Scheduling Parallel Tasks with Intra-communication Overhead in a Grid Computing Environment

With the improvements in wide-area network performance and powerful computers, it is possible to integrate a large number of distributed machines belonging to different organizations as a single system, for example, a grid computing environment. A grid computing environment involves cooperation and sharing resources among distributed machines. Users may dispatch their tasks to remote computing resources instead of just computing locally. Hence, task scheduling is an important issue in a grid computing environment. If a task is processed in parallel in a parallel machine, intracommunication overhead is inevitable and is an important factor affecting the task’s processing time. Thus, instead of a free communication assumption, this paper takes intra-communication overhead into account. According to the models of parallel task and intra-communication overhead, the problem of scheduling independent parallel tasks with intra-communication overhead in a grid computing environment is investigated. In this paper, a heuristic algorithm for this problem is proposed, and the performance bounds of the heuristic algorithm for scheduling parallel tasks in the environment with message passing machines and in the environment with shared memory machines are derived as 9/2 and 5/2 respectively. Although this problem is mainly about assigning tasks to a grid computing environment, however, by proper transformation it can be applied to the workforce assignment problem.