一种混合并行遗传算法在网格资源分配中的研究  

A Kind of Research on Hybrid Parallel Genetic Algorithm in Grid Resource Distribution

作　　者：荣巧梅[1] 李兰[1] 王怀彩[1] 

机构地区：[1]青岛理工大学计算机工程学院,山东青岛266033

出　　处：《电脑开发与应用》2009年第4期45-47,61,共4页Computer Development & Applications

摘　　要：在近年的资源选择算法研究中,有几种较为常见的算法。考虑到算法的性能和在网格领域中使用的频度以及实现等因素,目前研究集中在遗传选择算法,禁忌搜索算法和蚂蚁算法。首先讨论了这三种算法,并且在此基础上提出了一种较好的算法——混合并行选择算法,其基本思想是首先通过网格资源选择框架,采用静态预测的方法和贪心算法来实现与应用无关的资源预选择,然后用粗粒度并行遗传算法生成资源集合中的初始信息素分布,再利用蚂蚁算法求出全局最优解。实验表明混合并行遗传算法比普通算法在同等或更少的迭代次数就能获得更优的解。In the research of grid resource selecting algorithms, there are some algorithms which are often used. Taking into account the performance of the algorithms and the used frequency in the field of grid as well as factors, the research is currently focused on the genetic algorithm , tabu search algorithm and ant algorithm. The article discussed the three algorithms at first , and on this basis, a better algorithm - the parallel hybrid algorithm is brought up, its basic idea is first of all using the grid selecting framework, static and predictable method and greedy algorithm to achieve the the pre-selection of resources unrelated to application. And then use the coarse-grained parallel genetic algorithm to generate the initial distribution of pheromones in collection of resources, at last use ant algorithm to derive global optimum. Experiments have showed that the parallel hybrid genetic algorithm than the usual algorithm will be able to obtain better solutions at the less number of iterations.

关 键 词：算法 资源分配 最优解 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]