Exam Sequencing and scheduling, 2P450,
January 23, 2013, 14:00-17:00
This test contains four questions on two pages, with 10 sub-items, for which a total of
50 points can be scored. Each good answer scores 5 points. The final grade is obtained
by dividing the amount of points by five. Questions may be answered in English and/or in
Dutch. Please explain your answers. You are NOT allowed to use the handouts distributed
in class/or on the web.
Question 1. Consider the problem 1|rj |Lmax of minimizing maximum lateness on a single
machine under release dates. In particular consider the instance given by
jobs j 1 2 3 4 5 6 7
pj 6 18 12 10 10 17 16
rj 0 0 0 14 25 25 50
dj 8 33 44 24 90 85 68
(a) Formulate and apply the preemptive EDD rule. Compute the optimal preemptive sched-
ule, and list the results in a table with the following entries:
jobs j 1 2 3 4 5 6 7
CjpEDD
dj 8 33 44 24 90 85 68
pEDD
Lj
(b) Prove optimality of your schedule by pointing out the correct critical job subset.
Question 2. Consider a problem with m uniform machines, with speeds s1 ≥ s2 ≥ . . . ≥
sm > 0. We have n tasks that may be preempted. The jobs have processing requirements pj
with p1 ≥ p2 ≥ . . . ≥ pn > 0, for j = 1, . . . , n. That means that if a machine i is to process
p
the whole job j, it will take sji time units. We are interested in minimizing the makespan.
(a) Argue that if, for instance I, a schedule of length ` exists, then, for any integer k,
0 < k < min{m, n}, there is also a schedule of length ` or less for the instance Ik ,
defined by considering only the k first machines, and the k first tasks. Use the result
OPT (I).
on Ik to express a lower bound LBk (I) on Cmax
(b) Argue that if a schedule of length ` exists, then
Pn
j=1 pj
` ≥ max{ Pm , max LBk }
i=1 si k
(c) For the following instance on 5 machines and 7 jobs compute the (highest) lower bound
and find a schedule of the corresponding length. Machine speeds are: s1 = 5, s2 = 4,
s3 = 3, s4 = 2 and s5 = 1; job requirements are p1 = 470, p2 = 390, p3 = 340, p4 = 160,
p5 = 60, p6 = 40 and p7 = 20.
1
Question 3. Consider the problem (P ): O5||Cmax of minimizing maximum completion time
on a set of five identical machines. In particular consider the instance I on 6 jobs where
process requirements are given by
jobs j 1 2 3 4 5 6
p1j 1 0 1 0 1 2
p2j 1 1 2 0 2 0
p3j 2 1 1 1 1 0
p4j 1 2 1 2 0 0
p5j 1 1 0 1 0 2
We further consider the problem (Q): O5|pmtn|Cmax of minimizing maximum completion
time on a set of five identical machines, where we allow operations to be preempted.
Let P (I 0 ) and Q(I 0 ) denote the optimal solution values of the two problems, defined on
the same instance I 0 .
(a) How do the values P (I 0 ) and Q(I 0 ) relate, for general instances I 0 ? Justify your claim.
(b) For the general case I 0 , and for the particular instance I defined above, give a non-trivial
lower bound on the solution values Q(I 0 ) and Q(I).
(c) Does there exist a solution of Q with a matching value? If so, compute such a schedule.
(d) If we would consider instances I” of the open shop problem defined on TWO machines,
for example on the instance given by
jobs j 1 2 3 4 5 6
p1j 4 1 3 1 3 2
p2j 2 4 2 3 1 2
what is then the relation between P (I”) and Q(I”), where P and Q refer to the non-
preemptive and the preemptive version, respectively? Substantiate your answer.
Question 4.
Consider the use of the LPT-rule for solving the problem of minimizing makespan on m
parallel identical machines processing n jobs. We observe the following
LEMMA: for any instance I of P ||Cmax : if there exists an optimal schedule with at most
two operations per machine, then LPT when applied on I yields a schedule that has optimal
length.
You do not have to prove this lemma!! Give the worst case ratio between the makespans
LPT (I)
Cmax OPT (I) in terms of m and prove that ratio making use of the above Lemma.
and Cmax
