论文部分内容阅读
虚拟化的数据中心主要采用预拷贝(Pre-copy)算法进行虚拟机实时迁移.当虚拟机中运行负载较高或者网络传输带宽较低时,预拷贝算法固定的停机阈值严重影响实时迁移的性能.针对这个问题,提出一种自适应停机阈值机制,根据之前各拷贝轮中脏页率构成的时间序列,首先利用动态指数平滑法来预测后轮的脏页率,在预测脏页率超过网络传输带宽的情况下,再采用Mann-Kendall检验模型对脏页率变化趋势进行判断,根据判断结果确定停机切换的时机.实验结果证明,采用基于自适应阈值机制的预拷贝算法能在高负载低延迟场景下有效提高实时迁移性能.
The virtualized data center mainly adopts the pre-copy algorithm for real-time virtual machine migration, and when the running load of the virtual machine is high or the transmission bandwidth of the network is low, the fixed shutdown threshold of the pre-copy algorithm seriously affects the performance of the real-time migration In order to solve this problem, an adaptive threshold stop mechanism is proposed. According to the time series of dirty pages in each copy wheel, the dynamic index smoothing method is used to predict the dirty page rate of the rear wheel. When the predicted dirty page rate exceeds the network And the transmission bandwidth, the Mann-Kendall test model is used to judge the change trend of the dirty page rate, and the timing of the shut-down switching is determined according to the judgment result.Experimental results show that the pre-copy algorithm based on the adaptive threshold mechanism can be used in high- Delays scenario effectively improves real-time migration performance.