SMART information and deep learning-based hard disk damage prediction method and device

Abstract

The invention discloses a SMART information and deep learning-based hard disk damage prediction method and device. The method comprises the following steps of: collecting SMART information of different hard disk in different time periods as original sample data, and selecting two types of training sample data such as static data and time sequence data from the original sample data; screening the training sample data, extracting an effective parameter set related with hard disk faults and carrying out labeling to obtain labeled training data; respectively constructing at least one neural network model to carry out deep learning on the labeled static data and the labeled time sequence data so as to obtain corresponding neural network models for optimizing model parameters; and obtaining SMART information data of a hard disk in a storage system according to a set timing strategy, and extracting the effective parameter set to input into the neural network model so as to predict whether thehard disk is going to be damaged. The method and device are capable discovering correctly hard disk conditions in time before mechanical hard disks are going to be damaged, so as to carry out timelychange before hard disk failure, thereby greatly improving the high availability and data safety of data centers.

Description

technical field [0001] The present invention relates to the technical field of storage security, in particular to a method and device for predicting imminent damage of a hard disk based on hard disk SMART information and deep learning. Background technique [0002] Disk arrays or storage systems use a large number of mechanical hard disks, and mechanical hard disks (collectively referred to as hard disks below) are the most vulnerable components in the entire storage system. Although RAID technology can effectively deal with hard disk damage, hard disk damage still affects the reliability of the system. performance and high availability with a huge impact. [0003] Hard disk failures are generally divided into two types: predictable and unpredictable. The latter happens occasionally, and there is no way to prevent it, such as sudden chip failure, mechanical impact, etc. However, such things as motor bearing wear and disk magnetic medium performance degradation are predicta...

AI Technical Summary

Problems solved by technology

[0005] The SMART signal involves dozens of (more than 30) parameters of hard disk work and wear status. Disk array manufacturers usually set a threshold value for each parameter for early warning. Although this method is simple, it cannot accurately reflect the overall health status of the hard disk and is easy to Causes false positives or no positives, and cannot protect the security of the disk array or storage system

Images