Transcript No Slide Title
Chapter 11 Queuing Models
• Waiting and Managing Waiting Time • Features of Queuing Systems • Single Server M/M/1 Model • Little’s Formula • Multiple Server M/M/s Model • Economic Analysis of Queues
1 - Chap 11
Waiting is Everywhere
• How much time did you spend waiting for lunches/suppers
last week?
• How much time did you spend waiting for buses to school
last week?
• How much time did you spend waiting for a banking service? • How much time did you spend waiting for checkout at
Parknshop?
• Do you know how much of
your life queues?
will be spent waiting in
― It will be in terms of years.
2 - Chap 11
Waiting is Everywhere
3 - Chap 11
Waiting is Everywhere
4 - Chap 11
Time Well Spent?
In a life time, an average person will spend- SIX MONTHS Waiting at stoplights EIGHT MONTHS Opening junk mail ONE YEAR Looking for misplaced objects TWO YEARS Unsuccessfully returning phone calls FOUR YEARS FIVE YEARS Doing housework Waiting in line SIX YEARS Eating 5 - Chap 11
Flow Time Examples
Industry
Life Insurance Consumer packaging Bank Hospital Automobile
Process
New policy application Graphic design Consumer loan Patient billing Car painting
Processing time
7 min 2 hrs 34 min 3 hrs 10 min Consumer electronics Mould making 36 hrs
Why are the flow times so long? flow time = waiting time + processing time Actual flow time
72 hrs 18 days 24 hrs 10 days 2 hrs 25 days
6 - Chap 11
Remember Me
The Customer
• I am the person who goes into a restaurant, sits down,
and patiently waits while the wait-staff does everything but take my order.
• I am the person that waits in line for the clerk to finish
chatting with his buddy.
• I am the one who never comes back and it amuses me to
see money spent to get me back.
• I was there in the first place, all you had to do was to
show me some courtesy and service.
7 - Chap 11
Features of Queueing Systems
Balk (No waiting) Calling Population Arrival process Renege (Waited) Queue configuration Queue discipline Service process Departure No future need for service 8 - Chap 11
Elements of a Queueing System
• Calling population – Source of customers (a customer is a person or thing that wants
service from an operation)
– Infinite: large enough that one more customer can always arrive to be
served
– Finite: countable number of potential customers • Arrival rate
: average number of arrivals in unit time calculated over a long period of time, usually denoted by
l,
and the inter arrival time between two arrivals = 1/
l • Service rate
: average number of customers served by a server calculated over a long period of time, usually denoted by
m,
and the average time to serve one customer = 1/
m
9 - Chap 11
Arrival Process
• Paced (deterministic) arrivals
: products off the assembly line for inspection
• Random arrivals
: patients, cars for repair, calls, Internet requests
• Batch arrivals
: parts for milling bus loads, arriving passengers at the airports,
• Singleton arrivals
: traveler checking in at the hotel reception counter, semi-finished products to an assembly station
• Inter-arrival times:
time between 2 consecutive arrivals 10 - Chap 11
Arrival Characteristics in Queues
11 - Chap 11
Service Process
• Deterministic time:
beer bottling, some assembly operations,
• Random time:
checking passport, repairing car engine, haircut, milling a part
• Batch service:
processing bus, heat treatment of metal parts, food
• Singleton service:
immigration, doctor, assembly operation 12 - Chap 11
Queue Configurations
Multiple Queues Single Queue Take a Number 3 8 6 4 11 12 10 9 5 7 2 Enter 13 - Chap 11
Factors in a Queue Discipline
14 - Chap 11
Queue elements Arrival Server
Systems with Queues
Repair center Bank branch customers customers technicians tellers Production line parts machines Queue Service time single line random multiple lines random Queue discipline FCFS FCFS: first come first served SPT: shortest processing time FCFS Airport immigration travelers officers single buffer multiple lines constant or random FCFS, SPT random FCFS 15 - Chap 11
Service Facility Arrangements
Service facility Server arrangement Parking lot Self-serve Cafeteria Servers in series Toll booths Hospital Supermarket Servers in parallel Many service centers in parallel and series, not all used by each patient Self-serve, first stage; parallel servers, second stage 16 - Chap 11
A Service Process without Variability
arriving customer server
• Inter-arrival time = service time = 6 minutes, a constant • Arrival rate = 10 customers per hour • Service rate = 10 customers per hour • With one server:
no queue, no server idling 17 - Chap 11
A Service Process with Variability
arriving customer queue server with a customer
• Arrival rate = 20 customers per hour, service rate = 60
customers per hour, one server
• Inter-arrival times and service times are random • What happen if
- 100 customers arrive between 9:45 and 10 am?
- One customer takes much longer than 1 minute to be served?
18 - Chap 11
Waiting Realities
• Inevitability of Waiting:
- Waiting results from variations in arrival time between customers and the time required to serve customers
• Economics of Waiting:
- High utilization realized at the price of customer waiting. 19 - Chap 11
Approaches to Manage Customer Waiting
• Animate:
- Disneyland distractions, elevator mirror, recorded music
• Discriminate:
- Avis frequent renter treatment (out of sight)
• Automate:
- Use computer scripts to address 75% of questions
• Control arrival times:
- Appointments, pricing “Disney-world's management of waiting lines”.
http://www.youtube.com/watch?v=6OJIy-PzCgs
20 - Chap 11
How Technology Can Provide Faster Service
• Eliminate Customer Waiting Time (24x7 service) –Automated teller machines (ATMs) –Internet access to customer accounts • Reduce Customer Waiting Time –Bar-code scanners –Optical character recognition (OCR) –Menu-driven databases
21 - Chap 11
What probability distributions are often used for inter arrival times and service times?
• Suppose you start a service business. You haven’t seen the actual
customers arrival process, but you want to have some idea about the queue you will be facing.
• So, you need to make some
assumptions about the customers arrival process, and service time distribution.
• A most commonly used distribution is the
exponential distribution : 1/
l
~ exp(λ) 1/
m
~ exp(μ)
0
inter-arrival time 1/
l 0
service time 1/
m
22 - Chap 11
Why use these assumptions?
•
In many situations, the exponential distribution assumption is a good approximation of what really happens
– Such an arrival process is also called “
Poisson process ”
» Number of customers arriving per time unit is Poisson
distributed 23 - Chap 11
Arrivals
Single-Server M/M/1Model
Queue Service system Service facility Served units Ships at sea Ship unloading system Waiting ship line Dock Empty ships
24 - Chap 11
Single-Server M/M/1 Model
• Assumptions: – Inter-arrival times follow an
Exponential distribution with mean 1/λ
• Input rate follows a Poisson Distribution with rate λ
M
– Service times follow an
Exponential distribution with rate
m
(average time to serve one customer = 1/
m
)
– Single server
1 M
– Other technical assumptions • Single queue • No limit on queue length (unlimited waiting capacity) • All units that arrive enter the queue (no balking) • Any unit entering the system stays in the queue till served (no
reneging)
• First Come - First Served (FCFS) • All units arrive
independently of each other 25 - Chap 11
Single-Server M/M/1 System: The State of the System
• n = number of customers in system (in service plus in
waiting)
• If service rate (μ) > arrival rate (λ), system can reach the
steady state
• Steady state: the probabilities of observing any particular
number of customers in the system at any two arbitrary time
t
1 and t 2 are equal
• Steady system state probability distribution
P i
= P(n =i ), i = 0, 1, 2 , …
P i
is the probability of observing i customers in a steady system at an arbitrary time
• How to compute P
i
?
