Ozcan: Chapter 9 Productivity Lecture 2 ISE 468 ETM 568 Dr. Joan Burtner.
ISE 491 Fall 2009 Dr. Burtner
description
Transcript of ISE 491 Fall 2009 Dr. Burtner
11
Ozcan: Chapter 14.Queuing Models and Capacity Planning
ISE 491 Fall 2009Dr. Burtner
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 22
Outline
Learning Objectives Queuing System Characteristics
Population Source Servers Arrival Patterns Service Patterns
Specific Queue Characteristics Measures of Queuing System Performance Infinite Source-Models Model Formulations
Single Server (M/M/1) Multi Server (M/M/s>1)
Capacity Planning
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 33
Learning Objectives
Describe queuing systems and their potential applications in health care services.
Describe various queuing model formulations.
Analyze measures of queuing system performance
Recognize queuing concepts and their relationship to capacity planning.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 44
Location: Hospital Outpatient PharmacyDay & Time: Monday, 11:00 AMAverage number in line: 20Idle servers: 0Average wait time: 15 minutes
Location: Hospital Outpatient PharmacyDay & Time: Monday, 3:30 PMAverage number in line: 0Idle servers: 3Average wait time: 0 minutesHourly wage per idle pharmacist: $40
OZCAN: Figure 14.1 Queue Phenomenon
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 55
Queuing Models
Develop a mathematical approach to the analysis of waiting lines
Weigh the cost of providing a given level of service capacity (i.e., shortening wait times) against the potential costs of having customers wait
Reduce Wait Times Minimizing costs
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 66
Costs
Waiting Costs Salaries paid to employees while they wait for
service (e.g., physicians waiting for an x-ray or test result)
Cost of waiting space Loss of business due to wait
Balking customers Reneging customers
Capacity Costs-- the cost of maintaining the ability to provide service
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 77
Total cost
Cost of service capacity
Waiting line cost
Optimum capacity
Health care service capacity (servers)
Cos
tFigure 14.2 Healthcare Service Capacity and Costs
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 88
Queuing Model Characteristics
The population source Infinite Finite
The number of servers (channels) Single line or multiple line Single phase or multiple phase
Arrival and service patterns Queue discipline (order of service)
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 99
The Population Source
Finite source Customer population where potential number of
customers is limited Infinite source
Customer arrivals are unrestricted Customer arrivals exceed system capacity Exists where service is unrestricted The model used in our text
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1010
FacilityPopulation
(Patient Source)Queue
(Waiting Line)Service
Arrivals ExitServer
Ozcan: Figure 14.3 Queuing Conceptualization of Flu Inoculations
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1111
The Number of Servers
Capacity is a function of the capacity of each server and the number of servers being used
Servers are also called channels Systems can be single or multiple channel,
and consist of phases Most healthcare systems are multiple-line
multiphase systems.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1212
Population Receptionist(server)
Arrivals
Queue-1Nurse
(server)Physician(server)
Queue-2
Exit
Ozcan: Figure 14.4 Conceptualization of a Single-line, Multi-phase System
Phase -1
Queue-3
Phase -2
Phase -3
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1313
Ozcan: Figure 14.5 Multi-Line Queuing Systems
NursesProviding
Inoculations
Physicians-Initial Evaluation
DiagnosticTesting
Physicians-Interventions
Emergency Room
Phase-1 Phase-2 Phase-3
Line-1
Line-2
Line-3
Patients
Queue
Queue-3Queue-2Queue-1
Line-1
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1414
Arrival and Service Patterns
Both arrival and service patterns are random, thus causing waiting lines
Models assume that: Customer arrival rate can be described by a
Poisson distribution Service time can be described by a negative
exponential distribution
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1515
OZCAN: Figure 14.7 Measures of Arrival Patterns
9:00pm
Inter-arrival time
10:00pm
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1616
Ozcan: Figure 14.8 Poisson Distribution
0 1 2 3 4 5 6 7 8 9 10
Patient arrivals per hour
Pro
babi
lity
.20
.05
.10
.15
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1717
Exponential and Poisson Distributions
If service time is exponential, then the service rate isPoisson. If the customer arrival rate is Poisson, then the inter-arrival time (i.e., the time between arrivals) is exponential.
