ISyE 3232A Homework 3HINTS September 6, 2019 cat

Due Tuesday September 10

For job j let dj be its deadline or due date (this is a datum), let wj be its weight (this is a datum)
and let tj be the time at which it is completed, called the completion time or the finish time. tj depends
on the schedule even if process times are deterministic. The release time of a job (this is a datum) is the
earliest time at which processing of the job may begin.
The job j is late if tj > dj . The job’s lateness is tj − dj , which can be negative. The job’s tardiness is
[tj − dj ]+ ≡ max{tj − djP , 0}, the positive part of the lateness. The flowtime of a set of jobs P
J is the sum
of the completion times j∈J tj . It should be obvious to you that the weighted flowtime is j ∈ Jwj tj .
The makespan of a set of jobs J is maxj∈J tj . The total tardiness (lateness) of a set of jobs J is the sum
of the tardiness (lateness) of each jobs j ∈ J. It should be obvious that the weighted tardiness of a set
of jobs is the weighted sum of the jobs’ tardinesses.

1. India has 100 tribunals to handle the 1.9 million people who have been removed from the citizenship
rolls. Each of those people must appear before a tribunal to plead their case. The tribunals run
for 16 hours a day. If on average 5,000 people are handled per day, what is the average number of
hours it takes for a tribunal to hear a single case?

This is a Little’s Law or rate times time type problem based on current events.

If, instead, the average waiting time until a person pleads their case is 1 year, what is the average
number of hours it takes for a tribunal to hear a single case?

2. Select a real system that has random arrivals and departures. Examples:

• Cars coming up to a stop sign: the arrival occurs when the car gets in line behind the stop
sign; the departure occurs when the car has passed the stop sign.
• People getting in line at Starbucks or another fast food service. The arrival occurs when the
person gets in line; the departure occurs when the person leaves the counter with their drink
or food.
• People getting in a cashier’s line in a cafeteria, or store.
• People getting in line to enter a stadium.

(a) Observe the system for 10 minutes. Record the initial number in the system, arrival times and
departure times.
(b) Use the formula from Dai’s book and class to calculate wait times.
(c) Try to validate Little’s Law based on your data. If you cannot, explain why not.
(d) Turn in your raw data and your computed values. Comment on what you learned from this
activity.

As I said in class, I permit, nay, encourage, you to do this problem in pairs. Don’t wait until
Monday to do this.

3. There are 6 jobs indexed by j = 1, 2, . . . , 6.

1
 Job # Process Time Release Time Deadline Time Weight
1 20 0 50 2
2 40 0 65 3
3 10 30 40 1
4 50 30 100 10
5 10 30 100 3
6 60 0 200 8
For each of the following objectives, find the objective value if the jobs are run on a single machine
in the order 1,2,3,4,5,6.

(a) makespan 190

4. For the jobs in Problem 3, set all release times to zero and ignore deadlines. For each of the following
objectives, find a schedule that minimizes the objective and compute its value:

(a) makespan (this is trivial)

5. Repeat problem 4 for two identical machines that run in parallel. That is, each job must be run on
one of the two machines. Caution: makespan is no longer trivial.

6. Customers arrive at a checkout counter at times 1,4,5,10,20,22,23,28,29,35. The time it takes for
their checkout are the amounts 4,4,4,4,3,3,5,1,5. Compute for each customer the wait time from
when the customer arrives until the customer begins to get served. Use the formula given in class
wi+1 = [wi + pi − (ai+1 − a(i))]+ . The same formula is given in Dai&Park’s text.

7. Four jobs have uniformly distributed processing times, independently distributed on [10, 15],[15, 20],[10, 20],
and [15, 25] respectively. There are three identical processors. Whenever a processor completes a
job, it gets the next job on the schedule.
1
(a) Calculate the expected makespan for the schedule 1, 2, 3, 4. 37 12
1
(b) Calculate the expected flowtime for the schedule 1, 2, 3, 4. 82 12
(c) Calculate the expected makespan for the schedule 4, 1, 2, 3.
(d) Calculate the expected flowtime for the schedule 4, 1, 2, 3.
(e) Which schedule is better for flowtime? for makespan?

2
8. Four jobs have exponentially distributed processing times pi ∼ exp(1) : i = 1, . . . , 4. Job 5 has ex-
ponentially distributed processing time p5 ∼ exp( 14 ). There are two identical processors. Whenever
a processor completes a job, it gets the next job on the schedule.

(a) Calculate the expected makespan for the schedule 1, 2, 3, 4, 5. Remember job 5 on average
takes 4 times as long as job 1. I got 1582
195 .
(b) Calculate the expected flowtime for the schedule 1, 2, 3, 4, 5. Hint: You can extract this from
your makespan analysis – every time a job is completed, record its time. However, there is a
much simpler way to solve this problem. It relies on your best friend, linearity of expectation.

(c) Calculate the expected makespan for the schedule 5, 1, 2, 3, 4.