Report ID
1994-12
Report Authors
T. Yang and A. Gerasoulis
Report Date
Abstract
We present a low complexity heuristic named the Dominant Sequence Clusteringalgorithm (DSC) for scheduling parallel tasks on an unbounded number ofcompletely connected processors. The performance of DSC is comparable or evenbetter on average than other higher complexity algorithms. We assume no taskduplication and nonzero communication overhead between processors. Finding theoptimum solution for arbitrary directed acyclic task graphs (DAGs) isNP-complete. DSC finds optimal schedules for special classes of DAGs such asfork, join, coarse grain trees and some fine grain trees. It guarantees aperformance within a factor of two of the optimum for general coarse grainDAGs. We compare DSC with three higher complexity general schedulingalgorithms: the ETF by Hwang, Chow, Anger and Lee, Sarkar\'s clusteringalgorithm, and the MD by Wu and Gajski. We also give a sample of importantpractical applications where DSC has been found useful.Note: To appear in IEEE Trans. on Parallel and Distributed Systems.
Document
1994-12.ps631.01 KB