Algorithm for scheduling independent jobs on partitionable hypercubes

Abstract

The authors consider the problem of nonpreemptive scheduling of w independent tasks on an n processor partitionable hypercube system. Each task can be executed on a subcube of any dimension, with a smaller execution time on larger subcubes. The schedule must determine the subcube to be allocated to each task, with the objective of minimizing the overall finishing time. The authors present a polynomial time approximation algorithm which generates a schedule whose finishing time is within twice that of the optimal schedule.<>