33
Waiting Line Models (Queuing Theory)

7 Waiting Lines

Embed Size (px)

Citation preview

Page 1: 7 Waiting Lines

Waiting Line Models

(Queuing Theory)

Page 2: 7 Waiting Lines

© 2012 Lew Hofmann

Lines in Operations Management

• Assembly lines

• Production lines

• Trucks waiting to unload or load

• Workers waiting for parts

• Customers waiting for products

• Broken equipment waiting to be fixed

• Customers waiting for service

Page 3: 7 Waiting Lines

© 2012 Lew Hofmann

Costs The cost of waiting

Losing customers because of long lines• Reneging: Customers get tired of waiting and leave

• Balking: Customers see a long line and don’t get in line. Paying employees to wait for something they need.

(waiting for parts, supplies, deliveries, etc.)

Unusable (idle) equipment awaiting repairs• EG: Broken assembly line machinery.

The cost of service Paying people to provide service to customers Customers can be people, machines, or other objects needing service.

Page 4: 7 Waiting Lines

© 2012 Lew Hofmann

Cost of Providing Service

• Paying repairmen to fix broken machines

• Paying dock workers to load and unload trucks

• Paying customer-service people

• Using more production people to speed up the line

• Leasing of service equipment and facilities

• Paying checkout cashiers

Page 5: 7 Waiting Lines

© 2012 Lew Hofmann

Queuing System Costs

Number of Servers

Costs

Cost of Servicegoes up as you pay for more servers.

Costs of Waitinggoes down as service improves.

Total Cost

Optimal # of servers

Note that the lowest cost system requires some customer waiting.

Fewer servers often means longer waiting for customers. Many servers

means little or no waiting, but higher service costs.

Page 6: 7 Waiting Lines

© 2012 Lew Hofmann

What Queuing Models Tell Us.

• Average time in line for a customer.

• Average number of customers in line.

• Average time in the system for a customer.

• Average number of customers in the system at any time.

• Probability of n number of customers in the system at any given time.

• NOTE: “In The System” includes customers who are in line plus the customers being served.

Page 7: 7 Waiting Lines

© 2012 Lew Hofmann

ARRIVAL SYSTEM(How customers arrive)

QUEUE(The nature of the waiting line or lines of customers)

The Waiting Line System

SERVICE FACILITY(How customers progress through the service facility)

Page 8: 7 Waiting Lines

© 2012 Lew Hofmann

Waiting Line Models

Customer population

Served customers

Arrival System

Service System

Waiting line (Queue)

Priority rule

Service facilities

The sequence in which customers are admitted into the service facility.

Page 9: 7 Waiting Lines

© 2012 Lew Hofmann

Arrival System

• Arrival Populations are either…

• Limited (EG: Only people age 21 or over.)

• Unlimited (EG: cars arriving at a toll booth)

• Arrival Patterns are either…

• Random (Each arrival is independent)

• Scheduled (EG: Doctor’s office visits)

• Behavior of the Arrivals

• Balking (Seeing a long line and avoiding it.)

• Reneging (Get tired of waiting and leave the line)

• Jockeying (Switching lines)

Page 10: 7 Waiting Lines

© 2012 Lew Hofmann

The Queue

• Queue Length is either..

• Unlimited (EG: cars in line at a toll booth)

• Limited (Finite) EG: # of e-mail messages allowed.

• Queue Discipline (order of service)

• FIFO (First-In, First-Out)

• LIFO (Last-In, First-Out)

• SIRO (Service In Random Order)

• Priority

Page 11: 7 Waiting Lines

© 2012 Lew Hofmann

The Service Facility

• Channels• How many paths (ways to get through the

system) are there after getting in line?• EG: McDonalds drive-thru is one channel.

• Phases• How many stops must a customer make,

after getting in line?(Single-phase means only one stop for service.)

• McDonalds drive-thru is a three-phase system: 1. Order 2. Pay 3. Pick-up

Page 12: 7 Waiting Lines

© 2012 Lew Hofmann

Single-channel, Single-phaseOne way through the system

and one stop for service

Service Facility

Page 13: 7 Waiting Lines

© 2012 Lew Hofmann

Multi-channel, Single-phase

Service Facility

Service Facility

Once in line, you have at least two choices of how to get through the system, but only one stop.

