The Study Of The Adaptive Scheduling Algorithm In The Non-dedicated Grid Environment

The high speed networks develop very fast, which makes it possible to integrate the distributed and heterogeneous computing resources into Computational Grid. It provides an ideal platform to solve large-scale computing problems in scientific and engineering area. Because the resources in Grid are distributed, heterogeneous and dynamic, how to schedule tasks in Grid to meet users'requirements is a challenging problem.With the progress of the Grid, the dedicated property for the service resources has became an important factor needed to be considered in the procedure of scheduling. In this paper, the dedicated resources are the service resources provided by the parallel computing system and they could be deployed for free using. The mid dedicated resources are defined as being provided for free using only in its leisure time. That means they will process the paid tasks first, and only after all paid tasks have been disposed, the leisure resources can be used. The non-dedicated resources are defined as the service resources that could be used only by the paid tasks. How to synthesize the resources'characteristic of dedicated, mid dedicated and non-dedicated, and match all kinds of the tasks in order to improve the performance of the whole system is one of the most important factors to be considered.Firstly, this thesis presents the definition and research progress of risk assessment at home and abroad. Secondly, we research the architecture of Grid and many classicl dynamic scheduling algrithms, such as Min-Min, Max-Min and suffrage etc. Based on the analysis of the GHS, the scheduling algorithm of it, Min-Min, is a simple algorithm which runs fast and delivers the satisfactory performance but it unable to balance the load well since it usually schedules the smallest tasks first. Thus, we propose to bring Max-Min which has good banlance to resources into Min-Min, consequently, a new adaptive scheduling algorithm, an adaptive based on Min-Min and Max-Min, A-MM, has been presented. Finally, we propose to abstract the merit of A-MM and bring in the non-dedicated property of the resources, and A-MM based on the non-dedicated constraint grid, namely the adaptive Min-Min and Max-Min based on non-dedicated constraint grid (NDA-MM) algorithm, was proposed. Collecting datas from a quite few of experiements, we had analysed them and concluded that the new algorithm not only retains the advantage of the Min-Min algorithm but also achieves the good load balance.