On-Line Scheduling of Parallelizable Jobs

Christophe Rapine and Isaac Scherson and Denis Trystram

Abstract
We consider the problem of efficiently executing a set of parallel jobs on a parallel machine by effectively scheduling the jobs on the computer's resources. This problem is one of optimization of resource utilization by parallel computing programs and/or the management of multi-users requests on a distributed system. An instance of the problem consists in a set of n independent jobs and a parallel machine with m identical processors. We assume that each job is parallelizable and can be executed on any number of processors. Various on-line scheduling strategies of time/space sharing are presented here. The goal is to assign jobs to processors in space and time such that the total execution time is optimized.
Contact
Denis Trystram
LMC-IMAG,Domaine Universitaire, BP 53,F-38041 Grenoble Cedex 9,FRANCE
Denis.Trystram@imag.fr