一种改进樽海鞘群算法及其多目标云工作流调度应用  

AN IMPROVED SALP SWARM ALGORITHM AND IT'S APPLICATION ON MULTI-OBJECTIVE CLOUD WORKFLOW SCHEDULING

作　　者：李果[1] 陈信 吴迎来[3] Li Guo;Chen Xin;Wu Yinglai(Anhui Vocational College of Press and Pubilishing,Hefei 230601,Anhui,China;School of Computer,Hangzhou Dianzi University,Hangzhou 310018,Zhejiang,China;Institute of Innovation and Entrepreneurship,Hangzhou Dianzi University,Hangzhou 310018,Zhejiang,China)

机构地区：[1]安徽新闻出版职业技术学院,安徽合肥230601 [2]杭州电子科技大学计算机学院,浙江杭州310018 [3]杭州电子科技大学创业学院,浙江杭州310018

出　　处：《计算机应用与软件》2023年第12期263-271,331,共10页Computer Applications and Software

基　　金：安徽省教育厅自然科学重点研究项目(KJ2017A870);安徽省教育厅省级融媒体采编教学创新团队项目(2019cxtd101)。

摘　　要：为了优化云工作流应用的调度效率与代价,提出基于改进樽海鞘群算法的工作流调度策略。建立截止时间与预算约束的多目标优化模型,利用樽海鞘觅食的位置变化对工作流调度进行编解码,设计融合执行跨度与执行代价的权重适应度函数。为了增强樽海鞘群的寻优性能,引入基于疯狂算子的领导者更新模式,通过疯狂变量,减少领导者更新的停滞早熟现象;引入遗传算子的追随者更新模式,利用个体交叉和变异使樽海鞘群具有更均衡的搜索全局性和个体多样性,避免陷入局部最优。结果证明,改进樽海鞘群算法可以有效提升计算精度和收敛速度;应用于工作流调度求解后,其调度解收敛性更好,调度解集空间分布更加一致。For optimizing scheduling efficiency and cost of cloud workflow applications,we propose a workflow scheduling strategy based on improved salp swarm algorithm.A multi-objective scheduling optimization model meeting the deadline and budget was built.We used the location changes of salps foraging to code and decode the workflow scheduling.We designed a weighted fitness function intergrating the execution makespan and the execution cost.In order to enhance the performance of salp swarm,the algorithm introduced a leader update model based on crazy operator.By this crazy variable,we could reduce stagnation precocious phenomenon in leader update.This algorithm also introduced a follower update model based on genetic operators,which could use the individual crossover and mutation to make salp group having a more balanced global search and individual diversity.Experiments prove that the improved algorithm can promote the accuracy and convergence rate.When solving the workflow schedule problem,the new algorithm has a faster convergence rate and more uniform space distribution.

关 键 词：云计算 工作流调度 樽海鞘群算法 疯狂算子 遗传算子 多目标优化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]