虚拟化技术为用户提供了高可用性、动态、可扩展、可按需分配的逻辑资源，虚拟机迁移技术则减弱了虚拟化初次分配资源后物理资源和逻辑资源的耦合程度，使得物理资源池构建更为灵活。然而现存的虚拟机迁移技术存在着资源消耗多，物理磁盘负载重，迁移数据冗余等问题，使得迁移的稳定性和可用性大打折扣。提出基于内存缓存的动态块迁移算法。该算法主要关注以下两点：其一是如何在保证迁移性能不受明显影响的前提下，更加合理利用内存缓存，快速将虚拟机页面迁移到目的服务器，从而节省物理资源；其二是如何通过更细粒度的资源管理方式实现更有针对性的迁移时机优化。基于 QEMU 虚拟机实现了该算法，多种不同应用负载下的实验结果表明，该算法能有效降低资源消耗和物理磁盘的负载，稳定且迅速地实现虚拟机的迁移。%Virtualisation technology provides users with highly available,dynamic,scalable and allocation-on-demand logic resources, while the virtual machine migration technology weakens the coupling degree between the physical and logical resources after the initial virtualised allocation of resource,which enables the construction of physical resources pool more flexible.However,there are three main difficulties in existing solutions of virtual machines migration:too many resources are wasted,the hard disks are overloaded,and the redundancy of migration data,they greatly constrict the usability and stability of live migration of virtual machines.In the paper we propose a memory cache-based dynamic block migration algorithm,which focuses on two points:First,how to use the memory cache more reasonably to migrate the pages of virtual machine to the goal server rapidly on the premises of guaranteeing the migration performance not being apparently impacted so as to save physical resource;Secondly,how to achieve more targeted migration timing optimisation by the way of more fine-grained resources management.Based on QEMU emulator,we have implemented the algorithm.The results of experiments on various VM workloads show that the algorithm can effectively lower the resource consumption and the load of physical disks,and achieves the fast and stable migration of virtual machine.
展开▼
