摘要
针对数据中心规模和任务需求不断增加带来的服务效率降低等问题,提出了一种负载均衡多群落粒子群优化任务调度方法。通过改进的适应度函数对任务的最大完工时间和各机器完工时间方差进行组合优化以提升集群的负载均衡性;利用新的自适应惯性权重函数改进粒子搜索效率,提升算法收敛速率;采取新的粒子初始化方法提高初始解的质量和多样性,并利用多群落粒子协同搜索使得最终结果更加接近最优解。在阿里巴巴数据中心的公开数据集上对算法性能进行了验证和对比,实验结果表明,该方法能够提高数据中心在多样化供应链环境下的任务调度效率。
In order to deal with the problems of low service efficiency brought by the increasing scale of datacenters and task demands,a load balancing multi-swarm PSO task scheduling approach is proposed.Through the improved fitness function,the maximum completion time of the task and the variance of the completion time among machines are optimized to improve the cluster’s load balance.A novel adaptive inertia weights method is designed to enhance particle search efficiency and algorithm convergence speed.Meanwhile,a new particle initialization method is adopted to improve the quality and diversity of the initial solution.Multi-swarm particle collaborative search is further used to bring the final result closer to the optimal solution.The performance of the proposed algorithm is verified and compared with others based on the public dataset of Alibaba datacenter.The experimental results show that the method can improve task scheduling efficiency of datacenters in diversified supply chain environments.
作者
曾磊
白金明
刘琦
ZENG Lei;BAI Jinming;LIU Qi(School of Computer,Nanjing University of Information Science and Technology,Nanjing 210044,Jiangsu,China;School of Applied Meteorology,Nanjing University of Information Science and Technology,Nanjing 210044,Jiangsu,China;Engineering Research Center of Digital Forensics,Nanjing University of Information Science and Technology,Nanjing 210044,Jiangsu,China)
出处
《应用科学学报》
CAS
CSCD
北大核心
2023年第3期419-430,共12页
Journal of Applied Sciences
基金
国家自然科学基金(No.62002276,No.41911530242,No.41975142)
国家社科基金重大项目(No.17ZDA092)
英国爱丁堡皇家学会和中国自然科学基金委员会联合国际项目
江苏省基础研究计划基金(No.BK20191398)资助。
关键词
粒子群优化
供应链管理
数据中心
任务调度
负载均衡
particle swarm optimization
supply chain management
datacenter
task scheduling
load balancing
