Research On Group Of Task And Secondary Distribution And Load Balancing On Scheduling Algorithms In Grid

Grid as a distributed heterogeneous computing platform designed to share resources within the scope of the Internet and collaborative computing. Research on Grid Task Scheduling Algorithm of Grid technology as the core problem is the grid resource management an essential part. In this paper, the characteristics of resources and the status of research in-depth analysis grid environment task scheduling algorithm for grid task scheduling algorithms existing long span of time, unbalanced load and low resource utilization problems for grid task scheduling algorithm analysis and research, and ultimately get Makespan, load balancing and resource utilization and other performance relative to existing methods have improved algorithm to enhance higher.Firstly,the Max-Min grid resource scheduling algorithm optimization. Max-Min scheduling algorithm thinking is simple, easy to implement and achieved good results in the actual scheduling grid applications. However, due to Max-Min scheduling algorithm always takes precedence over a long task scheduling, resource allocation and strong computing power to the long task. When a larger number of tasks under centralized long task of case, since priority scheduling long task, and therefore may increase Makespan system. Max-Min algorithm for load imbalance, lack of resources and low efficiency, the use of standard deviation calculation task set task group, the task set into the length of the task force, and finally the use of the secondary distribution of ideas, assign tasks to the heavy load on light-load resources to perform scheduling strategy proposed GTSDSecondly,combined with the distribution of tasks set tasks, improve Seletion grid resource scheduling algorithm. Because the proportion of the size of the change in the length of the mandate of the task set, will affect the performance of the scheduling algorithm, when the long task than short mission Min-Min algorithm to achieve better performance, on the contrary, Max-Min algorithm has better performance. To select the algorithm with the best performance, make Double-SD scheduling algorithm. Using the prediction mechanism, depending on the task to predict the size distribution of the standard deviation of the task set, and then choose the most suitable scheduling algorithm according to the task set task distribution, at the end of selection algorithm, for Min-Min and Max-Min algorithm algorithms exist load imbalance, effective resource utilization is low and the system to complete a long time, the idea of using the second scheduling, load balancing fast, shorten time to complete.Finally, the use of simulation models Braun, who made reference to GTSD and Double-SD scheduling algorithm simulation, and Balance Max-Min scheduling algorithm and other comparative analysis, to verify the effectiveness of the algorithm.