Parallel machine scheduling problem based on improved Hierarchic Genetic Algorithm

This paper studies the problem of minimizing the maximum completion time of parallel machine scheduling based on MapReduce model. Each work piece contains Map and Reduce processes. Map process can be divided into parallel processing on multiple machines, while Reduce process is indivisible and can only be processed after the completion of Map process. The improved Hierarchic Genetic Algorithm is designed. The Adaptive Genetic Algorithm is used in the low level algorithm, and the Simulated Annealing is used in the high level algorithm. By comparing the merits and demerits of different populations in the low level, individuals of the high level population are selected to realize the information exchange among the populations. Simulation results show that, compared with the Simulated Annealing and the Water Wave optimization, the improved Hierarchic Genetic Algorithm designed in this paper has better convergence.