Queueing Models
Uploaded by
ScribdTranslationsQueueing Models
Uploaded by
ScribdTranslationsWAITING LINE MODELS
1- A queueing problem is one in which there is a succession of
units, (people or jobs) that arrive at a establishment to receive a
service.
Figure # 1
Queueing systems
Single channel model.
Station of
service
Multiple servers
(various channels)
Server
1
Server
2
Server
3
The previous figures show the general problem of waiting lines, known
as a single channel model. This consists of a queue and a single server.
As an example, we have the case of patients waiting to be seen in a clinic.
doctor. The multiple server model also consists of a queue, but to
the difference from the single channel model, there are several servers attending, in such a way
that the first task in the queue will be attended to by the first server that becomes free. It is
the case of the waiting lines at a bank where there are several tellers.
The case of the lines in a supermarket involves several systems of a
single channel.
On the other hand, when a task must go through several service stations in
sequentially, it is a modification of the known single channel model
as a multiphase system, and queues can be generated between stations, or servers
idle, as long as there is an imbalance in the attention capacity of each one of them
the stations. The system as a whole will have a service capacity equal to the
the capacity of the slowest station or in other words the one that requires the most time in
the demanded service. An example is when at a fast food restaurant you
They take the order at a window, ask him to pay at another, and to pick up the food.
at a third window. This type of model is shown in the following figure:
S S S S
#1 2 #3 #4
2- ANALYSIS OF WAITING LINES
The questions that can be asked in relation to a queue system
they revolve around four quantities and are the basis of queueing analysis:
a- The number of tasks in the system: Those that are receiving the service
currently, as well as those who are waiting.
b- The number of tasks in the queue: Those that are waiting to be addressed.
c- The waiting time: The period that elapses between the moment a task enters
to the system and the moment when the service is completed. Note that the period
include the duration of the service.
d- The waiting time in line: Time that elapses from the moment of entering the
system and the start of the service itself.
To answer the questions regarding these amounts, it is necessary to make
some basic assumptions of particular aspects of the system:
A- Arrival process: Each arrival is called a 'task'. Since it is not known the
time interval between two arrivals, we need to specify a distribution for it
probability. In the basic model, a particular distribution called
exponential distribution, sometimes called exponential distribution
negative." This distribution explains the reason why there are still waiting lines.
when the values of lambda and mu, which will be seen later, would suggest
that there should be no waiting lines.
The exponential distribution is specified by a parameter called
lambda (λ) and what is the average arrival rate; that is, how many tasks are presented,
on average during a given period.
Example:
If on average a task arrives every 20 minutes, the average time between arrivals
it's 20 minutes, therefore λ = 1 / 20 = 0.05, which means that they arrive in
average five hundredths of homework per minute.
B- Service process: In the basic model, the time it takes to complete the
a task (service time) is also managed through distribution
exponential. The parameter of this exponential distribution is called mu (μ) which
represents the average service rate in tasks per unit of time.
Example:
If a task is completed in 8 minutes on average, then μ=1/8=
0.125, which means that an average of 12.5 hundredths of the task is completed.
per minute.
C- Queue size: There is no limit to the number of tasks that
They can wait in line. It is said that the waiting line is infinite.
D- Queue discipline: Tasks are attended to based on the
First come, first served.
E- Time horizon: The system operates continuously in infinity.
3- OPERATIONAL CHARACTERISTICS OF THE BASIC MODEL
The basic model requires that λ < μ, as this would occur if it were given.
on the contrary, (λ > μ) the row would grow continuously, the system would collapse and therefore not
there is analysis to be done, (formulas to apply in it) since what is required is
a solution to the system incapacity problem.
The question that arises is: why if the arrival rate is lower than the rate of
service forms the line? The reason is explained by the fact that tasks do not arrive in
uniform way over time, but there will be periods when a certain amount arrives
of tasks at the same time forming the queue; and there will be other periods when they arrive
fewer tasks, or they do not arrive at all, which allows the server to keep "shortening" the queue. Thus
the line forms because the service time is not constant, there will be tasks
that require more time, and when attended to, a line forms. However, on average
the service time is less than the time between arrivals and therefore much greater than
lambda. If mu were not greater than lambda, there would be no possibility of shortening the line and the
the system would collapse as mentioned earlier. As the average capacity
if the server rate is greater than the arrival rate, the server will be able to meet all the
tasks; the system is capable.