Page 14: 7 Waiting Lines

© 2012 Lew Hofmann

Multi-channel, Multi-phase

Service Facility

Service Facility

Service Facility

Service Facility

Once in line, you have at least two choices (channels) of how to get through the system and at least two stops (phases).

Page 15: 7 Waiting Lines

© 2012 Lew Hofmann

Four Single-channel, Single-phase Systems(Once in line, you only have one channel and one stop.)

Service Facility

Service Facility

Service Facility

Service Facility

Page 16: 7 Waiting Lines

© 2012 Lew Hofmann

One, Multi-channel, Single-Phase System (Once in line you have four possible paths through the system, but only one stop.)

Service Facility

Service Facility

Service Facility

Service Facility

Page 17: 7 Waiting Lines

© 2012 Lew Hofmann

Assumptions of our Models

• The Rate of Service must be faster than the Rate of Arrivals. (It is unsolvable if customers arrive faster than they can be served.)

• FIFO (First In, First Out) (Customers are served in the order they arrive.)

• Arrivals are unlimited (infinite)

• Arrivals are random rather than scheduled.

• Customers arrive independently of each other.

• Service times can vary from one customer to another, and are independent of each other. (Customers may have different service needs and times.)

Page 18: 7 Waiting Lines

© 2012 Lew Hofmann

Queuing ProblemAt a large Naval Ship Repair Facility mechanics have to make frequent trips to the tool crib for parts and specialized equipment. (Arrivals are infinite since mechanics can come as often as need, even though the population of customers is finite.)

Records indicate that the tool crib serves an average of 18 mechanics each hour, but is capable of serving 20 per hour. If mechanics are paid $30 per hour and the tool crib attendants make $9 per hour, would it be more cost effective to have one or two attendants in the tool crib?

The service rate is always the average time for one server, regardless of how many servers there are in the system. Here it is 20, which is higher than the arrival rate of 18. If the service rate had been lower than the arrival rate, the problem would not be solvable, because customers would arrive faster than they could be served.

Page 19: 7 Waiting Lines

© 2012 Lew Hofmann

1st

Attendant

2nd

Attendant

Single

Attendant

Which system is less expensive?

(It depends on the relative costs of service versus waiting.)

Page 20: 7 Waiting Lines

© 2012 Lew Hofmann

Data Entry Information for POM/QM

I ran the POM-QM model using two servers, but I could have run it with any number of servers since you always enter the service rate for one server. The POM-QM model will do the computations for more than one server.

1st

Attendant

2nd

Attendant

Single

Attendant

Mechanics (customers) arrive at an average of 18 per hour and are paid $30 per hour.

One attendant can serve 20 mechanics/customers per hr.and is paid $9 per hour.

Page 21: 7 Waiting Lines

© 2012 Lew Hofmann

Lowest Cost!

Input data

Service rate for 1 server!

Page 22: 7 Waiting Lines

© 2012 Lew Hofmann

The probability that 4 or fewer mechanics are in the system is 97.45%

Probability that the system is idle (no customers) is 37.93%

Page 23: 7 Waiting Lines

© 2012 Lew Hofmann

# of Servers

1 2

What the Model Tells Us…

• Average # customers in the

system

• Average time in the system

• Average # customers in line

• Average time in line

• Probability that the system is

idle

9.0

1.128

0.5

0.063

8.1

0.228

0.45

0.013

0.1

0.38

Once you know the optimal # of servers, make sure you run it again for that many servers in order to get the right data. But always enter the service rate for one server, regardless of how many servers.

Note: You need to run the model with 1 server if you want the info for 1 server, and run it using 2 servers to get the info for 2 servers.

Page 24: 7 Waiting Lines

© 2012 Lew Hofmann

• Average # customers in the system

• Average time spent in the system

• Average # of customers in line

• Average time in line

Results for 2 serversMake sure you use the M/M model.

Page 25: 7 Waiting Lines

© 2012 Lew Hofmann

POM/QM or Excel Solver?

