Authors: Martin Niemeier Andreas Wiese
Publish Date: 2013/09/10
Volume: 71, Issue: 4, Pages: 837-858
Abstract
We address a scheduling problem that arises in highly parallelized environments like modern multicore CPU/GPU computer architectures where simultaneously active jobs share a common limited resource eg memory cache The scheduler must ensure that the demand for the common resource never exceeds the available capacity This introduces an orthogonal constraint to the classical minimum makespan scheduling problem Such a constraint also arises in other contexts where a common resource is shared across machinesWe study the nonpreemptive case of this problem and present a 2+ϵapproximation algorithm which relies on the interplay of several classical and modern techniques in scheduling like grouping jobclassification and the use of configurationLPs This improves upon previous bound of 3 that can be obtained by list scheduling approaches and gets close to the 3/2−ϵ inapproximability bound If the number of machines or the number of different resource requirements are bounded by a constant we obtain a polynomial time approximation scheme
Keywords: