A storage device time series classification early warning method

Abstract

The invention discloses a time series classification and early warning method for storage equipment, which comprises: real-time collection of storage equipment parameters; data cleaning; ARIMA time series analysis; logistic regression analysis and early warning mechanism output steps. Under the environment background of big data, the present invention uses the ARIMA model to carry out time series prediction analysis according to the historical data obtained by statistics and the SMART information of the hard disk, and analyzes the correlation between the SMART eigenvalue and the failure rate of the storage device, and then selects a more suitable Logistic model eigenvalues ​​for classification prediction. The invention adopts the method of machine learning to predict the failure rate of the storage device, solves the problem of single classification and low intensity of early warning in the final prediction of the storage device, and overcomes the hysteresis and low accuracy of the early warning mechanism for the disk in the prior art , The actual early warning is not effective, and it is difficult to apply to the defects of the big data environment. It can predict the probability of occurrence of each type of early warning intensity, and provides an effective solution for real-time operation and maintenance and monitoring in the data center environment.

Description

technical field [0001] The present invention relates to the field of storage devices in a data center environment, including storage devices such as mechanical hard disks (Hard DiskDrive, HDD), solid state disks (Solid State Drive, SSD), hybrid hard disks (Hybrid Hard Disk, HHD) and disk arrays, More specifically, it relates to a comprehensive prediction implementation method for performance analysis and load analysis of a storage device early warning mechanism in a data center environment that integrates a time series prediction analysis model and a logistic regression classification model. Background technique [0002] In today's information age, a large amount of new information is generated every day. The total amount of global data is increasing at a rate of 50% per year. Nowadays, more and more data are stored in data centers. Storage is an indispensable part of data centers. Any data center data must eventually be placed on storage devices. As data The scale of the c...

AI Technical Summary

Problems solved by technology

[0014] The present invention proposes a storage device timing classification early warning method, which is used to improve the accuracy of storage device fault prediction on the premise of satisfying the low false alarm rate of prediction, and solve the hysteresis and accuracy of the storage device fault early warning mechanism in the prior art The rate is low, the actual early warning is not effective, and it is difficult to apply to the technical problems of the big data environment

Images