26 - Chap 11
Formulas for Single-Server M/M/1 Model
Probability that the server is busy and the customer has to wait (utilization factor) Probability that the server is idle and no customers are in the system (either in the queue or being served)
=
l m
P
0 = 1 -
= 1 -
l m
Probability of exactly n customers in the system
P n n
= • P 0
l
n
= 1 -
l m
27 - Chap 11
Formulas for Single-Server M/M/1 Model
Average number of customers in the waiting line (average queue length Average number of customers in the system Average time a customer spends waiting in line to be served Average time a customer spends in the queuing system (average flow time)
L q W q
m
(
m l
-
l
) 1-
m l
-
l 1-
q
+
m
(
m l
-
l
)
m
-
l m
1 -
l
L
l
q
+ 1
m
28 - Chap 11
Example: Dr. Wang’s Clinic
• Dr. Wang runs a private walk-in clinic. She spends an
average of 10 minutes to see one patient. Four months ago, patients arrived at a rate of 4/hour.
• One afternoon, you were walking in … – What was the probability that you had to wait? – If you saw 3 or more patients waiting, you would go away.
How likely would this happen?
– What was the expected number of patients in the clinic? – What was the expected number of patients in the queue? – What was your expected waiting time? – What was your expected flow time?
29 - Chap 11
Example: Dr. Wang’s Clinic
Arrival rate
l
= 4, service rate
m
= 6, ρ= 2/3 Probability that you had to wait = probability that Dr. Wang was busy = 1– probability that she was idle = 1− (1−ρ) = 2/3 Probability of seeing 3 or more waiting {there are 4 or more patients in the clinic.} =1– P 0 – P 1 – P 2 – P 3 =1 – 1/3 – (2/3)x(1/3) – (4/9)x(1/3) – (8/27)x(1/3) =1 – 65/81=0.1975
30 - Chap 11
Example: Dr. Wang’s Clinic
• Expected number of patients in the clinic
L =
/(1 –
) = (2/3)/[1-2/3]=2 patients
• Expected number of patients in the queue
Lq= L –
= 2 – 2/3 = 1.33 patients
• Expected waiting time
Wq =
/(m
–
l)
= (2/3)/(6−4)= 1/3= 20 minutes
• Expected flow time
W = Wq + 1/
m
= 20 + 10 = 30 minutes 31 - Chap 11
Little’s Law of Service Systems
Little's Law: W - average flow time W L q q - average customer waiting time L - average number of customers in the system - average queue length The fundamental relations:
L
l W ,
L q
l
W q
32 - Chap 11
Use of Little’s Law for Dr. Wang’s Clinic Case
By Little’s Law,
• The average flow time is
W = L/λ = 1/(μ−λ) = 1/(6−4) = 30 minutes
• The average queue length is
L q
= λ
×
W q
= ρ 2 /(1−ρ) = (4/9)/(1/3) = 4/3 = 1.33 patients 33 - Chap 11
Example: Photo Express
• On average 10 customers arrive per hour at a Photo Express
to process film.
• There is one clerk in attendance that on average spends 4
minutes per customer.
1. What are the average queue length and the average number of customers in the system?
2. What are the average waiting time in queue and the average time spent in the system?
3. What is the probability of having 2 or more customers waiting in queue?
34 - Chap 11
Photo Express: Queue Length
• Poisson arrival: l
= 10 customers per hour
• Exponential service time: 1/ m
= 4 minutes
• Service rate = 15 customers per hour • Utilization: 10 / 15 0 .
6667 • Solution:
L q
2 1
0 .
6667 2 1
-
0 .
6667
1 .
3336 customers L
1
0 .
6667 1
-
0 .
6667
2 .
0003 customers 35 - Chap 11
Photo Express: Other Results
• Average waiting time
W q
L q
/ l 1 .
3336 / 10 0 .
1334 hours 8 minutes • Average time in system
W
L
/ l 2 .
0003 / 10 0 .
