摘要
传统的部署算法会受部署资源关系的影响导致部署时间过长,为此设计一个基于负荷感知的虚拟机集群部署方法。通过基于SVD(Singular Value Decomposition,奇异值分解)的系统性能预测模型判断和预测虚拟机资源的性能,根据资源性能的描述,以降低成本和减少资源浪费为目标,合理选择虚拟机。计算物理主机与虚拟机的相关系数,同时采用负荷感知法结合虚拟机内存与宽带之间关联性,通过加权处理计算资源权重和对主机排序,进而完成物理主机与虚拟机之间资源匹配,实现基于负荷感知的虚拟机集群部署。实验将部署时间与部署的失败次数作为对比指标,结果表明论文方法在短时间内就能够完成集群部署,并且失败次数较少,满足方法设计的需求。
The traditional deployment algorithm is affected by the relationship of deployment resources,and the deployment time is too long.For this reason,a virtual machine cluster deployment approach based on load perception is designed.This method describes the resource performance of the virtual machine before deployment,calculates the correlation coefficient between the server and the virtual machine,the ratio of available bandwidth,etc.,sorts the hosts according to the results,calculates the resource weight after sorting,and obtains the virtual According to the matching result between the machine and the server,the virtual machine is started according to the matching result,thereby completing the cloud computing virtual machine cluster deployment based on load perception.The experiment takes the deployment time and the number of deployment failures as a comparison index.The results show that the research method can complete the cluster deployment in a short time,and the number of failures is less,which meets the needs of method design.
作者
徐胜超
熊茂华
叶力洪
XU Shengchao;XIONG Maohua;YE Lihong(School of Date Science,Guangzhou Huashang College,Guangzhou 511300)
出处
《计算机与数字工程》
2022年第6期1167-1170,1195,共5页
Computer & Digital Engineering
基金
国家自然科学基金项目(青年基金)“非线性特性多智能体系统一致性研究及其应用”(编号:61403219)
广东省高等学校科学研究特色创新项目(编号:2021KTSCX167)
广州华商学院校内导师制科研项目(编号:2021HSDS15)资助。
关键词
奇异值分解
负荷感知
云计算
虚拟机
集群部署
singular value decomposition
load perception
cloud computing
virtual machine
cluster deployment