Upload
sean-wright
View
222
Download
0
Embed Size (px)
Citation preview
8/7/2019 D S Lecture #3 - Queing Theory
1/33
1
USE OF PROBABILISTIC
MODELS TO SOLVE REALWORLD PROBLEMS:
QUEUING THEORY
DECISION SCIENCE
LECTURE # 3
8/7/2019 D S Lecture #3 - Queing Theory
2/33
2
WHAT IS QUEUING THEORY ? Instances of waiting in a line:
Models developed to better understand& make decisions on the problem ofwaiting in lines.
A waiting line is called a queue
The body of knowledge dealing withwaiting lines is called QUEUINGTHEORY
8/7/2019 D S Lecture #3 - Queing Theory
3/33
3
Waiting Line/Queuing Theory Modelsconsist of Mathematical Formulas and
relationships that can be used todetermine the OPERATINGCHARACTERISTICS (performancemeasures) for a waiting line.
There are many such operatingcharacteristics only some will bestudied
WHAT IS QUEUING THEORY ?
8/7/2019 D S Lecture #3 - Queing Theory
4/33
4
OPERATING
CHARACTERISTICS(PERFORMANCE MEASURES)
1. The Probability that no units are in the system.
2. The Average Number of units in the waiting line3. The Average Number of units in the system (the
number of units in the waiting line plus the number ofunits being served)
4. The Average Time a unit spends in the waiting line5. The Average Time a unit spends in the system (the
waiting time plus the service time)
6. The Probability that an arriving unit has to wait forservice.
8/7/2019 D S Lecture #3 - Queing Theory
5/33
8/7/2019 D S Lecture #3 - Queing Theory
6/33
6
Single-Channel Waiting Line
STRUCTURE OF A WAITING
LINE SYSTEM
8/7/2019 D S Lecture #3 - Queing Theory
7/33
7
STRUCTURE OF A WAITING LINE
SYSTEM SINGLE CHANNEL
a server takes a customers order
Determines the total cost of the orderTakes the money from the customer
Fills the order
The server then takes the order of the next customerwaiting for service
8/7/2019 D S Lecture #3 - Queing Theory
8/33
8
STRUCTURE OF A WAITING LINE
SYSTEM MULTIPLE CHANNEL
8/7/2019 D S Lecture #3 - Queing Theory
9/33
9
Distribution of Arrivals
Defining the arrival process for a waiting line
involves determining the probabilitydistribution for the number of arrivals in agiven period of time
Arrivals usually occur randomly and
independently of other arrivals The Poisson Probability Distribution
provides a good description of the
arrival pattern.
STRUCTURE OF A WAITING
LINE SYSTEM
8/7/2019 D S Lecture #3 - Queing Theory
10/33
10
Distribution of Service Times
The time a customer spends at the service
facility once the service has started startswhen the customer begins to place the orderwith the server and ends when the customerhas received the order
Probability distribution is assumed to followan Exponential Probability distribution
STRUCTURE OF A WAITING
LINE SYSTEM
8/7/2019 D S Lecture #3 - Queing Theory
11/33
11
Queue Discipline
In general, for most customer-oriented
waiting lines, the units waiting forservice are arranged on a first-come,first served basis: referred to as anFCFS queue discipline.
Other types: Entering and leaving a elevator
Priority cases ( the aged, the disabled)
STRUCTURE OF A WAITING
LINE SYSTEM
8/7/2019 D S Lecture #3 - Queing Theory
12/33
12
Steady-State Operation
I.e. normal state.
When Best Patty Co. Ltd opens in themorning, no customers are in therestaurant. Gradually, activity builds up
to a normal or steady state.
Start-up period transient period
Normal period Steady-state
STRUCTURE OF A WAITING
LINE SYSTEM
8/7/2019 D S Lecture #3 - Queing Theory
13/33
13
QUEUING MODELS 5 TYPES
1. Single-Channel Waiting Line Modelwith Poisson Arrivals and ExponentialService Times (infinite callingpopulation - M/M/1) - Basic;
2. Multiple-Channel Waiting Line Model with PoissonArrivals and Exponential Service Times;
3. Single-Channel Waiting Line Model with PoissonArrivals and Arbitrary Service Times; (M/G/1)
4. Multiple-Channel Model with Poisson Arrivals,Arbitrary Service Times, and no waiting line;(M/G/k)
5. Waiting Line Models with Finite Calling Populations(M/M/1)
8/7/2019 D S Lecture #3 - Queing Theory
14/33
14
Operating Characteristics are calculatedbased on the method chosen as a result of
the queuing/waiting line trend; If the Operating Characteristics are
unsatisfactory in terms of meeting thecompany standards of service, thenManagement should consider alternativedesigns or plans for improving the waitingline operation.
