一种面向云的Volterra级数的自适应容错预测方法（英文）

Failure detection module is one of important components in fault-tolerant distributed systems,especially cloud platform.However,to achieve fast and accurate detection of failure becomes more and more difficult especially when network and other resources' status keep changing.This study presented an efficient adaptive failure detection mechanism based on volterra series,which can use a small amount of data for predicting.The mechanism uses a volterra filter for time series prediction and a decision tree for decision making.Major contributions are applying volterra filter in cloud failure prediction,and introducing a user factor for different QoS requirements in different modules and levels of IaaS.Detailed implementation is proposed,and an evaluation is performed in Beijing and Guangzhou experiment environment.     

一种基于Volterra级数的云平台自适应失效检测机制

失效检测是分布式系统特别是云平台容错的关键组成部分。然而由于网络状态的不断变化,要实现对失效的快速准确的检测变得比较困难。为应对这种情况,本文提出了一种基于Volterra级数的高效的自适应失效检测机制。该机制基于Volterra滤波器实现,能够只利用很少的训练样本就可以对时间序列做出很好的预测,且可根据预测误差来自适应地调整以使得预测结果更准确。为了适应云平台中各功能模块对失效检测的不同QoS需求,本文提出的失效检测机制在统计反馈部分引入调节因子α,可以方便地满足不同应用的QoS需求。