2 hours 12 minutes • Probability of 3 or more customers in system (2 or more
customers waiting in queue) P(n ≥ 3) = 1 – P 0 – P 1 – P 2 = 0.2963
36 - Chap 11
M/M/s Multiple-Server Model
l Arrival rate customers/hr Average throughput l customers/hr Servers queue Assumption : l
s
m s = number of servers = l / (
s
m
)
Service rate (per server) customers/hr m • Customers only form one queue • The first customer in the queue will be served by the next empty
server 37 - Chap 11
M/M/s Multiple-Server Model
• Two or more independent servers serve a single waiting line • Poisson arrivals, exponential service time, infinite population • Probability that no customers are in the system (all servers are
idle): 1
P
0 = n=s-1
n=0 1 n!
l m
n
+ 1 s!
l m
s s
m
s
m
-
l
38 - Chap 11
Basic M/M/s Multiple-Server Model
Probability of exactly n customers in the system
P n
= 1 s! s
n-s
1 n!
l m
n
l m
n P
0 , for n > s
P
0 , for 0
n
s
Probability that an arriving customer must wait
P w
Average number of customers in system 1 s!
l m
s s s
m
-
l
0
lm
(
l
/
m
)
s
(s - 1)!(s
m
-
l
) 2 0 +
l m
39 - Chap 11
Basic M/M/s Multiple-Server Model
Average time a customer spends in system Average number of customers in queue Average time a customer spends in queue Utilization factor W =
L
l
L q
= L
– l m
W q
= W
–
1
m
=
L q
l
=
l
/(s
m)
40 - Chap 11
MS Course Ware
• The formulas for determining the performance measures of
M/M/s model are complicated.
• An Excel template “MMs Economic Analysis.xls” for the M/M/s
model which is available from Moodle, will calculate all the measures of performance.
41 - Chap 11
Multiple-Server System: Example
Student Health Service Waiting Room
l
= 10 students per hour
m
= 4 students per hour per service representative s = 3 representatives s
m
= (3)(4) = 12 Probability no students are in the system Average number of students in the service area
P
0 = 0.045
L = 6 42 - Chap 11
Multiple-Server System: Example
Average waiting time students spend in the service area Average number of students waiting to be served Average time students wait in line W = L /
l
= 0.60 = 36 minutes
L W q q
= L -
l
/
m
= L
q
/
l
= 3.5
= 0.35 hours = 21 minutes Probability that a student must wait
P w
= 0.703
43 - Chap 11
Multiple-Server System: Example
Add a 4th server to improve service.
Re-compute operating characteristics:
P
0 = 0.073 (probability of no students) L = 3.0 students W = 0.30 hour, 18 minutes in service
L q W q
= 0.5 students waiting = 0.05 hours, 3 minutes waiting, versus 21 earlier
P w
= 0.31 (probability that a student must wait) 44 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
M/M/1 Queuing System Mr. Chan is a stock trader on the floor of the Hong Kong Stock Exchange for the firm SHK Security. Stock transactions arrive at a mean rate of 20 per hour. Each order received by Mr. Chan requires an average of two minutes to process.
Orders arrive at a mean rate of 20 per hour or one order every 3 minutes. Therefore, in a 15 minute interval the average number of orders arriving will be
l
= 15/3 = 5.
45 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
Service Rate Question: What is the mean service rate per hour?
Answer: Since Mr. Chan can process an order in an average time of 2
minutes (= 2/60 hr.), then the mean service rate, µ, is µ =
1/(mean service time), or 60/2 = 30/hr.
46 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
Average Waiting Time Question: What is the average time an order must wait for process?
Answer: This is an M/M/1 queue with
W q
=
/(µ - l
) = (2/3)/(30 - 20)
l
= 20 per hour and
m
= 30 per hour. The average time an order waits in the system is: = 1/15 hour or 4 minutes 47 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
Average Time in the System Question: What is the average time an order must wait from the time Mr. Chan receives the order until it is finished being processed (i.e. its flow time)?