QUEUING MODELS USE OF
THE RESULTS
8/7/2019 D S Lecture #3 - Queing Theory
15/33
15
Waiting line models often indicate whereimprovements in Operating
Characteristics are desirable. The decision of how to modify the waiting
line configuration to improve the operatingcharacteristics must be based on the
insights and creativity of the analyst. To make improvements in the waiting line
operation, analysts often focus on ways toimprove the service rate
QUEUING MODELS USE OF
THE RESULTS
8/7/2019 D S Lecture #3 - Queing Theory
16/33
16
Generally, service rate improvementsare obtained by making either or both
the following changes:1. Increase the mean service rate by
making a creative design change or byusing new technology
2. Add service channels so that morecustomers can be servedsimultaneously.
QUEUING MODELS USE OF
THE RESULTS
8/7/2019 D S Lecture #3 - Queing Theory
17/33
17
OPERATING
CHARACTERISTICS(PERFORMANCE MEASURES)
1. The Probability that no units are in the system.
2. The Average Number of units in the waiting line
3. The Average Number of units in the system (thenumber of units in the waiting line plus the number ofunits being served)
4. The Average Time a unit spends in the waiting line
5. The Average Time a unit spends in the system (thewaiting time plus the service time)
6. The Probability that an arriving unit has to wait forservice.
7. The probability of n units in the system
8/7/2019 D S Lecture #3 - Queing Theory
18/33
18
Single-Channel Waiting Line Model
with Poisson Arrivals andExponential Service Times -infinitepopulation M/M/1) - Basic
i.e.1. No finite calling populations
2. Steady-state operating characteristics fora single-channel waiting line;
3. Arrival follows a Poisson probabilitydistribution
4. Service times follow an Exponential
probability distribution
8/7/2019 D S Lecture #3 - Queing Theory
19/33
19
Single-Channel Waiting
Line Model with Infinitepopulation - M/M/1) - Basic Applicable only when the the mean service rate is greater than the mean arrival rate ,
I.e./ < 1. If this condition does not exist, the waiting
line will continue to grow without limitbecause the service facility does not have
sufficient capacity to handle the arrivingunits.
In the Operating Characteristics(formulas/performance measures, for this
model, >
.
8/7/2019 D S Lecture #3 - Queing Theory
20/33
20
=> the mean number of arrivals pertime per time period (the mean arrivalrate)
=> the mean number of services pertime period ( the mean service rate)
Single-Channel Waiting
Line Model with Infinitepopulation
8/7/2019 D S Lecture #3 - Queing Theory
21/33
21
1. The Probability that no units are in the system
Po= 1 2. The Average Number of units in the waiting line
Lq = 2 .( -
3. The Average Number of units in the system (the number ofunits in the waiting line plus the number of units being served)
L = Lq +
Single-Channel Waiting LineModel with Infinitepopulation
8/7/2019 D S Lecture #3 - Queing Theory
22/33
22
4. The Average Time a unit spends in the waitingline
Wq = Lq
5.The Average Time a unit spends in thesystem (the waiting time plus the service time
W = Wq + 1
Single-Channel Waiting Line
Model with Infinitepopulation
8/7/2019 D S Lecture #3 - Queing Theory
23/33
23
6. The Probability that an arriving unit has to wait forservice.
Pw=
7. The probability of n units in the system
Pn = (/)nPo
Single-Channel Waiting Line
Model with Infinitepopulation
8/7/2019 D S Lecture #3 - Queing Theory
24/33
24
In the case of Best Patty Co. Ltd, the mean
arrival rate = 0.75 customers perminute and a mean service rate =1customer per minute. Thus >;
Thus, the equations above can be used
to provide Operating Characteristics forthe Best Patty Co. Ltd single-channelwaiting line - M/M/1) - Basic .