Given this situation, queue analysis seeks to answer the questions that
were formulated in point # 2 to know: average of tasks waiting in the system and in
the line and average waiting time in the system and in the line. These responses are
they are obtained through the formulas that will be presented for the different models
later. For the moment we will focus on the basic single-channel model:
MODEL 1
It is characterized by being a single-phase single-channel. The queue is of unlimited size and
It is organized with a FIFO discipline. Customers coming from a population
infinite arrivals occur with times that are distributed according to a Poisson distribution. The
Service time is distributed exponentially.
Characteristic Symbol Formula
Expected number of tasks in the system L λ / (μ-λ)
Number of expected tasks in the queue Lq λ 2/ μ (μ-λ)
Expected waiting time in the system, (includes W 1 /(μ – λ)
the service time
Expected wait time in the queue Wq λ / μ ( μ – λ)
Probability that the server is busy,
(% server occupancy time) - λ/μ
Probability that the system is empty, By 1- (λ / μ)
( idle server )
Probability of there being more than k tasks in the Pn>k (λ / μ)k+1
system.
Probability that there are n tasks in the Pn (1-(λ / μ))(λ / μ)n
system.
WHAT HAPPENS WHEN λ > μ?
As explained earlier, if λ > μ the queue grows to infinity, so
it is necessary to increase the number of servers to 'decrease' the service time of the
system as a whole, and a system with several servers or system will have been established
multichannel, where different formulas are applied to answer analysis questions
previously indicated. This type of models will be studied later. Next, it
They will review some examples of type 1.
Solved Example No. 1:
Solution for the case presented whose data was exposed earlier, where
we define: λ = 0.05 tasks per minute and μ = 0.125 tasks per minute.
Expected tasks in the system: 0.05 / (0.125–0.05) = 0.67 tasks.
2-Expected tasks in the queue: 0.0520.27 tasks.
Expected time in the system: 1 / (0.125–0.05) = 13.33 minutes.
Expected time in line: 0.05 / [0.125 (0.125–0.05)] = 5.33 minutes.
Note that since λ and μ are expressed in terms of minutes, the waiting times
calculated are in the same units. On the other hand, it can be noted that the difference
between the waiting time in line and the waiting time in the system, (13.33–5.33),
it is equal to the service time itself, that is 8 minutes, (1/μ = 1 / 0.125 = 8).
Server occupancy time: 0.05 / 0.125 = 0.4 = 40% of the time.
Probability of occupancy or system utilization factor
6- Probability that the system is empty: 1 - (0.05/0.125) = 0.60 = 60%
7- As noted in point one, in this system 0.67 tasks would be expected.
system. As this is an average we might want to know what the probability of
that there is more than 1 task in the system.
Pn>1 = (0.05 / 0.125)1+10.16 = 16%
PRACTICE
Exercise #1
You manage a cinema complex in a shopping center. The company has four
rooms and in each of them a different movie is shown. The programming is done
so that the start times of the movies are staggered to avoid the long ones
queues that would be produced if all were started at the same time. The complex has a
the only ticket sales window that can maintain an average service of 235
spectators per hour. The arrival of spectators on a normal activity day has
an average of 210 per hour. The management of the Shopping Center has asked you to
I took the necessary measures so that the rows in the system do not exceed 10.
people as this obstructs the way to other businesses.
a- Determine the average number of spectators waiting in line to buy
your tickets.
b- What percentage of the time is the ticket agent occupied?
How long does a customer typically stay in the system?
d- How long does an average viewer stay in line?
What is the probability that there are more than the allowed number of clients in the system?
How do you evaluate the situation regarding this?
Exercise No. 2
In a large company that manufactures intimate apparel, it has been determined that every 5
Minutes report machines with minor faults, according to a distribution
Poisson. There is only one mechanic who takes an average of 4 minutes to solve the
problem, with service times that follow an exponential distribution.
a- Determine the average number of machines waiting to be repaired.
