基于混合蚁群遗传算法的Hadoop集群作业调度  

Hadoop Clusters of Job Scheduling Based on Mixed Ant Algorithm Genetic Algorithm

作　　者：楼涛[1] 杜文才[1,2] 钟杰卓[1] 

机构地区：[1]海南大学信息科学技术学院,海南海口570228 [2]澳门城市大学

出　　处：《海南大学学报（自然科学版）》2015年第4期340-346,共7页Natural Science Journal of Hainan University

基　　金：国家自然科学基金(61162010);海南大学青年基金(Qnjj1186)

摘　　要：提出了一种基于蚁群与遗传算法融合的自适应作业调度机制,将遗传算法全局收敛、快速搜索的优点与蚁群算法正反馈、高求精率的优势相结合,以变异策略来加快局部寻优,提高收敛速度.实验结果表明本文算法可快速找到最适合当前作业的节点,有效提高Hadoop集群作业调度的效率.In our report,a kind of self-adaptive job scheduling mechanism based on ant algorithm and genetic algorithm was proposed,which integrated the advantages of genetic algorithm,the global convergence and fast search,with the characteristics of ant algorithm,the positive feedback and efficient refinement,and the mutation strategy was performed to accelerate the speed of local optimization and convergence. The results indicated that the algorithm can obtain the most suitable nodes for current jobs and effectively improve the efficiency of job scheduling on Hadoop clusters.

关 键 词：蚁群算法 遗传算法 HADOOP集群 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]