate them, we must consider that we find different models of queues,
all of them of the type M/M/s/4, where s takes the values 1 or 2, with  = 12, and  =
12, 24, or 6.
The key point is to determine the proportion of motorcycles that cannot be serviced,
which is given by the probability that the workshop is full, that is, the probability that the
queue is in state 4. According to the formulas corresponding to this model, we have to

a) s = 1

 
n
1
p0  4 n , pn    p0
     
  
 
n 0  

Therefore:
1. If  = 12, = 1 ⇒ 𝑝 = 0.2 ⇒ 𝑝 = 0.2 𝜆 = 9.6
2. If  = 24, =2⇒𝑝 = ⇒𝑝 = 𝜆 = 11.613
3. If  = 6, = 0.5 ⇒ 𝑝 = ⇒𝑝 = 𝜆 = 5.807

November 2019
 QUEUEING THEORY

b) s = 2


n
1
p0  , pn  2   p0 n  1
 3 
n
 2 
1    
 n 0  2 
Therefore:
4. If  = 12, =1⇒𝑝 = ⇒𝑝 = 𝜆 = 11.478
5. If  = 24, =2⇒𝑝 = ⇒𝑝 = 𝜆 = 18. 6̑
6. If  = 6, = 0.5 ⇒ 𝑝 = ⇒𝑝 = 𝜆 = 5.972

With this data, the expected income is as follows:

1. If you decide to hire a single team for the whole summer, they are:
a. During the first month of summer, an average of 360 motorcycles arrive,
of which 80% are serviced. The income during this month is 360 ⋅ 0.8 ⋅
65 = 18720€
b. During the other two months, if the rate of motorcycles rises to 24, only a
proportion of 15/31 of the motorcycles that arrive on average are served,
which means additional income for those 2 months of 60 ⋅ 24 = 14401440 ⋅
(15/31) ⋅ 65 = 45290€. This gives a total income for the entire campaign
of €64010.
c. If the average number of motorcycles drops to 6, a ratio of 30/31 is served,
which translates into an income during these two months 60 ⋅ 6 ⋅ ⋅
65 = 22645€ and a global income of € 41365

1. If you decide to hire two teams for the whole summer, they are:
a. During the first month of summer, an average of 360 motorcycles arrive,
of which the 22/23 ratio is served. The income during this month is 360 ⋅
(22/23) ⋅ 65 = 22383€
b. During the other two months, if the rate of motorcycles rises to 24, only a
proportion of 7/9 of the motorcycles that arrive on average are served,
which means additional income for those 2 months of 60 ⋅ 24 = 14401440 ⋅
(7/9) ⋅ 65 = 72800€. This gives a total income for the entire campaign of
€ 95183.
c. If the average number of motorcycles drops to 6, practically all 212/213
are served, translating into an income during these two months 60 ⋅ 6 ⋅
65 = 23400€ and a global income of 45783 €.

1. if you decide to hire a team at the beginning and another one in July
a. During the first month of summer, an average of 360 motorcycles arrive,
of which 80% are serviced. The income during this month is 360 ⋅ 0.8 ⋅
65 = 18720€.+

November 2019 23
QUEUEING THEORY

b. During the other two months, if the rate of motorcycles rises to 24, only a
proportion of 7/9 of the motorcycles that arrive on average are served,
which means additional income for those 2 months of 60 ⋅ 24 = 14401440 ⋅
(7/9) ⋅ 65 = 72800€. This gives a total income for the entire campaign of
€91520.
c. If the average number of motorcycles drops to 6, practically all of them are
serviced, translating into an income during these two months 60 ⋅ 6 ⋅ 65 =
23400€ and a global income of € 42120.

With these data, the valuation of the different leaves of the tree is reflected in the
following figure

64000 €
 julio  24 1

0.75
0.25 41394 €
A1 2 42120 €
ra  julio  6
pa ano
o
p e r
ui l v € 3
eq o e 0 0  julio  6
1 d 0
to 45 1equipo para junio y 2
- 0.25
para el resto del verano
D1 A2
- 95000 €
2 -9 0.75
to equ 00
d o ip 00  julio  24
e o
lv p s €
er ar
an a
 julio  24 4
o A3
0.75 5 96100 € 91520 €

0.25
 julio  6

6 46800 €

We move on to assess the vertices of the tree. Those of chance with the weighted
average value of the values associated with the extreme vertices of arcs that come from
them are already valued. The decision-makers choose the alternative with the highest
expected value. This leads us to the optimal decision to hire a single team for the entire
campaign with an expected profit of 13349 €.

