You are on page 1of 24

Extra

Credit
Project
IE 417: Operations Research II

Travis Guzman
Amanda Smith
David Trigg

Presented to: Dr. Sima Parisay

California State Polytechnic


University, Pomona

Due: February 7th, 2011


Table of Contents
Problem Statement .......................................................................................................................................................3
Summary of the Problem ...............................................................................................................................................4
 Kendell-Lee Notation ..................................................................................................................... 4
 Queuing System Notations: .......................................................................................................... 4
 Formulas .......................................................................................................................................... 5
 Rate Diagram .................................................................................................................................. 5
Figure 1: Rate Diagram for M/M/s/GD/∞/∞ ........................................................................................... 5
Part A: ............................................................................................................................................................................6

Figure 2: Queuing System Diagram A ....................................................................................................... 6


Part B: ............................................................................................................................................................................8
Figure 3: Queuing System Diagram B........................................................................................................ 8
Part B: Performance Measures ....................................................................................................................................10
Table 1: Performance Measures Summary ............................................................................................. 10
Part C ...........................................................................................................................................................................11

Table 2: Part A and B Comparison .......................................................................................................... 11


Part D ...........................................................................................................................................................................13
WINQSB- Performance Measures & Cost Analysis ......................................................................................................14
Table 3: Cost Table .................................................................................................................................. 14
Figure 4: WinQSB Input ........................................................................................................................... 15
Figure 5: WinQSB Output- Performance Measure ................................................................................. 15
Figure 6: WinQSB Output- System Probability Summary ....................................................................... 16
WINQSB- Sensitivity Analysis .......................................................................................................................................18
Sensitivity Analysis #1: Arrival Rate ........................................................................................................ 18
Figure 7: WinQSB Input Parameters for SA #1........................................................................................ 18
Figure 8: WinQSB Output Table- SA #1 ................................................................................................... 19
Figure 9: WinQSB Output Utilization Graph- SA #1 ................................................................................ 20
Sensitivity Analysis #2: Service Rate ....................................................................................................... 21
Figure 10: WinQSB Input Parameters for SA #2...................................................................................... 21
Figure 11: WinQSB Output Table- SA #2 ................................................................................................. 22
Figure 12: WinQSB Output Utilization Graph- SA #2 .............................................................................. 23
Figure 13: WinQSB Output Utilization Graph- SA #1 & #2 Comparison ................................................. 23
Acknowledgments .......................................................................................................................................................24

2|Page
Problem Statement

The problem statement can be found on page 1094, problem #3 from Operations
Research by Wayne L Winston, 4th edition.

In this problem, all interarrival and service times are exponential.

a. At present, the finance department and the marketing department each have one
typist. Each typist can type 25 letters per day. Finance requires that an average
of 20 letters per day be typed, and marketing requires that an average of 15
letters per day be typed. For each department, determine the average length of
time elapsed between a request for a letter and completion of a letter.

b. Suppose that the two typists were grouped into a typing pool; that is each typist
would be available to type letters for either department. For this arrangement,
calculate the average length of time between a request for a letter and
completion of the letter.

c. Comment on the results of part (a) and (b).

d. Under the pooled arrangement, what is the probability that more than .200 day
will elapse between a request for a letter and completion of the letter?

3|Page
Summary of the Problem

Before beginning the problem, we must have an understanding of the type of queuing
system presented in the problem, as well as associated formulas that will be employed
with this type of queuing system.

