计算集群中一种基于任务运行时间的组合预测方案

发布时间：2018-07-24 18:10

【摘要】：针对现有单一预测策略不适用于所有异构任务的问题,提出一种基于本地任务与远程任务运行时间的组合预测方案(CPS)和预测精度保证(PAA)的概念。使用Grid Sim工具集来实现CPS,将PAA作为定量评价由某一特定预测策略提供的预测运行时间精度的标准。仿真实验表明:与本地任务预测策略如Last和滑动窗口中值(SM)相比,CPS的平均相对残差下降了1.58%、1.62%;与远程任务预测策略如平均运行时间(RM)和加权移动平均值(ES)相比,CPS的平均相对残差下降了1.02%、2.9%。因此,PAA能从综合策略所提供的结果中选择接近最优值的预测,CPS增强了计算环境中本地任务和远程任务运行时间的PAA。
[Abstract]:In order to solve the problem that the single prediction strategy is not suitable for all heterogeneous tasks, this paper presents the concept of (CPS) and (PAA), which are based on the running time of local task and remote task. The Grid Sim toolset is used to realize PAA, which is used as a standard to quantitatively evaluate the precision of prediction running time provided by a particular prediction strategy. The simulation results show that compared with the local task prediction strategy such as Last and sliding window median (SM), the average relative residuals are reduced by 1.58 and 1.62, and compared with the remote task prediction strategies such as average running time (RM) and weighted moving average (ES). The average relative residuals decreased by 1.02% and 2.9%. Therefore, PAA can select a prediction close to the optimal value from the results provided by the synthesis strategy. CPS enhances the PAAs of local and remote tasks running time in the computing environment.
【作者单位】：衡阳师范学院计算机科学系;湖南大学信息科学与工程学院;
【基金】：国家自然科学基金资助项目(61370095,61370098,61070057,90715029) 湖南省教育厅科学研究项目(13C074) 衡阳市科技发展计划项目(2011KJ22) 湖南省教育科学“十二五”规划课题(XJK014CGD006)
【分类号】：TP38

【参考文献】