A Hybrid GA-based Scheduling Algorithm for Heterogeneous Computing Environments

We design a hybrid algorithm to schedule the execution of a group of dependent tasks for heterogeneous computing environments. The algorithm consists of two elements: a genetic algorithm (GA) to map tasks to processors, and a heuristic-based approach to assign the execution order of tasks. This algorithm takes advantage of both the exploration power of GA and the heuristics embedded in the scheduling problem, so it can effectively reduce the search space while not sacrificing the search quality. The experiments show that this algorithm performs consistently better than heterogeneous earliest-finish-time (HEFT) without incurring much computational cost. Multiple runs of the algorithm can further improve the search result.