Grid Resources Scheduling Based On QoS

There have been more and more grid users with different Quality of Service requirements in the grid resource management. Only improving availability of systems and resources or considering schedulability of resources directly, most scheduling policies now rarely consider scheduling from user's view. Optimization methods merely take single cost or time into account, not optimization among multiple targets and multiple tasks in an application. Consequently, they cannot better improve availability of systems and the success rate of the application. All of the above makes the resource management and scheduling a complex problem and challenging.Motivated by this fact, the dissertation studies how to satisfy the users with QoS requirements and improve the reliability of the resource scheduling. The contributions of this dissertation are as follows:Firstly, grid Quality of Service's fundamental conception is introduced and grid Quality of Service is classified. The current grid resource scheduling policies are studied, including static scheduling algorithms, dynamic scheduling algorithms and Deadline and Budget Constrained scheduling algorithms.Secondly, analyzing and studying characteristics of traditional resource scheduling policies, this dissertation proposes resource scheduling with user's Quality of Service preference. It solves the problem of user's Quality of Service preference with weighting method effectively. For this reason, the user's actions can be described more accurately.Finally, based on user's preference and Quality of Service parameter's characteristics, a resource scheduling algorithm implemented between multiple tasks with time and cost is proposed. The algorithm doesn't simply use time or cost of a single task as the constraint, but not beyond budgetary completion time and cost premise, and when a scheduling failure appears, the algorithm calculates the QoS parameter's fault-tolerant capability to determine whether to continue scheduling.This scheduling algorithm is simulated with GridSim toolkits. The result shows that the scheduling algorithm has better performance and can better meet the requirements of users and improve the reliability of scheduling.