First, we begin with the type of queuing system given. The Kendell-Lee Notation for
Queuing Systems can be described by six characteristics:

 Kendell-Lee Notation
1/2/3/4/5/6 
Arrival process/service process/parallel servers/ queue discipline/max customers/population size

For our problem the Kendell-Lee Notation looks like the following:

M/M/s/GD/∞/∞

Where

M= Interarrival times are independent, identically distributed (iid)


s = number of parallel servers
GD = General queue discipline
∞ = infinite numbers of customers; infinite population size
 Queuing System Notations:
λ = Arrival rate approaching the system
λe= Arrival rate (effective) entering the system
µ = Maximum (possible) service rate
µe= Practical (effective) service rate
L = Number of customers present in the system
Lq = Number of customers waiting in the line
Ls = Number of customers in service
W = Time a customer spends in the system
Wq= Time a customer spends in the line
Ws= Time a customer spends in service
ρ = Traffic Intensity

4|Page
𝜋j = P(j) = Probability that j units are in the system

𝜋0 = P(0) = Probability that there are no units (idle) in the system


 Formulas (for M/M/S)

Traffic Intensity: Probability that a Server is Idle:


𝜆
ρ = 𝑠µ < 1 1
0 
s 1 ( s ) ( s ) s
i

Number Waiting in Line:


i0 i! s!(1   )
𝑃(𝑗>𝑠)𝜌
Lq = (1−𝜌)

Number in Service: Probability that Arriving Unit has to Wait:


𝜆
Ls = µ ( s ) s  0
P( j  s ) 
s!(1   )
Number in System:
L = Lq + Ls

Time Spent in Line: Probability that J Units are in Service:


𝐿𝑞
Wq = 𝜆 (s ) j  0
j 
j!
Time Spent in System:
1
Ws = µ

Time in System: Probability that Time in System is > t


𝐿 𝑃(𝑗>𝑠) 1 1−𝑒 −𝜇𝑡(𝑠−1−𝑠𝜌)
W=𝜆= + 𝑃(𝑊 > 𝑡) = 𝑒 −𝜇𝑡 {1 + 𝑃(𝑗 ≥ 𝑠) 𝑠−1−𝑠𝜌
}
𝑠µ− 𝜆 µ

***Note: Cost Formulas are located in cost analysis section for clarity****

 Rate Diagram

Figure 1: Rate Diagram for M/M/s/GD/∞/∞

5|Page
Part A:
First we extract the information the problem gives us:

s = 2 typist

Service Rate Arrival Rates

µ = 25 letters/day λFinance = 20 letters/day

λMarketing = 15 letters/day

The problem asks, “For each department, determine the average length of time elapsed
between a request for a letter and completion of a letter.” In essence this means find W,
the total time each letter is in the system for each department. The diagram below will
help in analyzing the problem:

Figure 2: Queuing System Diagram A

𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λFinance=20 μFinance=25 𝑑𝑎𝑦
𝑑𝑎𝑦 Finance Dept.
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦

𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λMarketing=15 𝑑𝑎𝑦 MarketingDept. μMarketing=25 𝑑𝑎𝑦

𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦

For this part, number of servers is equal to one when we are analyzing traffic intensity.
There is only one server in each department, and we are analyzing the departments
separately. Keep in mind that when we are analyzing departments separately we are

6|Page
looking at a M/M/1/GD/∞/∞. With the given information, we only need one formula, W,
from the M/M/1 system. We proceed to make the following calculations:

FINANCE MARKETING

Traffic Intensity: Traffic Intensity:


𝜆 𝜆
ρ=µ ρ=µ

𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
20 15
𝑑𝑎𝑦 𝑑𝑎𝑦
ρ= 𝑙𝑒𝑡𝑡𝑒𝑟𝑠 = .80 ρ= 𝑙𝑒𝑡𝑡𝑒𝑟𝑠 = .60
(25 ) (1 𝑠𝑒𝑟𝑣𝑒𝑟)(25 )
𝑑𝑎𝑦 𝑑𝑎𝑦

ρ = 0.80 ρ = 0.60
0< ρ<1  0< 0.80<1so a steady state exists 0< ρ<1  0< 0.60<1so a steady state exists
Time in System: Time in System:
1 1 1 1 1 1
W = µ−𝜆 = = day W= µ−𝜆 = = 10 day
25−20 5 25−15