Answer: The average time an order stays in the system is: W = W
q
+ service time = 4 + 2 = 6 minutes 48 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
Average Queue Length Question: What is the average number of orders waiting to be processed?
Answer: The average number of orders waiting in the queue is:
L q
=
l
x W
q
= 20 x 1/15 = 4/3 49 - Chap 11
Example: SHKS (A)
Utilization Factor Question: What percentage of the time is Mr. Chan processing orders?
Answer: The percentage of time Mr. Chan is processing orders is equivalent to the utilization factor,
l
/
m
. Thus, the percentage of time he is processing orders is:
l
/
m
= 20/30 = 2/3 or 66.67% 50 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (A)
M/M/2 Queuing System SHK Security has begun a major advertising campaign which it believes will increase its business 50%. To handle the increased volume, the company has hired an additional floor trader, Mr. Wong, who works at the same speed as Mr. Chan.
Note that the new arrival rate of orders,
l
than that of problem (A). Thus,
l
, is 50% higher = 1.5(20) = 30 per hour.
51 - Chap 11
Example: SHKS (B)
Sufficient Service Rate?
Question: Why will Mr. Chan alone not be able to handle the increase in orders?
Answer: Since Mr. Chan processes orders at a mean rate of µ = 30 per hour, then
l
= µ = 30 and the utilization factor is 1. This implies the queue of orders will grow infinitely large. Hence, Mr. Chan cannot handle this increase in demand.
52 - Chap 11
Economic Analysis of Queuing Systems Example: SHKS (C)
The advertising campaign of SHK Security (see problems (A) and (B)) was so successful that business actually doubled. The mean rate of stock orders arriving at the exchange is now 40 per hour and the company must decide how many floor traders to employ. Each floor trader hired can process an order in an average time of 2 minutes.
Based on a number of factors the brokerage firm has determined the average waiting cost per minute for an order to be $5.0. Floor traders hired will earn $200 per hour in wages and benefits. Using this information compare the total hourly cost of hiring 2 traders with that of hiring 3 traders.
53 - Chap 11
Economic Analysis of Waiting Lines ― A Simple Tradeoff Model
• Parameters:
- c w
= Hourly waiting cost of one customer
- c s
= Hourly cost per server - s = Number of servers as the decision variable Total Hourly Cost = (Total cost of servers per hour) + (Total cost of waiting)
min
TC
(
s
) subject to
c s s s
l
c w
m
L q
We can also use service rate as a decision variable.
54 - Chap 11
Cost Relationship in Queue Analysis
Total Cost Cost of Servers Cost of Waiting Optimal Number Of Servers 55 - Chap 11
Example: SHKS (C)
Economic Analysis of Waiting Lines Total Hourly Cost = (Total salary cost per hour) + (Total waiting cost for orders in the system) = ($200 per trader per hour) x (Number of traders) + ($300 waiting cost per hour) x (Average number of orders waiting in the system) = 200s + 300L q .
56 - Chap 11
Example: SHKS (C)
• Use the Excel template “MMs Economic Analysis.xls to
calculate L q with
l
= 40/hr and
m
= 30/hr s 1 2 3 4 Lq #N/A 1.067
0.145
0.026
57 - Chap 11
Example: SHKS (C)
Cost with Two Floor Traders (Servers) L q = 1.067
Total Cost = (200)(2) + 300(1.067) = $720 per hour Cost with Three Floor Traders (Servers) L q = 0.145
Total Cost = (200)(3) + 300(0.145) = $643 per hour Cost with Four Floor Traders (Servers) L q = 0.026
Total Cost = (200)(4) + 300(0.026) = $808 per hour 58 - Chap 11
Example: SHKS (C)
System Cost Comparison Wage 2 Traders 3 Traders 4 Traders Cost/Hr $400 $600 $800 Waiting Cost/Hr $320 $43 $8 Total Cost/Hr $720 $643 $808 Thus, the cost of having 3 traders is the lowest. SHK should hire 3 floor traders.
59 - Chap 11