November 2019
 QUEUEING THEORY

64000 €
 julio  24 1

58348 € 0.75
0.25 41394 €
A1 2 42120 €
ra  julio  6
pa ano
o
p e r
ui l v € 3
q
e oe 0 0  julio  6
1 d 0
to - 45 1equipo para junio y 2
0.25
para el resto del verano
D1 A2
- 95000 €
2 - 79170 € 0.75
to equ 90
do ip 00  julio  24
e l os 0
€
ve pa
ra ra  julio  24 4
no A3
0.75 5 96100 € 91520 €
83775 €
0.25
 julio  6

6 46800 €
The last section asks to calculate the VEIP, the difference between the expected profit
if the future were known and the expected gain under risk.
If the arrival rate for the last two months is exceeded and the optimal decision is
made, the profit would be the maximum between 19010, –3480, and 6100, i.e., 19010 €.
If it were known that the rate is, then, the optimal decision is the maximum between 𝜆 =
24𝜆 = 6–3635, –52880, –43200, i.e., –3635 €. The expected profit with perfect information
is €. The return at risk is €13349. Then, the VEIP is the difference between the two,
which, in this case, is 0. This result is produced by the coincidence in the optimal decision,
the first, both when the rate is 𝜆 = 24, 𝜆 = 6, and when it is as low risk 19010 ⋅ 0.75 −
3635 ⋅ 0.25 = 13349.

SOLUTION OF PROBLEM 30
a) The state variable is the vehicles in the main workshop and in the annex.
b)
0 1 2
0 1 2 3
1 2 3

210  
p3  p0 , p2  1 0 p0 , p1  0 p0 , 𝑝 + 𝑝 + 𝑝 + 𝑝 = 1
321 21 1
2  1  0   , 𝜆 = 0 𝑛 ≥ 3, 𝜇 = 𝜐 + 𝜐 , 𝜇 = 𝜇 = 𝜐 ,
c) 𝑝 = 𝑝 = 𝑝 = 2/7 ≃ 0.286, 𝑝 = 1/7 ≃ 0.143

November 2019 25
QUEUEING THEORY

d) The probability of not being able to be repaired in the main workshop or in the annex
is 𝑝 ≃ 0.143.
e) 𝑑 = 𝜆(1 − 𝑝 ) = 6(6/7) ≃ 5.143 pieces/hour, 𝑐 = 500 €, 𝑐 = 4/(9 ⋅ 250) €/piece
and hour, 𝑙 = 16 days.
9 ⋅ 250 ≃ 1700.86 pieces. 𝑇 = 1700.86/5.143 = 330.71ℎ/9 ≃
⋅ . ⋅
𝑄∗ =
36.7 days and reorder point 𝑙𝑑 = 16 ⋅ 5.143 ⋅ 9 = 740.592 pieces.

SOLUTION OF PROBLEM 31
How many administrative staff will be required to be in office B if the probability of
a wait of more than 10 minutes is less than 0.05?
Given systems A and B linked in series, if the first system A receives users according
to a Poisson law with an arrival rate 𝜆 and that system is stable, system B receives users
from system A receives them according to a Poisson process of arrival rate 𝜆 . The
following figure represents the two linked systems with their input and output rates,
assuming that both systems, A and B, are stable.
A B
A A A
A B

The value of the entry fee to system A is 12 people/hour. As the service rate of this
system is 15 people/hour (60/4), system A turns out to be stable. System B with a single
administrative staff would not be stable since the service rate is 8 people/hour compared
to an arrival rate of 12 people/hour. Therefore, with two administrative staff, system B
would be stable.
The calculation of the probability that there is a waiting time of more than 10 minutes
for two administrative staff in system B is calculated as:
𝜆 12
= = = 0.75
𝑆𝜇 2 ⋅ 8

1 1
𝑝 = = = 0.142857
(𝑆 ) (𝑆 ) (2 ⋅ 0.75) 2 ⋅ 0.75
+∑ 𝑛!
+1+ 1
𝑆! (1 − ) 2! (0.25)

(𝑆 ) (2 ⋅ 0.75)
𝑃(𝑡 > 0) = 𝑝 = 0.142857 = 0.64285
𝑆! (1 − ) 2! ⋅ 0.25

 10 
P tc    e S x 1P tc  0  e 28 60.250.64285  0.33
1

 60 
As the probability is greater than 0.05, the number of administrative staff must be
increased to assess the probability that the waiting time will exceed 10 minutes with three
administrative staff.

November 2019
 QUEUEING THEORY

𝜆 12
= = = 0.5
𝑆𝜇 3 ⋅ 8

1 1
𝑝 = = = 0.21
(𝑆 ) (𝑆 ) (3 ⋅ 0.75) 3 ⋅ 0.5 (3 ⋅ 0.5)
+∑ 𝑛!
+1+ 1 + 2!
𝑆! (1 − ) 3! (0.5)

(𝑆 ) (3 ⋅ 0.5)
𝑃(𝑡 > 0) = 𝑝 = 0.21 = 0.23625
𝑆! (1 − ) 3! ⋅ 0.5

 10 
P tc    e S x 1P tc  0  e 38 60.50.23625  0.03197
1

 60 
With three administrative staff, it is obtained that only 3% of the people who access
system B will have a waiting time of more than 10 minutes.

November 2019 27