Memory disk fault prediction method and prediction system

Abstract

The invention discloses a memory disk fault prediction method and system, and belongs to the field of computer storage, and the method comprises the steps: (1) taking the state data, collected in realtime, of a storage disk to be predicted as input, and predicting the fault condition of the storage disk to be predicted through a trained fault prediction model; (2) judging whether a prediction result is normal or not, and if so, concluding fault prediction; if not, exuecting the step (3); (3) interpreting and analyzing the prediction result to obtain all fault cause sets of the to-be-predictedstorage disk; (4) filtering out the fault cause set with low detection rate or high false alarm rate, thereby obtaining a set S formed by the residual fault cause set; and (5) taking a correspondingfault processing measure on each fault edge in the set S from the attribute item with the lowest centralized processing expense to remove the fault of the storage disk to be predicted, and ending thefault prediction. According to the invention, accuracy and reliability of storage disk fault prediction can be improved, and waste of storage resources is reduced.

Description

technical field [0001] The invention belongs to the field of computer storage, and more specifically relates to a storage disk failure prediction method and a prediction system. Background technique [0002] At present, storage disks such as magnetic disks are important storage devices for data storage in data centers. Once a storage disk fails, it will cause huge data loss if there is no backup or copy of the data. Data can incur huge overhead, affecting the uninterrupted online service of data. [0003] When storing data, there are many reasons for the failure of the storage disk. Common failure reasons include temperature, the number of remapped sectors, power-on times, and accumulated power-on time. In order to avoid data loss or huge data recovery costs caused by storage disk failures, data centers usually collect SMART data and I / O load statistics of storage disks in real time, and then use machine learning algorithms to build fault prediction models to Predict the w...

AI Technical Summary

Problems solved by technology

[0004] At present, there are already some methods and systems for fault prediction of storage disks, but these methods and systems only focus on predicting whether the disk is faulty and the degree of wear and tear, and lack further analysis of the prediction model. Therefore, only a single fault can be provided. Predict the results and take a single fault handling measure, which can easily cause model overfitting, resulting in low accuracy and reliability of fault prediction results, and may lead to waste of disk resources

Images