We also make the following extra calculations for later comparison purposes:
Number Waiting in Line: Number Waiting in Line:
𝑝2 .802 𝜌2 .602
Lq = (1−𝜌) = (1−0.80) = 3.2 letters . Lq = (1−𝜌) = (1−0.60)= 0.90 letters

Number in System: Number in System:

𝑝 .80 𝑝 .60
L = (1−𝜌) =(1−0.80)= 4 letters L = (1−𝜌) =(1−0.60)= 1.5 letters

1
As you can see, it takes the finance department 5 day between a request for a letter and
1
completion of a letter. It only takes the marketing department 10 day between a request
for a letter and completion of a letter. This is expected since the marketing department
receives 5 less letters a day as compared to the finance department and both
departments have the same process time.

7|Page
Part B:

Essentially, Part B asks the same question as Part A, but in this scenario, the typists
can type letters for either department.

First we extract the information the problem gives us:


s = 2 typist
Service Rate Arrival Rates

µ = 25 letters/day λFinance = 20 letters/day Now, since the


typists are working
λMarketing = 15 letters/day as a team, the
Now, the service
rate doubles since arrival rates are
two typists are additive. The new
working together; arrival rate is then
together their λFinance +λMarketing =
service rate is twice 35 letters/day
as fast or 2µ =2(25
letters/day) = 50
letters/day
The diagram below will help in analyzing the problem:

Figure 3: Queuing System Diagram B

Finance Dept.
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦
𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λM&F=35 𝑑𝑎𝑦
μM&F=50 𝑑𝑎𝑦

MarketingDept.
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦

8|Page
Then, we proceed to the calculations. Now, we don’t have to perform calculations for
separate departments since we are viewing the departments as one team.

FINANCE & MARKETING


Traffic Intensity:
𝜆
ρ = 𝑠µ < 1

𝑙𝑒𝑡𝑡𝑒𝑟𝑠
35
𝑑𝑎𝑦
ρ= 𝑙𝑒𝑡𝑡𝑒𝑟𝑠 <1
(2 𝑠𝑒𝑟𝑣𝑒𝑟)(25 )
𝑑𝑎𝑦

ρ = 0.7 < 1 so a steady state does exist


Number Waiting in Line:
𝑃(𝑗>𝑠)𝜌 𝑃(𝑗>2).70 (.57)(.70)
Lq = = = = 1.33 letters
(1−𝜌) (1−0.70) (.30)

*P(j>2) value gotten from Table 6 on page 1088 of


Operations Research by Wayne L Winston, 4th edition
when ρ=.7 and s = 2.

Number in System:
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
35
𝑑𝑎𝑦
L = Lq + Ls = 1.33 𝑙𝑒𝑡𝑡. + 𝑙𝑒𝑡𝑡𝑒𝑟𝑠 = 2.73 letters
25
𝑑𝑎𝑦

Time in System:
𝐿 2.73 𝑙𝑒𝑡𝑡.
W=𝜆= 𝑙𝑒𝑡𝑡𝑒𝑟𝑠 = .078 days
35
𝑑𝑎𝑦

Thus, when the typists were grouped the new average length between a request for a
letter and completion of the letter is .078 days.

9|Page
Part B: Performance Measures
Performance measures were added for Part B of this problem because in this section of the
problem the typists reflected two true parallel servers with one arrival rate and one service rate.
All performance measures learned in lecture regarding queuing theory thus far were
calculated and placed in the summary table below. Thus you can ask any question of
the name given and find its corresponding value from this table.