What percentage of the time is the mechanic occupied?
What is the average time a machine remains out of service?
d- What is the average time a machine remains out of service before it
Did you start your repair?
What is the probability that there are more than four machines out of service?
What is your opinion on this situation? What would you recommend?
Exercise #3
Two mechanics are being considered to service the machines in a plant.
industrial. The first mechanic, category A, would be paid $10 per hour and can
repair machines at the rate of 8 per hour. The second mechanic, category B, is assigned to
They would pay $7 per hour, but he would repair the machines at a rate of 7 per hour. It is estimated
that the time a machine is down has a cost for the company of $15 per
hour.
Assume that machines break down according to a Poisson distribution with an average
4 per hour, and the repair time is exponential. For each mechanic
determine
a. The number of machines that would be expected to be idle in the system?
b. The time that one would expect a machine to be idle in the system?
c. What percentage of the time would the system be empty?
d. What is the probability that more than 2 machines are down?
Which of the two mechanics would you recommend?
COST ANALYSIS IN QUEUES
The costs in a queueing system are given by the cost of providing the
service, represented by the cost of having a server multiplied by the number of
servers. The second cost to identify in a waiting line system is the cost
what it means to have tasks waiting in the system or waiting in the queue. For this
Overall, the cost of waiting in the system will be. For example, in an industry, the cost of
having machines waiting to be repaired incurs the cost of idle labor, delays in
production, others. Another example of costs, although more difficult to determine, is the cost
of the people waiting in a bank: cost of image loss and consequently
loss of customers.
If the number of servers in a system is increased, we could reduce costs of
wait; but it would be increasing the costs of providing the service and vice versa. The
the ideal number of servers is the one that represents a balance between both costs; but
difference from other models, such as inventory models, in this case there is no formula
that allows us to determine the ideal number of servers, but this is done by
test of trial and error by varying the number of servers and calculating L or Lq, (according to
the case), and calculating the total costs of the system using the following formula:
CT = L * Ce + S * Cs ó
CT = Lq * Cq + S * Cs
Where:
CT = Total costs of the system.
L = Number of tasks waiting in the system.
S = Number of servers.
Ce = Cost of waiting in the system per unit of time.
Cq = Cost of waiting in line per unit of time.
Cs = Cost of having a server per unit of time.
Note: When applying these formulas, special care must be taken that the costs are
expressed for the same units of time in which λ and μ have been defined.
Resolved example
The truck drivers of a company earn $10 per hour. The loaders that
they work on the loading dock, with space for only one truck, earning $6 per
time and carry out the loading with a semi-automatic machine. The drivers of the
trucks that are waiting in line or at the loading dock, they have a salary but are
productively inactive during that waiting time in the system. On the platform of
charging stations can be located one or more chargers increasing the charging capacity as
It is indicated in the following table. On average, 3 trucks arrive at the plant per hour and
One charger can charge 2 trucks per hour. What is the number of chargers?
that minimizes the system costs?
a.λ = 3 y μ = 2
b. As λ > μ we need at least 2 servers in which case μ = 4.
let's remember the assumption that increasing the number of servers does not modify the
single-channel queue system, in this case since the platform only has
space for a truck.
c. Since the unproductivity of drivers occurs whether they are waiting in line.
or while the truck is being loaded, the waiting costs are considered for the
system, therefore we must calculate L and not Lq.
Amount of Tasks Cost of waiting Cost of Total cost
chargers μ waiting $ service (CT)
L $ $
1 2 - - - -
λ/(μ-λ)= L * Ce S * Cs 30 + 12 =
2 4 3/(4-3)= 3 3 * 10 = 30 2 * 6 = 12 42
λ/(μ-λ)= L * It S * Cs 10 + 18 =
3 5 3/(5-3)= 1.5 1.5 * 10 = 15 3 * 6 = 18 33
λ/(μ-λ)= L * Ce S * Cs 6 + 24 =
4 6 3/(6-3)= 1 1 * 10 = 10 4 * 6 = 24 34
Conclusion:
a- You cannot use just one charger because the line grows indefinitely, the
system collapses.
b- With two chargers, the total cost of the system is $42 and decreases to $33 with three.
chargers.
c- By using four chargers, the total cost of the system rises to $34.
consequence, to minimize system costs, three should be employed.
chargers.
