Homework #1 solutions / IEOR4405 1

3.1. (a) By Theorem 3.1.1

P in the textbook, the WSPT (weighted shortest processing time) rule is
optimal for 1|| wj Cj . In order to apply this rule, we have to calculate the values of wj /pj .
These values are given below:
Job 1 2 3 4 5 6 7
wj /pj 0 3 2 1.6 2 2.2125 1.7778
As a result,
P both the sequences 2-6-3-5-7-4-1 and 2-6-5-3-7-4-1 are optimal with objective
value wj Cj = 1610.
(b) If p2 is increased from 6 to 7, the ratio w2 /p2 decreases from 3 to 2.57. Since w2 /p2 is still
the largest among all the ratios, this change will has no effect on the optimal sequences.
P
(c) The new objective value under (b) is wj Cj = 1689.
3.11 Let us assume that S is an optimal sequence and S is not WSPT. Then there exist in S at least
ω
two adjacent jobs i and j such that pjj < ωpii and job j is scheduled before job i. Given that by
assumption r is sufficiently close to zero, then it follows that ex ≈ 1 + x. Let us denote as S 0
the schedule in which job i is scheduled before job j and all the other jobs are scheduled in the
same order than in S. The only terms that differ in the objective value of schedule S and S 0 are
the corresponding to job i and job j. Let t be the time in which job j is scheduled in S. We will
argue that the sum of those two terms in schedule S 0 :
   
ωi · 1 − er·(t+pi ) + ωj · 1 − er·(t+pi +pj ) (1)

is strictly less than the contribution of the corresponding terms in S, which will contradict
optimality of S. Using ex ≈ 1 + x in equation 1 it follows that:

   
ωi · 1 − er·(t+pi ) + ωj · 1 − er·(t+pi +pj ) ≈ ωi · (r · (t + pi )) + ωj · (r · (t + pi + pj ))
= ωi · r · (t + pi ) + ωj · r · (t + pi + pj )
< ωi · r · (t + pi ) + ωj · r · (t + pj ) + ωi · pj
= ωi · r · (t + pi + pj ) + ωj · r · (t + pj )
   
≈ ωi · 1 − er·(t+pi +pj ) + ωj · 1 − er·(t+pj )

Which is the desired result.

1. In class we argued that the EDD rule finds an optimal schedule for 1||Lmax . Suppose jobs have
release dates so that job j cannot be started before time rj . This problem is denoted 1|rj |Lmax .
Does that EDD rule still minimize maximum lateness? Prove or find a counterexample.

EDD doesnt minimize maximum lateness in 1|rj |Lmax . Let us consider an instance with two
jobs. Job 1 has released time 0, deadline 5 and workload 3. Job 2 has released time 1, deadline
3 and workload 2. EDD will process first job 1 and then job 2. Having a maximum lateness of
max{3 − 5, 5 − 3} = 2. If job 2 is processed before job 1 then the resulting maximum lateness is
max{3 − 3, 6 − 5} = 1.
2.3 (a) The optimal schedule has ΣUj = 2. The following schedules are optimal {1, 3, 2, 4}, {1, 3, 4, 2},
{3, 4, 1, 2}, {3, 4, 2, 1}, {2, 3, 1, 4}, {2, 3, 4, 1}, {2, 4, 3, 1}, {2, 4, 1, 3}. Direct computation re-
veals that all the above schedules have ΣUj = 2.

This study source was downloaded by 100000831952622 from CourseHero.com on 02-06-2022 02:06:58 GMT -06:00

https://www.coursehero.com/file/7267737/sol1-2012/
 Homework #1 solutions / IEOR4405 2

(b) The optimal schedule is discussed in Algorithm 3.3.1 in chapter 3.

This study source was downloaded by 100000831952622 from CourseHero.com on 02-06-2022 02:06:58 GMT -06:00

https://www.coursehero.com/file/7267737/sol1-2012/
Powered by TCPDF (www.tcpdf.org)