Você está na página 1de 2

Exercise 5 : EDD rule solves the problem 1| r j | Lmax for the

case that the release dates and due dates are similarly ordered.
We will show the optimality of EDD rule to the problem 1| r j | Lmax
r n a nd d 1
d n by contradiction using
under the condition r 1
pairwise exchange argument.
Let

be an optimal schedule to the problem in which the jobs are not


processed under EDD rule. Therefore, there are at least two adjacent jobs such
d j where r k
that job k is processed before job j even though d k

rj .

.......

.......

{Li ( )}, L j ( ), Lk ( )}
Lmax ( )= max { max
i k , j
L j ( )=r k +s + pk + p j - d j
Lk ( )=r k + s+ p k - d k

Since

L j ( )

Lk ( )

pj +

where

( k d j)

equal to Lk ( ) .
Therefore, there are two cases which are

>0 ,

Lmax ( ) =L j ( )

Lmax ( ) can not be

and

Lmax ( ) =max {Li ( ) } .


ik ,j
Case 1 :

Lma x ( )=L j ( )

{Li ( )} , Lk ( ) (*)
+ p j - d j > max
i k , j
We now perform the exchange between k and j without changing the
completion times/job processing orders of the others to create a new
schedule( ' re
s
garding EDD rule.
dj ,
Since d k
we force the
machine to start
j
k
.......
.......
processing job j
instead of job k as
follows:
Lmax ( ) =r k +s + pk

':

.....

.....

Since the completion times of the others will remain the same,
Li ( )=Li ( ' )
i j ,k.
{Li ( )}, r k + s+ p jd j ,r k + s+ p j + pk d k }
Lmax ( ')= max { max
ik , j

Lmax ( ' ) may be equal max { Li( )},


i k , j

r k + s+ p jd j or r k + s+ p j+ p k d k

Let Lmax ( ' ) =max {Li ( ) }: Lmax ( )Lmax ( ' )=

1)

from (*).

i k , j

Let Lmax ( ' ) =r k + s+ p jd j

2)

d j (r k + s+ p jd j)

r k + s+ pk

: Lmax ( ) Lmax ( ')=

pk > 0

Let Lmax ( ' ) =r k + s+ p j+ p k d k

3)

Lmax ( ) - max {Li ( )}


ik, j

: Lmax ( ) Lmax ( ')=

r k + s+ pk

>0

pj +

pj -

d j (r k + s+ p j + pk d k ) = d k d j > 0 from the assumption .


Lmax ( ) Lmax ( ' ) >0 contradicts the optimality of schedule .
Case 1 : Lmax ( ) =max {Li ( )}
ik ,j

Lmax ( ) =max {Li ( ) }


ik ,j

> Lk ( ) ,

Lj ( )

(**)

{Li ( )}, r k + s+ p jd j ,r k + s+ p j + pk d k }
Lmax ( ')= max { max
ik , j

Lmax ( ' ) may be equal max { Li( )},


i k , j

r k + s+ p jd j or r k + s+ p j+ p k d k

1 L et Lmax ( ' )=max { Li ( )}: Lmax ( )Lmax ( ' )=


i k , j

2 Let L max ( ' )=r k + s +p jd j


max {Li ( )}(r k +s + p j d j )

3 Let Lmax ( ' ) =r k +s + p j + p k d k


max {Li ( )}(r k +s + p j + pk d k )
the assumption

i k , j

{Li ( )}
- max
i k , j

: Lmax ( ) Lmax ( ')=

ik, j

i k , j

max {Li ( )}

L j ( )r k +s + p j d j = pk >0.
: Lmax ( ) Lmax ( ')=
r
( k + s+ p j + pk d k )
d k d j > 0 from

L j ( )

Lmax ( ) Lmax ( ' ) 0 contradicts the unique optimality of schedule .

Você também pode gostar