Note:
It should be clarified that the payment the company makes to the drivers of the
trucks it will not vary, regardless of whether 2, 3, or 4 are used
chargers, this type of analysis allows us to make a decision by assessing the cost
of opportunity that represents having tasks pending, in this case the cost of
opportunity to have the truck drivers on standby.
PRACTICE
Exercise #1
American Vending Inc. supplies food dispensers at a university
of large size. As students tend to kick the machines in anger and
frustrated, management has to face a constant problem of repairs. In
on average, three machines break down per hour and the breakdowns are distributed in a way
Downtime costs the company $25 per hour per machine and to each
maintenance employee is paid $4 per hour. An employee can repair the
machines at an average rate of five per hour, distributed exponentially; two
employees who work together can repair seven machines per hour, distributed
exponentially; and a group of three workers can repair eight per hour,
exponentially distributed.
What is the optimal size of the maintenance team to repair the machines?
Exercise #2
For the case of exercise No. 3 of the first practice, consider the salaries of the
mechanics A and B as well as the cost of having a machine stopped: what
Which mechanic do you recommend? Compare your decision with the one made earlier for
this case.
Exercise #3
The rice harvesting season in the central Pacific (Zone of Quepos and Parrita),
it's short and the farmers take their trucks loaded with rice to a giant silo of
central storage over a two-week period. Therefore, the trucks that
full of rice wait to unload and return to the rice fields, they form lines of
more than one 'block' in the reception of the silo. The silo is cooperatively owned by the
farmers and make the unloading-storage process as
The possible efficiency will benefit all farmers. The cost of deterioration
due to delays in unloading and the cost of renting the truck and the
driver downtime is important for the members of the
cooperative. Although farmers have difficulties in quantifying the damages
in the harvest, it is easy to assign a waiting and unloading cost for the trucks and
drivers who are evaluated at $10 per hour of waiting both in line and during the
download process proper. The storage silo is open and
working 16 hours a day seven days a week for the two weeks of
collection and is capable of unloading 35 trucks per hour, following a distribution
exponential. The trucks arrive filled throughout the day (during the hours that the
the store is open) at an average of one every two minutes, following a law of
Fish.
Determine:
1- The total daily cost for farmers to have their trucks parked in
the entire downloading process, (queuing and the download itself).
2- The cooperative, as mentioned, uses the storage silo intensively.
only for two weeks a year. Farmers estimate that remodeling the silo
it has a cost of $9000 and would increase the capacity by 15 trucks/hour
download. Will it be worthwhile for farmers to carry out this renovation? Justify.
your answer with the respective calculations.
Note: The operating cost of the silo will not change with the remodeling.
MODEL 2
It is characterized by being a single channel of single phase. The queue is of unlimited size.
is organized with a PEPS discipline. The clients that come from a population
infinitely arrive with times that are distributed according to a Poisson distribution. The
service time is constant.
Characteristic Symbol Formula
Expected number of tasks in the system L Lq + (λ / μ)
Expected number of tasks in the queue Lq λ 2/ 2μ (μ-λ)
Expected wait time in the system, W Wq +(1 / μ)
(includes service time)
Estimated wait time in line Wq λ / 2μ (μ - λ)
Probability that the server is
busy, (% of occupancy time of P λ/μ
server
Probability that the system is empty, Po 1-(λ / μ)
(unoccupied server)
Probability of having more than k tasks Pn>k (λ / μ) k+1
in the system.
Probability of there being n tasks in the Pn (1-(λ / μ))(λ / μ)n
system.
Solved example
A university has decided to place a microwave for the students
heat up your lunch. The oven has been programmed for exactly 105 later.
seconds to heat each lunch. During lunch periods, students
They arrive at an average rate of one every 2.25 minutes following a Poisson law.
a. Determine the average number of students in line waiting to heat their
lunch?
b. The average total time that a student is expected to take in the process of
heat your lunch? (time in the system).
c. The total number of students waiting in the lunch warming system?
d. What is the probability that there are more than 3 students in the system?
e. What is the probability that a student finds the oven available?
