| Abstract |
|---|
| Parallel job scheduling is the problem of how to run a workload of multiple parallel jobs in a single parallel machine. Dynamic means that the possibility of arbitrary arrival times for new jobs is allowed. The scheduler is responsible for finding the best scheduling allocation, both temporal and spatial, as a function of the existing workload. In this study jobs are assumed to be data parallel with large degrees of parallelism, and the machine is assumed to have an MIMD architecture. The dynamic parallel job scheduling is considered here as an optimization problem, and after defining some metrics a theoretical solution is proposed anda lower bound on its complexity is found. Simulation is then used to compare the proposed solution with already existing policies. |
| Contact |
| Fabricio Alves Barbosa da Silva Laboratoire ASIM,Couloir 65-55, 2eme etage,Universite Pierre et Marie Curie,4, Place Jussieu,75252 Paris Cedex 05,France, fabricio.silva@asim.lip6.fr |