摘要
在网格环境下,资源选择的策略是决定应用程序性能的关键因素之一。目前的资源选择一般采用贪心算法,而在资源选择中引入遗传算法是一种新的尝试,其目的是提高资源选择的效率。这种资源选择首先产生若干个网格环境中可能的配置(第一代),预测应用程序在每种配置下的表现。再以此为依据,应用多种交叉和变异策略,产生下一代的配置,直到得到满意的结果。对所选应用程序的两种不同规模的测试表明,在相同或更短的时间内,基于遗传算法的资源选择策略比贪心算法的资源选择策略得到更好的结果。
In environment of grid, the strategy of resource selection is one of the key fac tors determining performance of application.At present, greedy algorithm is usua lly used as the strategy of resource selection. While introducing genetic algori thm into the resource selection is a new attempt for efficiency promotion.In thi s resource selection, several available configurations(first generation) are gen erated at first, and performance of application is predicted in each configurati on. Then, according to the prediction, this generation will evolve to the next o ne by crossover and mutation in multiway, till the satisfy result is got.The exp erimentation of two scales of the same application present that, by the same or smaller time cost, the strategy based on genetic algorithm produces better resul t than the one based on greedy algorithm does.
出处
《计算机应用》
CSCD
北大核心
2003年第5期20-23,共4页
journal of Computer Applications
关键词
资源选择
遗传算法
网格计算
resouce selection
genetic algorthm
grid computing