λ = 1 / 2.25 = 0.444 students per minute.
μ = 1 / 1.75 = 0.571 lunches (students) per minute.
a. Lq = (0.444)2(2x 0.571(0.571 - 0.444)) = 1.36 students
b. Wq = (0.444) / (2 x 0.571(0.571 - 0.444)) = 3.06 minutes
W = 3.06 + 1/0.571 = 4.81 minutes
c. L = 1.36 + 0.444 / 0.571 = 2.14 students
d. (0.444 / 0.571)3+136.5%
e. Po = 1–0.444 / 0.571) = 22.2%
PRACTICE
The Departmental Bank has decided to install an automated service kiosk.
drivers for people who wish to make a single deposit; the manufacturer has
informed the Bank that in these cases the service time is constant at 7.5
minutes. To determine the operating characteristics of this new system, they have
evaluated the arrivals of the cars and it has been found that they behave in a way
from a Poisson distribution with an arrival of 4 cars per hour.
a. The average idle time of the system.
b. The average number of people in the system.
c. The average number of customers in line.
d. The average time a person stays in the system.
e. The average time of a customer in line.
MODEL 3
It is characterized by being a single-phase multiple channel. The tail is of size
unlimited and is organized with a FIFO discipline. It assumes an infinite population with
Poisson arrivals and exponentially distributed service time.
Characteristic Symbol Formula
Expected number of tasks in the system L Lq + (λ / μ)
Number of expected tasks in the queue Lq λ μ (λ/μ)sPo
(s–1)! (sμ-λ) 2
Expected waiting time in the system, W Wq +(1 / μ)
(includes service time)
Expected waiting time in line Lq/ λ
Probability that the server is busy,
(% server occupancy time) P λ/μs
Probability that the system is empty, _____________1________________
(idle server)
s-1
Po
∑ 1/i! (λ/μ)i+ 1/s! (λ/μ)s sμ/(sμ-λ)
i=0
Probability that all servers Pw _1λ_Sμ_ S Po
busy
S!μ Sμ-λ
Probability of having n tasks in the system Pn ((λ/μ)n/n!) P0 For n <= S
n n-s
((λ/μ) /(S! S )) P0 For n > S
Solved example:
A certain bank receives 100 customers per hour during the mornings.
Poisson distribution. There are three 'open' boxes and the average time to attend to a
The customer arrival time is 1.5 minutes, distributed exponentially. Perform the evaluation of this.
system calculating the different performance measures:
a.
1
Po = (1/0! (100/40)0 + 1/1! (100/40)1+1/2! (100/40)2) + 1/3! (100/40)3-1.0
1
(1 + 2.5 + 0.5 * 6.25 ) + ((1/6 * 2.53* (3 * 40 / 3*40–100)
1
(1 + 2.5 + 0.5 * 6.25 ) + 0.1667 * 15.625 * 120/20
1 1
(6.625 + 15.625) = 22.25 = 0.045 = 4.5%
Once the Po value is established, it is substituted in the rest of the formulas along with the
values of λ and μ.
b. Lq = 100 * 40 (100 / 40)30.045
(3-1)! (3* 40–100)2
= 4000 * 15.625 * 0.045= 2812.50 = 3.52 clientes
2 * (20)2 800
c. L = 3.52 + 100 /40 = 6.02 clients
d.. Wq = 3.52 / 100 = 0.0352 hrs 2.1 minutes
e.. W = 0.0352 + 1 / 40 = 0.062 hrs (3.61 minutes)
The occupancy of the system is 83.3%, or
percentage of the time that will be occupied
employee.
g.. Pw = 1/3! (100/40)3* 3 * 40 * 0.045 = 0.703
3*40 - 100
The probability that all employees are busy, (the probability that when
a customer should wait to be served), is 70.3%.
To calculate the probability of having 2 customers in line, P5 should be calculated.
