Title: Scheduling real-time fork-join structures in cluster computing

Authors: Reda Ammar, Abdulrahman Alhamdan

Addresses: U 2155, Department of Computer Science and Computer Engineering, University of Connecticut, 371 Fairfield Rd., 1155 Unit, Storrs, CT 06269 2155, USA. ' U 2155, Department of Computer Science and Computer Engineering, University of Connecticut, 371 Fairfield Rd., 1155 Unit, Storrs, CT 06269 2155, USA.

Abstract: Scheduling a large number of high performance computing applications on cluster computing environment is a complex task especially in real-time systems. Efficient scheduling strategies are critically important for achieving a good performance. A cluster scheduler should adequately manage the cluster resources. In this paper, we present an algorithm for scheduling real-time applications represented as parallel task graphs on a cluster. We utilise the available processing power on each processor to maximise the degree of parallelism while satisfying the required tasks| deadlines. The algorithm reduces the communication cost among tasks and the possibility of processing power fragmentation.

Keywords: cluster scheduling; cluster computing; real-time systems; fork-join structures; high performance computing; parallel task graphs; control flow graphs.

DOI: 10.1504/IJHPCN.2005.008568

International Journal of High Performance Computing and Networking, 2005 Vol.3 No.4, pp.262 - 271

Published online: 30 Dec 2005 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article