POM/QM will do the cost analysis for you, so it is easier. Select the “M/M” model.

You can use the Excel Solver but it won’t do the cost analysis.

You also need to run the Excel solver once each server number you use, and then do a manual cost analysis for each in order to see which number of servers has the lowest cost.

In multiple server problems, you might have to run it for a half-dozen or more scenarios.

Page 26: 7 Waiting Lines

© 2012 Lew Hofmann

Excel Solver: 1 Server

Page 27: 7 Waiting Lines

© 2012 Lew Hofmann

Manual cost computations for one server (9 customers in the system)

• Attendants get paid $9 per hour(Cost of service is thus 1 x $9 = $9 per hour)

• Mechanics get paid $30 per hourCost of waiting is: $30 x 9 customers in system = $270

• Total cost of one server is: $270 + $9 = $279

POM-QM model

Page 28: 7 Waiting Lines

© 2012 Lew Hofmann

Excel Solver: Two Servers

Page 29: 7 Waiting Lines

© 2012 Lew Hofmann

Manual cost computations for Two servers (9 customers in the system)

• Attendants get paid $9 per hourCost of service is: $9 x 2 servers = $18 per hour

• Mechanics get paid $30 per hourCost of waiting is: $30 x 1.28 customers in system = $33.84

• Total cost of two servers is: $18 + $33.84 = $51.84

POM-QM model

Page 30: 7 Waiting Lines

© 2012 Lew Hofmann

Homework Assignment

Due next Tuesday

Problem 1: Car Repair

Problem 2: The Quarry

Use POM/OM or Excel Solver software and submit printouts to support your decisions

Page 31: 7 Waiting Lines

© 2012 Lew Hofmann

Car Repair In the service department of a car repair shop, mechanics requiring parts for auto repair or service present their request forms at the parts department counter. The parts clerk fills a request while the mechanic waits.

Mechanics arrive in a random fashion at the rate of 40 per hour, and a clerk can fill requests at the rate of 20 per hour. If the cost for a parts clerk is $6 per hour and the cost for a mechanic is $12 per hour, determine the optimum number of clerks to staff the counter. (Because of the high arrival rate, an infinite source may be assumed.)

Always enter the service rate for one server. The program will do the math once you enter the number of servers. If you enter fewer servers than can handle the arrival needs, the program will give you an error message because the computed service rate must be higher than the arrival rate.

Page 32: 7 Waiting Lines

© 2012 Lew Hofmann

The Quarry (no cost analysis in model)You are in charge of a quarry that supplies sand and stone aggregates to construction sites. Empty trucks arrive at the quarry and wait in line for loading either sand or aggregate. At the loading station they are filled with material, weighted, checked out, and then proceed to a construction site.

Currently 9 empty trucks arrive each hour (on average). In addition to waiting in line, it takes 6 minutes for a truck to be filled, weighed and checked out.

Concerned that trucks are spending too much time waiting and being filled, you evaluate the current situation and compare it to the 2 alternatives below.

Alternative 1: Speed up the loading process and add side boards to the trucks so that more material can be loaded faster. This will improve the speed of loading, but cost $50,000. Since the trucks hold more, their arrival rate would be reduced to 6 per hour and the loading time would be reduced to 4 minutes each.

Alternative 2: Add a second loading station at a cost of $80,000. The trucks would arrive at the current rate of 9 per hour. They would then wait in a common line and the truck at the front of the line would move to the next available loading station. Loading time at each of the stations is 6 minutes.

Which alternative do you recommend? (Select “No Cost” in the POM/QM waiting line model. You must decide which of the three situations would the most cost effective based on time in the system and upgrade costs.

Page 33: 7 Waiting Lines

© 2012 Lew Hofmann

Facts About Queuing

QUEUE originally (15th century) referred to the “tail of the beast” or “a tail piece.” (Not to be confused with a piece of tail.)

In the 17th century a queue became “a braid of hair.” Later it was used to refer to a pigtail.

In the 18th century a billiard stick became a queue, later changed to “que” and then to “cue”.

In the early 19th century England, to Queue was “to line up.” And that is how it is used today in England.