IPDPS '05 Proceedings of the 19th IEEE International Parallel and Distributed Processing Symposium (IPDPS'05) - Workshop 1 - Volume 02
Journal of Parallel and Distributed Computing
Comparison of genetic representation schemes for scheduling soft real-time parallel applications
Proceedings of the 8th annual conference on Genetic and evolutionary computation
Stochastic DAG scheduling using a Monte Carlo approach
Journal of Parallel and Distributed Computing
Hi-index | 0.00 |
Abstract: The fact that the scheduling problem is NP-complete has motivated the development of many heuristic scheduling algorithms. These heuristic algorithms often neglect the stochastic nature of tasks' execution times. Contrary to existing heuristics, in this study, tasks' execution times are treated as random variables and the stochastic scheduling problem is formulated accordingly. Using this formulation, it is theoretically shown that current deterministic scheduling algorithms may perform poorly in a real computing environment. In order to support the theoretical foundations, a genetic algorithm based scheduling algorithm is devised to make scheduling decisions either stochastically or deterministically by changing only the fitness function of chromosomes. The simulation studies conducted show that deploying a stochastic scheduling algorithm instead of a deterministic one can improve the performance of meta-tasks in a heterogeneous distributed computing system.