王国豪,李庆华,刘安丰.多目标最优化云工作流调度进化遗传算法[J].计算机科学,2018,45(5):31-37
多目标最优化云工作流调度进化遗传算法
Evoluation Genetic Algorithm of Multi-objective Optimization Scheduling on Cloud Workflow
投稿时间:2017-07-18  修订日期:2017-10-03
DOI:10.11896/j.issn.1002-137X.2018.05.005
中文关键词:  云计算,遗传算法,工作流调度,多目标优化,适应度函数
英文关键词:Cloud computing,Genetic algorithm,Workflow scheduling,Multi-objective optimization,Fitness function
基金项目:本文受浙江省教育厅科研备案项目(Y201534160),浙江省公益性应用研究计划项目(2016C31G2260015)资助
作者单位E-mail
王国豪 丽水学院工学院 浙江 丽水323000 lsxywgh2011@163.com 
李庆华 丽水学院工学院 浙江 丽水323000  
刘安丰 中南大学信息科学与工程学院 长沙410083  
摘要点击次数: 391
全文下载次数: 689
中文摘要:
      为了实现云环境中科学工作流调度的执行跨度和执行代价的同步优化,提出了一种多目标最优化进化遗传调度算法MOEGA。该算法以进化遗传为基础,定义了任务与虚拟机映射、虚拟机与主机部署间的编码机制,设计了满足多目标优化的适应度函数。同时,为了满足种群的多样性,在调度方案中引入了交叉与变异操作,并使用启发式方法进行种群初始化。通过4种现实科学工作流的仿真实验,将其与同类型算法进行了性能比较。结果表明,MOEGA算法不仅可以满足工作流截止时间约束,而且在降低任务执行跨度与执行代价的综合性能方面也优于其他算法。
英文摘要:
      To implement the synchronous optimization of makespan and execution cost of scientific workflow scheduling in cloud environment,this paper proposed a multi-objective optimization evoluation genetic scheduling algorithm named MOEGA.Based on evoluation genetics,MOEGA defines the encoding mechanism of the mapping between tasks and virtual machines,virtual machines and hosts placement,and designs the fitness function satisfying multi-objective optimization.Meanwhile,for meeting the diversity of population,the crossover operation and mutation operation are introduced into the scheduling scheme,and the heuristics is used to initialize the population.Through the experimental tests of four types of scientific workflow in reality,its performance was compared with the same types of algorithms.The results show that MOEGA not only can meet the deadline constraint of workflow,but also outperforms other algorithms in overall performance of reducing the execution makespan and execution cost.
查看全文  查看/发表评论  下载PDF阅读器