Grid Resources Scheduling Based On Multiple QoS

In the grid enviroment, there have been more and more grid users with different Quality of Service requirements. Most of traditional resource scheduling algorithms aim at minimizing the makespan and improving system performance, not considering users'QoS requirements, so some tasks are scheduled into resources which can't meet their requirements. Therefore, resource scheduling algorithms need to consider different QoS requirements of users to meet the needs of users.Motivated by this fact, this paper investigates the grid resource scheduling algorithms with multiple QoS requirements, in order to make the greatest efforts to meet the multiple QoS requirements of users. According to the level of user's QoS requirements, the task set is splited into three disjoint sets, namely tasks with hard QoS requirements in the timeliness, tasks with hard QoS requirements except timeliness, and tasks without hard QoS requirements. Firstly, tasks with hard QoS requirements in the timeliness are scheduled, tasks are sorted in ascending order according to their average slack values ensuring that urgent tasks are scheduled first. Each task select resources in accordance with its minimum requirements, then select the one that can attain the maximum value of the overall utility function. If this task attain the same value of the two resources,the one with the earlier completion time will be selected. Secondly, tasks with hard QoS requirements except timeliness are scheduled. Tasks without hard QoS requirements are scheduled lastly. For the rest two task sets,each task selects resources in accordance with its minimum requirements, then the one that can attain the max value of the overall utility function is written. Finally, the task with the maximum value is selected. If the two tasks have the same value,the task with the earlier completion time will be selected.This scheduling algorithm is simulated using GridSim tookits. The results show that the scheduling algorithm has better performance and can better meet the requirements of users.