A Study On The Market-mechanism Based Scheduling Model For The Spatial Information Grid

How to finish computation tasks by using the grid resource effectively is one of the key points of research on grid system. In this thesis, a market-mechanism based scheduling model(MBSM) was proposed, which was adapted to the characteristics of the computation tasks of spatial information grid(SIG) and the using policy of the SIG resource. The core idea of this model is: a)Partitioning the job submitted by user into tasks, the tasks form a workflow by ordering the dependence relationships of the tasks, and several independent parallel tasks set can be composed based on a workflow graph; b)Then every tasks was scheduled to appropriate resources, the scheduling decision was based on the information provided by the resource information service and the resource monitoring service; c)The work load of the resources was reflected by the price of the resources, the goal of optimized scheduling is to make the deadline of the tasks shortest and the budget of finishing tasks as lower as possible, and the workload of resources can be balanced by using the price mechanism.Firstly, the characteristics of the SIG resources and tasks and the factors that affecting the cost of finishing tasks were analyzed, and the difficulties of SIG scheduling was founded. Based on the domestic and foreign research work, a market-mechanism based scheduling model was proposed.Secondly, the process of MBSM was presented. A workflow graph was proposed to describe the process of job partitioning and dependence relationship, and the method of repartitioning of workflow graph was discussed. The information that can be provided by resource information service and resource monitoring service were analyzed.Thirdly, the implementation of MBSM was discussed. The structure of MBSM, the function of modules and a optimized scheduling algorithm were presented. The data model of main modules and main scheduling flow were presented.Lastly, a simulation experiment was implemented by using the GridSim simulation toolkit, the results of the experiment was analyzed and summarized.