Example: If a lab processes 10 customers per hour(rate), the average service time is 6 minutes. If the arrival rate is 12 per hour, then the average time between arrivals is 5 minutes.
Thus, service and arrival rates are described by the Poisson distribution and inter-arrival times and service times are described by a negative exponential distribution.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1818
Queue Behavior
•Balking• Patients who arrive and see big lines (the
flu shot example) may change their minds and not join the queue, but go elsewhere to obtain service; this is called balking.
•Reneging• If patients join the queue and are
dissatisfied with the waiting time, they may leave the queue; this is called reneging.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 1919
Queue Discipline
•Queue discipline refers to the order in which patients are processed.•Adaptations in healthcare:
First-Come, First-Served (may be the most common practice) Most Serious First-Served (emergency department triage system) Shortest Processing Time, First Served (sometimes applied in operating room scheduling)
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2020
System Metrics
Average number of patients waiting (in line or in system)
Average time patients wait (in line or in system)
System utilization (% of capacity utilized)
Implied costs of a given level of capacity and its related waiting line
Probability that an arriving patient will have to wait for service
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2121
0 100%System UtilizationAvg.
Num
ber o
r tim
e wa
iting
in li
ne
Increases in system utilization are achieved at the expense of increases in both the length of the waiting line and the average waiting time.
Most models assume that the system is in a steady state where average arrival and service rates are stable.
Tradeoffs
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2222
Ozcan: Exhibit 14.1 Queuing Model Classification
A: specification of arrival process, measured by inter-arrival time or arrival rate. M: negative exponential or Poisson distributionD: constant valueK: Erlang distributionG: a general distribution with known mean and variance
B: specification of service process, measured by inter-service time or service rateM: negative exponential or Poisson distributionD: constant valueK: Erlang distributionG: a general distribution with known mean and variance
C: specification of number of servers -- “s”
D: specification of queue or the maximum numbers allowed in a queuing system
E: specification of customer population
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2323
Examples of Typical Infinite-Source Models
•Single channel, M/M/s
•Multiple channel, M/M/s>1,where “s” designates the number of channels (servers).
•These models assume steady state conditions and a Poisson arrival rate.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2424
OZCAN: Exhibit 14.2 Queuing Model Notation
arrival ratearrival rate service rateservice rateLLqq average number of customers waiting for service average number of customers waiting for serviceL L average number of customers in the system average number of customers in the system (waiting or being served)(waiting or being served)WWqq average time customers wait in line average time customers wait in lineW average time customers spend in the systemW average time customers spend in the system system utilizationsystem utilization1/1/ service time service timePPoo probability of zero units in system probability of zero units in systemPPnn probability of n units in system probability of n units in system
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2525
Infinite Source Models: Model FormulationsFive key relationships that provide basis for queuing formulations and are common for all infinite-source models:
1.The average number of patients being served is the ratio of arrival to service rate.
r
2.The average number of patients in the system is the average number in line plus the average number being served.
rLL q
3. The average time in line is the average number in line divided by the arrival rate.
q
q
LW
4. The average time in the system is the sum of the time in line plus the service time.
1
qWW
5. System utilization is the ratio of arrival rate to service capacity.
s
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2626
)(
2
qL
10P
Performance MeasuresSingle server or single team
n
n PP
0
Single Channel, Poisson Arrival and Exponential Service Time (M/M/1).
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2727
Example 14.1
A hospital is exploring the level of staffing needed for a booth in the local mall where they would test and provide information on diabetes. Previous experience has shown that, on average, every 15 minutes a new person approaches the booth.
A nurse can complete testing and answering questions, on average, in 12 minutes. If there is a single nurse at the booth, calculate system performance measures including the probability of idle time and of 1 or 2 persons waiting in the queue.
What happens to the utilization rate if another workstation and nurse are added to the unit?
Single Channel, Poisson Arrival and Exponential Service Time (M/M/1).