(probability of 5 customers in the system), since three customers would be attended to
and two in a row. R/ 8.1%
The probability that 3 or fewer people are in line would be the sum of the
probabilidades de que haya 0,1,2,3,4,5 y 6 personas en el sistema. R/ 66.1%
Use of Table:
(1)
To facilitate calculations, the value of Lq can be obtained from Table I. Using
this value, and the previous formulas, can be used to calculate certain performance measures.
The remaining ones would be obtained by isolating the value Po in the Lq formula.
2
Po = Lq(s–1)! (sμ-λ)
λ μ (λ/μ)s
PRACTICE
Exercise No. 1
Customers arrive at the American Airlines ticket counter. at a rate of 50 per hour,
following a Poisson distribution. An agent takes an average of 6 minutes to assist
a customer, (exponentially distributed). American Airlines has 7 agents in
box office. Calculate the different performance measures for this case.
Exercise No. 2
For the previous case, suppose that the agents earn $20 per hour. On the other hand
American Airlines has estimated that the commercial cost, (of image), because a customer
waiting for attention, (in line), it is $30 per hour. Determine what would be the
What is the optimal number of agents that this airline should have attending to its customers?
(1) Taken from: Chase. Aquilano J. Production and Operations Management. Eighth Edition.
McGraw-Hill Publishing 2000.
MODEL No. 4
Assume a single server, single phase, and that there is a small number of possible clients.
(finite population). If there is a small number of possible clients, the number of clients
present in the system will affect the rate at which new clients arrive; this is because
When a customer enters the system, the size of the source population decreases and therefore
the probability of a new customer arriving also decreases. When it is completed
the task, the client joins the source population and again increases the
probability of occurrence of a new task. This situation forces the use of formulas
different from those used in infinite population.
Finite row notation
N = Number of units in source population or size of the population.
H = Average number of units being attended to.
M= Number of service channels.
T = Average time to provide the service.
U = average time between services.
X = Service factor, or proportion of service time required.
F = Efficiency factor, a measure of the effect of having to wait in line.
CHARACTERISTIC SIMBO FORMULA
LO
Expected number of tasks in the system L N-(λ / μ) (1-Po)
Number of tasks expected in the queue Lq N- λ + μ* (1- Po)
λ
Expected wait time in the system, W Wq+ ( λ / μ)
(includes the time of service)
Expected wait time in line Wq Lq / (μ (1 – Po))
Use of the system P λ / μ
1
Probability that the system is empty, By N
n
(idle server) ∑((N / N-n)!)( λ / μ)
n=0
Probability of having more than k tasks Pn ( λ / μ)n*((N / N-n)!) P 0
in the system.
This type of model is solved more easily using finite tables and formulas.
specifics:
X = T / (T+U)
2. H = F*N*X
3. F = (T+U) / (T+U+W)
4. Lq = N* (1–F)
5. L = Lq + H
6. W = LT / LH(T+U)
or / (N-L)
7. Wq = W - (1/μ)
Resolved example:
In a certain workplace, there are 4 machines, each of which must be
adjusted on average every 2 hours. A single mechanic performs the adjustments
corresponding, a task that takes 10 minutes. Calculate the different measures of
performance in this case and also calculate the probability of there being two
machines in the system.
λ = 1 / 2 = 0.5 machines per hour N = 4 machines.
μ = 60 /10 = 6 machines per hour. M = 1 server.
Po = 1
4!/(4-0)!(0.5) + 4!/(4-1)!(0.5) 4!/(4-2)! (0.5)24!/(4-3)!(0.5/6) 3+ 4!/(4-4)!(0.5/6) 4
0 1
1
=
1 + 24/6*0.83 + 24/2*0.0069 + 24*0.00579 + 24*0.000482
1
=
1 + 0.332 + 0.0828 + 0.01389 + 0.011
= 1 / 1.44 = 0.6987
a. Lq = 4 -0.5 + 6 * (1- 0.698) = 4 - 3.9168 = 0.0831 machines
0.5
b. L = 4 –12 * (1–0.6987) = 0.3844 machines
c. Wq = 0.0831 0.046 hours
6 * (1 - 0.6987)
d. W = 0.046 + 1/6 = 0.2126 hours
e. P20.5/624! / (4-2)! * 0.6987
0.0069 * 24 / 2 * 0.6987 = 5.8%
Solution using finite row tables
X = 10 / (10 + 120) = 0.077
Lq = 4 ( 1- 0.98) = 0.08 machines
H = 0.98 * 4 * 0.077 = 0.30
L = 0.08 + 0.30 = 0.38 machines
W = 0.38 * 0.1666 / 0.30 = 0.21 hours
Wq = 0.21 - (1/6) = 0.043 hours
MODEL No. 5
This model is similar to model 4, that is to say, it assumes a single phase and a finite population, but
of multiple channels.
