一种基于引力搜索机制的云工作流调度算法  

A CLOUD WORKFLOW SCHEDULING ALGORITHM BASED ON GRAVITATIONAL SEARCH MECHANISM

作　　者：王旖旎[1] 李明[2] Wang Yini;Li Ming(Chongqing Business Vocational College,Chongqing 401331,China;School of Computer,Chongqing University,Chongqing 400044,China)

机构地区：[1]重庆商务职业学院,重庆401331 [2]重庆大学计算机学院,重庆400044

出　　处：《计算机应用与软件》2019年第12期29-35,59,共8页Computer Applications and Software

基　　金：重庆市社会科学规划项目(2018PY22);重庆市高等教育教学改革研究项目(163274);重庆商业职业学院校级重点项目(2018XJKTZD06)

摘　　要：为了同步优化云环境中工作流调度长度和代价,提出一种基于引力搜索算法的工作流任务调度算法。算法以异构最早完成时间机制生成引力搜索的部分初始代理,并结合随机生成方式,得到初始种群;利用引力搜索的进化机制,通过代理适应度的评估,得到最终在调度时间和调度代价上综合性能最优的任务映射方案。利用一个算例对算法的有效性进行了论证与评估,并以四种实际科学工作流模型对算法进行了大规模仿真实验。结果表明,该算法不仅可以得到最小的调度代价,且调度时间在所有算法中也是较小的,其综合性能是最优的。For synchronously optimizing the scheduling makespan and cost of workflow in cloud environment,we propose a workflow tasks scheduling algorithm based on gravitational search mechanism.Based on heterogeneous earliest finish time,our algorithm generated some initial agents of gravitational search.Combined with the random way,we could get the initial population.The evoluation mechanism of the gravitational search was applied to get the final tasks mapping solution with optimal comprehensive performance on the scheduling makespan and the scheduling cost through evaluating the fitness of the agent.We used an example to evaluate and argument the effectiveness of the proposed algorithm.And we carried on some large-scale simulation experiments with four kinds of actual scientific workflow models.The experimental results show that the proposed algorithm can get the minimum scheduling cost and smaller scheduling makespan among all algorithms.Its comprehensive performance is optimal.

关 键 词：任务调度 引力搜索算法 调度长度 调度代价 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]