Table 1: Performance Measures Summary


Performance Measures Summary
Notation Name Given Calculation Value
1 35 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λ=λE Arrival Rate = Effective AR n/a 𝑑𝑎𝑦
2 50 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ Service Rate n/a 𝑑𝑎𝑦
3 35 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μE Effective Service Rate n/a 𝑑𝑎𝑦
4
Lq Number in Queue see Part B 1.33 letters
5 𝜆 35
=
Ls Number in Service 𝜇 25 1.4 letters
6 L Number in System see Part B 2.73 letters
7 𝐿𝑞 1.33
= 35
Wq Time in Queue 𝜆 .038 days
8 1
=
1
Ws Time in Service 𝜇 25 .04 days
8 W Time in System Wq + Ws = .038+.04 .078 days
10
Probability 0 Units in
𝜋0 Service 0.1764
11 Probability 1 Unit in ( s )  0 (2 x.7) .1764
j 1
j  
𝜋1 Service j! 1! 0.2470
12 Probability 2 Units in ( s )  0 (2 x.7) .1764
j 2
j  
𝜋2 Service j! 2! 0.1729
13 P0 Probability Teller Idle P0 = 𝜋 0 0.1764
14 Probability Arriving Unit P( j  s ) 
( s )  0 (2 x.7) .1764
s

2

Pw Must Wait s!(1   ) 2!(1  .7) .5762


15 U Teller Utilization 1 - 𝜋0 = 1-0.1764 0.8236
16 Probability More than 1 1-P(J<1)=1-( 𝜋0+ 𝜋1) = 1-
P(J>1) Unit (.1764+.2470) 0.5766
17 Probability More than 2 1-P(J<2)=1-( 𝜋0+ 𝜋1+ 𝜋2)
P(J>2) Units = 1-(.1764+.2470+.1729) 0.4037

10 | P a g e
Part C
The results of parts (a) and (b) can be summarized in the following table:

Table 2: Part A and B Comparison


Finance Department Marketing Department
λ= 20 ltrs/day λ= 15 ltrs/day
Part A
W = 1/5 day = .2 days W= 1/10 day = .1 days
λ= 35 ltrs/day
Part B
W = .078 days

When we analyzed part (a) by itself, we see that the marketing department has a
shorter time in system value (W) than the finance department. This is expected because
the marketing department receives less letters a day than the finance department, while
both have the same processing capabilities. However, we can note that the effective
service rate for each department is equal to the arrival rate of letters to each
department. Therefore, within the same time span, the marketing department must type
15 letters while the finance department must type 20.

When the typists for each department work together, we have the lowest time in system
value (W) of 0.078 days. Letters arrive at a rate of 35 letters per day while they can be
processed at 50 letters per day. Because a letter can go to either typist, the W value is
lower since a letter is not waiting on only one typist. This queuing system is the more
opportune system.

The additional performance measures calculated can be compared by placing the


diagrams above and below each other. Pay attention to the white boxes that show traffic
intensity, number in line, number in system, and time in system.

11 | P a g e
ρ = .80

𝑙𝑒𝑡𝑡𝑒𝑟𝑠 Lq = 3.2 Letters 𝑙𝑒𝑡𝑡𝑒𝑟𝑠


λFinance=20 μFinance=25 𝑑𝑎𝑦
𝑑𝑎𝑦 Finance Dept.
L = 4 Letters
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
W = 1/5 day μ = 25 𝑑𝑎𝑦

WAVG = 0.15
day
ρ = .60
𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λMarketing=15 𝑑𝑎𝑦
Lq = .9 Letters MarketingDept. μMarketing=25 𝑑𝑎𝑦
L = 1.5 Letters
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦
W = 1/10 day

VS

Finance Dept.
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25 𝑑𝑎𝑦
ρ = .70
𝑙𝑒𝑡𝑡𝑒𝑟𝑠 𝑙𝑒𝑡𝑡𝑒𝑟𝑠
λM&F=35 Lq = 1.33 Letters μM&F=50
𝑑𝑎𝑦 𝑑𝑎𝑦

L = 2.73 Letters

W = .078 day

MarketingDept.
𝑙𝑒𝑡𝑡𝑒𝑟𝑠
μ = 25
𝑑𝑎𝑦

12 | P a g e
Part D

The problem states, “Under the pooled arrangement, what is the probability that more
than .200 day will elapse between a request for a letter and completion of the letter?”

