《计算机技术与发展》  2015年09期 169-173   出版日期：2015-09-10   ISSN:1006-6977   CN:61-1281/TN

---

 网格依赖任务重调度资源选择算法研究


 Selecting a stable resource for grid-dependent task is the key issues to reduce the frequency of tasks rescheduling. Aiming at this problem,present a marking method that count the number of resources to perform tasks successfully. Based on this tag value classify resource dynamically. Then put forward the improved method based on the tax mechanism successful index rescheduling method. This method uses different tax rates for different tag value range,segmenting the tag values of resources according to the size,allocating the rel-atively stable resources for task according to the current tag values. For the tag value in different segments,when resources are in the suc-cessful completion of a task or fails,the tag value will be different for increasing or decreasing degree of tag values. Experiment results show that the method can make the resources fully utilized and improve load imbalance,effectively improving the accuracy of current sta-bility judgment. The possibility of rescheduling trigger is reduced.