Load Balanced, Efficient Scheduling With Parallel Job Submission in Computational Grids Using Parallel Particle Swarm Optimization

摘要

Management of resource and application scheduling in a highly distributed heterogeneous Grid environment is a complex and challenging task. Processing jobs at the grid resources in a fine grained form results in a low computation - communication ratio. This necessitates the dynamic assembly of fine grained jobs into groups of jobs before dispatching them to the resources. Recent advances in computer and network technologies have led to parallel optimization algorithms. Here a novel job grouping method using Parallel Particle Swarm Optimization (PPSO) is proposed to reduce the communication overhead, enhance the speed of completion of processes, improve resource utilization, and parallel efficiency. The proposed approach uses PPSO to group the jobs and to submit them in parallel to the grid resources. Trust based parallel job submission is also proposed to ensure security and improve on job submission time. The proposed approach has been implemented and tested by extending the features of GridSim, a simulation toolkit for grid environment.