In other words, what is the probability that the total time in the system, W, is more than
.200 day.

First off, we must locate the necessary formula to calculate the distribution of waiting
time. To determine the probability, we need to know the distribution of a letter’s waiting
time. On page 1090 the following formula is given:

1 − 𝑒 −𝜇𝑡(𝑠−1−𝑠𝜌)
𝑃(𝑊 > 𝑡) = 𝑒 −𝜇𝑡 {1 + 𝑃(𝑗 ≥ 𝑠) }
𝑠 − 1 − 𝑠𝜌

The t value given in the problem statement is .200 day. From Part C, we know all other
values in the equation. Now, we just have to plug and chug numbers:

𝟏 − 𝒆−𝟐𝟓(.𝟐)(𝟐−𝟏−𝟐(.𝟕))
𝑷(𝑾 > .200) = 𝒆−𝟐𝟓(.𝟐) {𝟏 + (. 𝟓𝟕) }
𝟐 − 𝟏 − (𝟐)(. 𝟕)

𝑷(𝑾 > .200) = . 𝟎𝟕

13 | P a g e
WINQSB- Performance Measures & Cost Analysis
Cost values were chosen for this problem and their relative calculations were made in
the table below. The following costs given are:

 Cs (Cost of Busy Server) at $72/day. This value was chosen based on an 8 hour
Notice work day and a typist working at $9/hour.
these  Ci (Cost of Idle Server) at $72/day. This value was chosen based on an 8 hour
values are work day and an idle typist costing the company $9/hour.
the same  Customer Waiting- It costs the company $1/day for a letter to be waiting in the
queue to be typed.
 Loss in Good Will- It costs the company $.08/letter of the goodwill of the
customer bringing the letters to be typed.
 Cost of Space- Since letters are so small, the cost of space is almost negligible
at $.02/letter. However, to reflect understanding of material, it was calculated in
the cost table.
 Loss of Customer- N/A with ∞ Capacity

Table 3: Cost Table


Cost Table
Cost Calculation In WinQSB? Value per Day
Busy Server/Typist Csρs =
($72)(.70)(2) Yes $100.80
Idle Server/Typist Ci(1-ρ)s =
($72)(1-.70)2 Yes $43.20
Customer/Letter Waiting ($/unit of time)λeWq =
𝑙𝑡𝑟𝑠
($1/day)(35 𝑑𝑎𝑦 )(.0384dy) Yes $1.34
Customer/Letter being Served
WinQSB Yes $0.70
Loss in Goodwill ($/part)λePw=
𝑙𝑡𝑟𝑠
($.08/ltr)(35 𝑑𝑎𝑦 ) (.5762) No $1.61
Cost of Space = Almost Neglible ($/part)Lq=
($.02/letter)(1.33 letters) No $0.03
Loss of Customer N/A with ∞ Capacity -----
TOTAL
$147.68

14 | P a g e
The Queuing System problem was input into WinQSB. The values calculated in the
Cost Table (Table 3) and Performance Measure Summary (Table 1) match that of
WinQSB.

Figure 4: WinQSB Input

Figure 5: WinQSB Output- Performance Measure

Additionally, WinQSB contains a probability and cumulative probability feature that can
be seen Figure 6 below:

15 | P a g e
Figure 6: WinQSB Output- System Probability Summary

The application/use of the system probability summary in Figure 6 can be utilized in situations
like:

 Finding the probability that there are more than a specific number of letters in the
system. For example, what is the probability that there are more than 5 letters in the
system? 7 letters? 10 letters? 20 letters?

P(j>5) = 1-P(j<5) = 1 – 0.8616 = .1384 = 13.84%


P(j>7) = 1-P(j<7) = 1 – 0.9322 = .0678 = 6.78%
P(j>10) = 1-P(j<10) = 1 – 0.9767 = .0233 = 2.33%

 Finding the probability that there are a specific number of letters in the system. For
example, what is the probability that there are 0 letters in the system? 11 letters? 15
letters? 20 letters?

