基于脏页率的虚拟机动态内存迁移  

Live Memory Migration of Virtual Machine based on Dirty Page Probability

作　　者：邹潇[1] 訾小超[1,2] 姚立红[1,2] 潘理[1] 

机构地区：[1]上海交通大学信息安全工程学院,上海200240 [2]南京大学计算机软件新技术国家重点实验室,江苏南京210093

出　　处：《通信技术》2013年第3期101-103,106,共4页Communications Technology

基　　金：上海市自然科学基金项目(No.11ZR1418500);上海市自然科学基金项目(No.10ZR1416700);国家科技部973计划课题(No.2013CB329603)

摘　　要：虚拟机动态迁移中,为保证透明性和一致性,内存迁移多采用能平衡总迁移时间和宕机时间的迭代预拷贝。为减少预拷贝过程中的数据传输量、缩短总迁移时间,这里提出了一种基于脏页率预测的预拷贝机制,该机制根据局部性原理,对虚拟机内存页的脏页率做出预测,优先传输脏页率低的内存页,避免高脏页的重传。在Xen中的实验结果表明,该机制可行有效,减少了数据传输量,缩短了总迁移时间。In live virtual machine migration, the pre-copy approach is usually adopted in memory migration, and this approach could balance the contradictions of total migration time and downtime for transparency and consistency. shortening total migration time, an improved For reducing pages transferred in pre-copy and pre-copy mechanism based on dirty page probability forecasting is proposed, in this mechanism, the pages＇ dirty probability is forecasted according to the principle of locality, and the pages with low dirty probability is preferentially transmitted for avoiding unnecessary transmission of high dirty pages. The experiment results in Xen indicate that the proposed mechanism is feasible and effective, and could reduce the number of transferred pages and the total migration time.

关 键 词：动态迁移 预拷贝 局部性原理 脏页率 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]