Task Migration Enabling Grid Workflow Application Rescheduling

摘要

This paper focuses on the task migration enabling grid workflow application rescheduling problem, presents a reduced task graph model, and implements a performance oriented rescheduling algorithm based on immune genetic algorithm. The experiment shows that, compared with Adaptive Heterogeneous Earliest Finish Time static rescheduling algorithm and the classical dynamic Max-Min scheduling algorithm, the performance advantage of the proposed rescheduling algorithm is obvious, on the one hand because of the performance contribution of global optimization and task migration, and on the other hand because of the efficiency contribution of task graph reduction and immune genetic algorithm's convergent speed. It also shows that task migration improves grid application's adaptability of dynamics further.