尽管预拷贝技术由于其毫秒级的停机时间及对应用影响较小而被广泛使用到各大虚拟化平台的内存数据迁移中,但是当网络处于低带宽高负载的情况下则会导致虚拟机服务的一些性能下降。为了解决以上问题,本文首先将传统的预拷贝机制与马尔科夫模型结合旨在对后轮脏页率进行预测,利用Mann-Kendall检验模型对迁移中脏页率的变化趋势进行判断,据此引入自适应阈值机制确定最优迁移策略,更好地解决虚拟机中网络传输带宽的大小和运行负载的高低对传统预拷贝技术固定停机阈值机制的影响,实验结果显示,改进后的迁移机制可有效地提高虚拟机实时迁移性能。
Although the pre-copy technology because of its millisecond application downtime and the impact on small and widely used virtual memory to the major data migration platform, but when low-bandwidth network in high-load will result in a virtual machine some of the performance of the service decline. To solve this problem, this paper will be the traditional pre-copy mecha- nism combined with the Markov model seeks to predict the rate of the rear wheel dirty pages using Mann-Kendall test model for the migration of dirty pages rate trends to judge, pursuant to the introduction of adaptive threshold mechanism to determine the op- timal migration strategy to better address the impact of the virtual machine network bandwidth size and the level of operating loads on the traditional pro-copy technology threshold fixed stop mechanism, the experimental results show that the improved transfer mechanism can be effectively improve virtual machine live migration performance.