基于压缩感知的云存储系统状态监测方法

State detection method for cloud storage system based on compressive sensing

为了解决大规模状态监测数据处理中高精度与大规模之间的矛盾,利用压缩感知处理宽带信号优于奈奎斯特采样定律的特性,提出了一种适合于测量云存储系统状态的压缩感知状态监测方法 SDCS.该方法是在经典匹配追踪MP算法的基础上,增加贝努利矩阵行和为零的约束条件而得到的,可用于测量含直流分量的稀疏信号,并保证原始的重构算法依然满足改进后的目标函数.然后,利用仿真实验测评了该方法在蚁群文件系统FFS状态监控中的应用效果.实验测试结果表明,针对稀疏度为10的状态信息,当测量次数大于70时,所有异常结点可被精确定位,且压缩比率达到3.5%,说明该方法能有效压缩监测流量,满足大规模数据高精度检测的要求.

To solve the contradiction between precision and scale when dealing with large-scale data,by means of compressive sensing in signal processing which is superior to the Nyquist,a method called SDCS（state detection with compressive sensing） for detecting the state of cloud storage system is proposed.Based on the typical MP（matching pursuit） algorithm,this method is developed by adding the constraint condition that the sum of the rows in Bernoulli measurement matrix equals zero.The SDCS can measure sparse signals containing direct current component and ensure the equivalence between the improved target function and the original one.Then,this method is applied to state detection of FFS（formicary file system）.The experimental results show that for the status signs with the sparse degree of 10,when the number of detection is more than 70,all the abnormal nodes can be fixed accurately and the compression ratio is 3.5%,indicating that this method can improve the efficiency of locating the abnormal nodes and satisfy the requirement of high precision detection for the large scale system.