Single-Channel Waiting LineModel with Infinite
population Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
25/33
25
1. The Probability that no units are in the system
Po= 1 =1 0.75 = 0.25
1
2. The Average Number of units in the waiting line
Lq = 2 . = 0.752 . = 2.25 customers( ) 1(1 0.75)
3. The Average Number of units in the system (thenumber of units in the waiting line plus the number ofunits being served)
L = Lq + = 2.25 + 0.75 = 3 customers
1
Single-Channel Waiting LineModel with Infinite population
Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
26/33
26
4. The Average Time a unit spends in the waitingline
Wq = Lq = 2.25 = 3 minutes 0.75
5.The Average Time a unit spends in the system(the waiting time plus the service time
W = Wq + 1 = 3 + 1 = 4 minutes 1
Single-Channel Waiting LineModel with Infinite population
Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
27/33
27
6. The Probability that an arriving unit has to wait forservice.
Pw= = 0.75=0.75
1
7. The probability of n units in the system
Pn =(/
)
nPo
This equation can be used to determine theprobability of any number of customers in thesystem.
Single-Channel Waiting LineModel with Infinite population
Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
28/33
28
The probability of n units in the system
Pn = (/)n
Po
P1= (0.75/1)1 x 0.25= 0.75 x 0.25 = 0.1875
P2 = (0.75/1)2
x 0.25
= 0.5625 x 0.25 = 0.1406
Single-Channel Waiting LineModel with Infinite population
Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
29/33
29
The probability of n units in the system
Pn = (/)n
Po
Number of Customers Probability
0 0.2500
1 0.1875
2 0.1406
3 0.1055
4 0.0791
5 0.0593
6 or more 0. 2373
Single-Channel Waiting LineModel with Infinite population
Worked Example
8/7/2019 D S Lecture #3 - Queing Theory
30/33
30
1. Customers have to wait an average of3 minutesbefore beginning to place an order, which
appears somewhat long for a business based onfast service
2. The facts that the average number of customerswaiting in line is 2.25 and that75% of the
arriving customers have to wait for service areindicators that something should be done toimprove the waiting line operation.
Single-Channel Waiting LineModel with Infinite population
Worked Example -Interpretation
8/7/2019 D S Lecture #3 - Queing Theory
31/33
31
3. The table shows 0.2373 probability that sixor more customers are in the Best Patty
system at one time. This condition indicates afairly high probability that Best Patty willexperience some long waiting times if itcontinues to use the single-channel operation.
4. If the operation characteristics are
unsatisfactory in terms of meeting companystandards for service, Best Pattys Managementshould consider alternative designs or plans forimproving the waiting line operation
Single-Channel Waiting LineModel with Infinite population
Worked Example -Interpretation
8/7/2019 D S Lecture #3 - Queing Theory
32/33
32
In the Single-Channel Infinite CallingPopulation Model, denotes the meanarrival for the system.
For the Single-Channel with a finite callingpopulation, the mean arrival rate for thesystem varies, depending on the numberof units in the system. Instead ofadjusting for the changing system arrivalrate, in this model indicates the mean
arrival rate for each unit.
Single-Channel Infinite Calling
Population vs Single-ChannelFinite Calling Populations (M/M/1)
8/7/2019 D S Lecture #3 - Queing Theory
33/33
33
A TAXONOMY OF
QUEUING MODELSThere are many possible queuing models. For example, if the interarrival time in the basic model had been given a
different distribution (not the exponential) we would have had a different model, in the sense that the previous formulas for L,
Lq , and so on, would no longer hold. To facilitate communication among those working on queuing models, D. G.
Kendall proposed a taxonomy based on the following notation:
A/B/s
where A = arrival distribution
B = service distribution
s = number of servers
Different letters are used to designate certain distributions. Placed in the A or the B position, they indicate the arrival
or the service distribution, respectively. The following conventions are in general use:
M = exponential distribution
D = deterministic numberG = any (a general) distribution of service times
GI = any (a general) distribution of arrival times
We can see, for example, that the Xerox model is an M/M/1 model; that is, a single-server queue with exponential
inter-arrival and service times.