摘要
虚拟机资源是云计算的基本调度单位。当前云计算中虚拟机资源调度策略仅关注系统当下的负载状态,负载均衡服务器会获取CPU利用率和内存空闲率来计算新权值并写入管理模块。但这种方式忽视了系统在未来一段时间的负载变化,导致资源调度的结果不太理想,容易出现负载失衡的情况。因此,提出多云异构环境下CPU并行计算动态负载均衡调度方法。通过搭建MCM平台,实现CPU并行计算资源的状态监测和管理;定期从平台监控系统中收集CPU并行计算负载数据,使用一元线性回归方法对其实施拟合,预测其未来一段时间内的负载变化。设计基于最小化任务迁移次数的动态负载均衡调度方法,根据负载预测结果智能迁移过载任务至轻载虚拟机,实现CPU并行计算动态负载均衡调度。实验结果表明,所提方法能够在准确预测CPU并行计算负载值并拟合负载变化趋势的前提下,显著降低调度能耗,提高负载均衡性能。
Virtual machine resources are the basic scheduling units of cloud computing.Currently,the virtual ma-chine resource scheduling strategies in cloud computing only focus on the current load status of the system.The load balancing server will obtain the CPU utilization rate and the memory free rate to calculate a new weight and write it into the management module.However,this approach ignores the load changes of the system in the future period,re-sulting in less than ideal resource scheduling results and prone to load imbalance.Therefore,a dynamic load balancing scheduling method for CPU parallel computing in a cloudy heterogeneous environment is proposed.By building an MCM platform,the state monitoring and management of CPU parallel computing resources are realized;Regularly col-lect CPU parallel computing load data from the plaform monitoring system,and use a univariate linear regression method to fit the data and predict its load changes over the next period.A dynamic load balancing scheduling method based on minimizing task migration times is designed,which intelligently migrates overloaded tasks to lightly loaded virtual machines based on load forecasting results,achieving dynamic load balancing scheduling for CPU parallel com-puting.Experimental results show that the proposed method can significantly reduce scheduling energy consumption and improve load balancing performance while accurately predicting CPU parallel computing load values and fitting the load change trends.
作者
邹昊东
黎俊杰
吴旺东
王磊
ZOU Hao-dong;LI Jun-jie;WU Wang-dong;WANG Lei(State Grid Jiangsu Electric Power Co.,Ltd.Information&Telecommunication Branch,Nanjing Jiangsu 210002,China;State Grid Electric Power Research Institute Co.,LTD,Nanjing Jiangsu 211000,China)
出处
《计算机仿真》
2025年第9期186-190,共5页
Computer Simulation
基金
国家电网有限公司科技项目(5700-202318307A-1-1-ZN)。
关键词
多云异构环境
并行计算
负载预测
负载均衡调度
Cloudy Heterogeneous Environment
Parallel Computing
Load Forecasting
Load Balancing Schedu-ling
