面向可扩展集群环境的快速虚拟机迁移方法  

Rapid migration method of virtual machine for extensible cluster environment

作　　者：刘诗海[1] 孙宇清[1] 石维琪[1] 高云伟[2] 

机构地区：[1]山东大学计算机科学与技术学院,济南250101 [2]中国科学院计算技术研究所计算机系统结构重点实验室,北京100190

出　　处：《东南大学学报（自然科学版）》2011年第3期468-472,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61070017);山东省自然科学基金资助项目(Y2008G28);中国科学院计算机系统结构重点实验室开放课题资助项目

摘　　要：在集群环境下,为了提高虚拟机Xen的迁移效率,提出了一种面向可扩展集群环境的快速可靠的虚拟机迁移方法.通过在分布式文件系统GlusterFS中增加Heartbeat监测模块,实时监测集群环境中物理机节点的加入与退出,动态创建或修改GlusterFS文件系统的配置文件.为了使GlusterFS文件系统能够自动管理节点,设计实现了基于gSOAP的内部角色通信,比传统方法增加了动态挂(卸)载集群环境中物理机节点的功能,达到了实时更新文件存储的拓扑结构的目的,在保证迁移性能和文件可靠性的同时提高了系统的扩展能力.实验结果表明,该方法能够在虚拟机迁移过程中自动分配资源,在充分利用集群资源和保证文件完整性的前提下,提高了迁移效率.In order to improve the efficiency of Xen virtual machine migration in cluster environments,a fast and reliable virtual machine（VM） migration solution is proposed based on GlusterFS for clusters with high scability.The Heartbeat mechanism is introduced into GlusterFS to monitor the changes of network topology and the information about node mount and node unmount.It can dynamically create and modify the GlusterFS configuration.The communication of inner roles is implemented by using gSOAP in order to automatically manage the nodes and mount（unmount） the dynamic physical nodes.The file storage topology then can be updated in real-time.The migration performance can provide more guarantee for system scalability without any loss of the reliability of FS.Some experiments were performed to verify the proposed method and the results show that it can automatically allocate system resources during the process of VM migration.The efficiency of VM migration and the utilization of cluster resources are also improved.

关 键 词：集群环境 虚拟机 迁移 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]