P(j=0) = .1765 = 17.65%


P(j=11)= .0070 = .7%
P(j=15)= .0017 = .17%

 Finding the probability that there are less than a specific number of letters in the system.
For example, what is the probability that there are less than 3 letters in the system? 5
leters? 10 letters?

P(j<3) = 1-P(j>2) = 1-.5965 = .4035 = 40.65%

16 | P a g e
P(j<5) = 1-P(j>4) = 1-.8023 = .1977 = 19.77%
P(j<10)= 1-P(j>10) = 1-.9767 = .0233 = 2.33%

 Finding the probability that there are less than or equal to a specific number of letters in
the system. For example, what is the probability that there are less than or equal to 5
letters in the system? 10 letters? 20 letters?

P(j<5) = 0.8616 = 86.16%


(j<10) = 0.9767 = 97.67%
P(j<20) = 0.9993 = 99.93%

17 | P a g e
WINQSB- Sensitivity Analysis

Sensitivity Analysis #1: Arrival Rate

We chose to do our first sensitivity analysis on Arrival Rate. Currently, the arrival rate is
35 letters/day. We created a range above and below this number. From past research,
we know that a minimum of 25 letters/day arrive for the typists to type and at maximum
50 letters/day. We decided to stop at 50 because when arrival rate equals the service
rate (which is currently 50), the system will blow up due to variation.

Figure 7: WinQSB Input Parameters for SA #1

Just as expected in the screen below, as arrival rate increases:

λ↑, U↑, L↑, Lq↑, W↑, Wq↑, Pw↑

and all costs go up. This was expected because naturally, having more letters in the
system causes such a relationship (common sense).

18 | P a g e
Figure 8: WinQSB Output Table- SA #1

WinQSB also has the capabilities of showing all the performance measures given in the table above in a graphical form.
We chose to graph the system utilization graph. As was already discussed, as arrival rate increase so does the system
utilization. This is indicated by the positive slope of the graph in Figure 9.
Figure 9: WinQSB Output Utilization Graph- SA #1
Sensitivity Analysis #2: Service Rate
We chose to do our second sensitivity analysis on Service Rate. Currently, the service
rate is 25 letters/day. We created a range above and below this number. From past
research, we know that a minimum of 10 letters/day can be typed by the typists and a
maximum 35 letters/day.

Figure 10: WinQSB Input Parameters for SA #2

From the output, at service rates ranging from 10 letters/day to 17 letters/day the
system is unstable. At a service rate of 18/letters per day, the system can begin to
handle the 35 letters/day arrival rate.

In the screen below, as service rate increases:

μ↑, U↓, L↓, Lq↓, W↓, Wq↓, Pw↓

and all costs decrease. This was expected because naturally, processing letters at a
faster rate causes such a relationship (common sense).

21 | P a g e
Figure 11: WinQSB Output Table- SA #2

Again, we chose to graph the system utilization graph. As was already discussed, as service rate increases system
utilization decreases. This is indicated by the negative slope of the graph.
Figure 12: WinQSB Output Utilization Graph- SA #2

When we put the two graphs from Figures 9 and 12 side by side in Figure 13, we can
see that there is an inverse relationship of system utilization when we either increase
arrival rate (left graph) system utilization goes up or increase service rate (right graph)
system utilization goes down.

Figure 13: WinQSB Output Utilization Graph- SA #1 & #2 Comparison


Acknowledgments

Bibliography:

Chang, Yih-Long, and Kiran Desai. WinQSB Version 2.0. New York: Wiley, 2003. Print

Winston, Wayne L. Operations Research Application and Algorithms. 4th Edition. New York;

Duxbury, 2003. Print

Software Used:

 Microsoft Word 2010 (Windows)

 Microsoft Excel 2010 (Windows)

 Win QSB, Version 2.0 (Windows)

24 | P a g e