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2828
Arrival rate: λ = 1(hour) ÷ 15 = 60(minutes) ÷ 15 = 4 persons per hour.Service rate: μ = 1(hour) ÷ 12 = 60(minutes) ÷ 12 = 5 persons per hour.
8.54
rAverage persons served at any given time
2.3)45(5
42
qL
Number of persons in the system 48.2.3
qLL
488.042.3
q
q
LW Minutes of waiting time in the queue
601248560481
qWW Minutes in the system
(waiting and service)
Persons waiting in the queue
Example 14.1 Solution: 1
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 2929
2.8.154110
P
16.)8)(.2(.)8)(.2(.54)2(. 1
11
01
PP
128.)64)(.2(.)8)(.2(.54)2(. 2
22
02
PP
Calculate queue lengths of 0, 1, and 2 persons Arrival rate: λ = 1(hour) ÷ 15 = 60(minutes) ÷ 15 = 4 persons per hour.Service rate: μ = 1(hour) ÷ 12 = 60(minutes) ÷ 12 = 5 persons per hour.
Example 14.1 Solution: 2
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3030
Example 14.1 Solution: 3
Current system utilization (s=1)
%.805*1
4
s
System utilization with an additional nurse (s=2)
%.405*2
4
s
System utilization decreases as we add more resources to it.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3131
FIG.14.10 EXCEL SETUP AND SOLUTION TO THE DIABETES INFORMATION BOOTH PROBLEM
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3232
Multi-Channel, Poisson Arrival and Exponential Service Time (M/M/s>1)
02!1P
ssL
s
q
1
0
0
1!!
1
s
n
sn
ssn
P
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3333
Multi-Channel, Poisson Arrival and Exponential Service Time (M/M/s>1)
a
qw W
WP
Wa = the average time for an arrival not immediately servedPw = probability that an arrival will have to wait for service.
s
Wa1
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3434
Multi-Channel, Poisson Arrival and Exponential Service Time (M/M/s>1)
Expanding on Example 14.1: The hospital found that among the elderly, this free service had gained popularity, and now, during week-day afternoons, arrivals occur on average every 6 minutes 40 seconds (or 6.67 minutes), making the effective arrival rate 9 per hour.
To accommodate the demand, the booth is staffed with two nurses working during weekday afternoons at the same average service rate.
What are the system performance measures for this situation?
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3535
Multi-Channel, Poisson Arrival and Exponential Service Time (M/M/s>1)
Solution:This is an M/M/2 queuing problem. The WinQSB solution provided in Exhibit 14.5 shows the 90% utilization. It is noteworthy that now each person has to wait on average one hour before they can be served by any of the nurses.
On the basis of these results, the healthcare manager may consider expanding the booth further during those hours.
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3636
Figure 14.12 System Performance Summary for Expanded Diabetes Information Booth with M/M/2
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3737
Figure 14.13 System Performance Summary for Expanded Diabetes Information Booth with M/M/3
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3838
Figure 14.14 Capacity Analysis
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 3939
Multi-Channel, Poisson Arrival and Exponential Service Time (M/M/s>1) Table 14.1 Summary Analysis for M/M/s Queue for Diabetes Information Booth
Performance Measure2 Stations 3 Stations 4 Stations
Patient arrival rate 9 9 9
Service rate 5 5 5
Overall system utilization 90% 60% 45%
L (system) 9.5 2.3 1.9
Lq 7.7 0.5 0.1
W (system) – in hours 1.05 .26 0.21
Wq – in hours 0.85 .06 0.01
Po (idle) 5.3% 14.6% 16.2%
Pw (busy) 85.3% 35.5% 12.8%
Average number of patients balked 0 0 0
Total system cost in $ per hour 790.53 294.91 302.89
ISE 491 Fall 2009 Dr. Burtner Ch. 14 Ozcan 4040
Assumptions and Limitations
Steady state of arrival and service process
Independence of arrival and service is assumed
The number of servers is assumed fixed
Little ability to incorporate important exceptions to the general flow
Difficult to model all but simple processes