An Adaptive QoS Group Guided Grid Scheduling Algorithm with Task Replicas

To balance resource loads and minimize makespan are two vital goals in grid scheduling. However, it comes to be difficult for the dynamicity of grid resources, especially with meeting the QoS requirements of tasks considered. In this paper we propose a scheduling algorithm called the QoS group guided grid scheduling algorithm with task replicas (QGTR). The proposed algorithm makes scheduling decision that bases on recent QoS status feedback of resources, dividing tasks into groups according to the feedback QoS status of resources and then scheduling tasks in different groups to resources accordingly. In addition task replica is adopted to improve the resource utilization and gain a better schedule result. The simulation results show QGTR can effectively reduce the makespan and enhance the resource utilization in addition to meeting the QoS requirement of tasks with its best efforts.