Parallel Greedy Genetic Algorithm for Job Scheduling in Cluster Enviornments

Recently, many scientific researchers and applications work on large amounts of data or use high performance computing resources. A high performance cluster is developed to handle massively parallel processes. To manage the resources for dynamic requests with optimal usage, we have to maximize the utilization rate of clusters. In this paper we provide a parallel genetic algorithm to schedule the jobs for different classes of clusters. The greedy approach is used to create an initial population for the genetic algorithm. We applied the master/slave method in parallelism to manage the schedulers and improve the performance of the main scheduler. Analyzing the complexity of the algorithm shows that it can be more efficient than similar algorithms.