The Design And Implementation Of Scheduling System For Service Grid

Service Grid which is the outcome of combing Web service and Grid technologies represents the developing trend of computing model. As a kernel technology of Service Grid, the scheduling system aims to develop the utilization ration of resources and improve the performance of grid application. The research of scheduling system has been the hot issue and problem in Grid research.The thesis first introduces the background and related technologies of service grid. Then, according to the disadvantages of current popular scheduling systems and the nature of grid which is dynamic, uncertain and heterogeneous, the thesis designs and implements a scheduling system for Service Grid. The system adopts market-oriented scheduling policy to implement efficient scheduling for grid resources and provides a series of functions, such as optimizing process, information service, scheduling and fault-tolerant mechanisms. Therefore, a dynamic service resource selection model based on historical information feedback (DRHF) is proposed. The model adjusts predicting model automatically by using historical information of services execution, and then it selects the best one as an executing node to run the service. The experimental results show this method can effectively increase the reliability and efficiency of service execution, and balance the system workload. Moreover, the thesis introduces the idea of DRHF into job scheduling and proposes an adaptable dynamic job scheduling approach based on historical information (ASHI). This approach adjusts the predicting model automatically by using the latest jobs execution historical information and then selects the appropriate resource to execute the job considering dynamic and real-time factors of the Grid. The experimental results demonstrate that this method not only can schedule the jobs effectively and timely, but also can improve the throughput and load balance of the Grid. At last, the thesis analyses the reasons of failure in the process of scheduling and proposes a fault-tolerant schema under the Service Grid. The schema puts both exception handling strategies of task-level and workflow-level together, and as to satisfy users'requirements for handling failures flexibly, the schema provides two kinds of exception strategies: automatic process and user designation.There are two contributions in this research. On one hand, it implements a dynamic, flexible and real-time scheduling system based upon the characteristics of grid resources and experimental results indicates it achieves effectively. On the other hand, the thesis proposes a flexible fault-tolerant mechanism which combines the task-level and workflow-level techniques of failure handling. The work in this paper has certain meanings for reference to related study on the scheduling mechanism in the Service Grid environments.