The formulas corresponding to the different performance measures are:
1. P = λ/ sμ
N
2.Lq = ∑ (n –s) Pn
n=s
s-1 N s-1
3.L = ∑ n * Pn + ∑ (n-s) Pn + s ( 1 - ∑Pn )
n=0 n=s n=0
4.Pn = (λ / μ)n* N1/ ((N-n)!n!) * Po
For 0 <= n <= s
n-s
5. Pn = (λ/μ)* N!/ ((N-n)!s!s ) * Po
For s <= n <= N
1
6. Po =
s-1 N! (λ/μ)n N N! (λ/μ)n
∑ (N-n)! n! + ∑ (N-n)! s! sn-1
n=0 n=s
Solved example:
Consider a service center that serves a population of four machines that
must be adjusted every 10 hours. There are two mechanics, each of whom
can repair a machine in an average of 2 hours.
1
Po =
24/(24-1) (1) + 24/(6*1) * (0.20) + 24/(2*2*1) * 0.04 + 24/(1*2*2) * 0.008 + 24/(1*2*4) * 0.0016
1
= 0.48
1 + 0.80 + 0.24 + 0.048 + 0.0048
P1 = (0.10 / 0.5)124 / (3! * 1!) * 0.48 = 0.384
P2= (0.10/0.5)2* 24 / ( 2!*2!)) * 0.48 = 0.1152
P3 = (0.10 / 0.5)324 / 1*2!* 2 1* 0.48 = 0.023
P4= (0.10 / 0.5)4* 24 (0! * 2 * 22* 0.48 = 0.002
L = 1 * 0.384 + 0.023 + 2 * 0.002 + 2 ( 1 - (0.48 + 0.384))
= 0.384 + 0.023 + 0.004 + 0.28 = 0.69 machines.
Lq = 1 * 0.023 + 2 * 0.002 = 0.027 machines
Use of finite population table:
X = 2 / (10 + 2) = 0.166
Lq = 4 * (1 - 0.993) = 0.028 machines
H = 0.993 * 4 * 0.166 = 0.666
L = 0.028 + 0.666 = 0.688 machines.
Practice:
1. For the previous case:
a. Calculate the other performance measures.
A textile factory that has four machines needs to determine if it is economical
hire one more mechanic. Currently, he only has one capable mechanic.
make the adjustments in 7.5 minutes per machine, (distributed
exponentially). The machines need to be adjusted on average every
hour. A machine out of service represents a cost for the textile company of $40
per hour. Mechanics are paid $7 per hour.
For one and two mechanics determine:
a. The number of machines waiting in line.
b. The number of machines out of service.
c. The expected time a machine will be out of service?
d. From an economic point of view: how many mechanics should be hired?
Explain.
MODEL NO. 6.
It is characterized by being a simple channel in series, in tandem, or of multiple phases. That is to say
customers must go through different service stages in succession. The
performance measures are as follows:
1. Pi = λ / μi
F
2.W = W1+ W2+ W3+ …WF=∑((1 / (1-Pi)) (1/μ i))
i=1
F
3. Lf = L1+ L2+ L3+LF= ∑ (Pi / (1-Pi))
i=1
F
4. Wq = Wq1 + Wq2 + Wq3 + Wqf = ∑((Pi/(1-Pi)) (1/μ i))
I=1
Example:
A clinic offers a packaged service that the client must follow the procedure.
what is indicated below. On average, it is estimated that 10 customers arrive per hour at
request the services of the clinic.
Activity or service Time in minutes
Register 3
X-ray taking 2
Take blood pressure 1
Take a blood sample 2
Take a urine sample 1
Physical examination by the doctor 5
Determine: