多目标云工作流调度的协同进化多群体优化  

Coevolutionary multi-swarm optimization of multi-objective cloud workflow scheduling

作　　者：刘雨潇[1] 王毅[1] 袁磊[1] 吴钊[1] LIU Yu xiao;WANG Yi;YUAN Lei;WU Zhao(School of Mathematical and Computer Science, Hubei University of Arts and Science, Xiangyang 441053, China)

机构地区：[1]湖北文理学院数学与计算机科学学院,湖北襄阳441053

出　　处：《计算机工程与设计》2018年第5期1350-1357,共8页Computer Engineering and Design

基　　金：湖北省襄阳市科技计划基金项目(2015zd26)

摘　　要：为实现云工作流调度的多目标最优化,提出一种协同进化多群体优化调度算法。以执行跨度、代价和能耗同步最优化为目标,建立基于激素的协同进化多群体优化模型;通过多群体方式,使每个群体通过多目标粒子群优化寻找单目标非占优解;为避免局部最优,粒子进化中,引入激素激励机制和多群体竞争与协作机制,得到多目标最优解。通过仿真实验,与多目标调度算法MOHEFT和CMPSO作分析比较,结果表明,该算法在综合性能上实现了更好的Pareto最优解,具有更好的有效性和可行性。For implementing the multi-objective optimization of workflow scheduling in cloud,a coevolutionary multi-swarm optimization scheduling algorithm was proposed.The synchronization optimization of workflow scheduling execution makespan,cost and energy consumption was defined as the objective and an endocrine-based coevolutionary multi-swarm optimization model was established.Multi-swarm method was adopted and each swarm employed multi-objective particle swarm optimization to find out non-dominated solutions with on objective.To avoid falling into local optima,an endocrine-inspired mechanism was embedded in the particles’ evolution,and a competition and cooperation mechanism among swarms was designed,which obtained the multi-objective optimization solutions.Through simulation experiments,the performance using the proposed algorithm was compared with that of multi-objective scheduling algorithms such as MOHEFT and CMPSO.The results show that the proposed algorithm can implement better Pareto solution in overall performance and it has better feasibility and effectiveness.

关 键 词：工作流调度 多群体 协同进化 粒子群优化 Pareto边界 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]