Abstract
A fuzzy clustering based scheduling heuristic (FCBSH) is presented to solve the scheduling problem in large-scale heterogeneous computing environments. A group of features is defined to describe the synthetic performance of processing cells and distinguish their heterogeneousness in large-scale heterogeneous computing environments. This paper mainly focuses on the detail performance analysis of FCBSH. A task graph generator and a target system generator are presented to randomly produce task graphs and target systems according to custom-made parameters. Through experiments, this paper finally reduces features from 7 features to 3 notable features, which not only can effectively partition the target system with fuzzy clustering, but also greatly reduce the processing time of the clustering process. Finally, the FCBSH algorithm is compared with some algorithms in performance of make-span, runtime and proportion of producing better scheduling results.