BaNHFaP: A Bayesian Network Based Failure Prediction Approach for Hard Disk Drives

Abstract

A Hard Disk Drive (HDD) failure may lead to serious consequences for users and companies. Hence, predicting failures in HDDs became a topic that attracted much attention in recent years. Monitoring a HDD status can provide information about its degradation, so as to let the user or a system manager know about a failure before it happens, preventing loss of information. In this paper, we propose a failure prediction method using a Bayesian Network. Our method uses the deterioration over time of a HDD, calculated via SMART (SelfMonitoring Analysis and Reporting Technology) attributes, for predicting eventual failures. To demonstrate practical usefulness, this method was applied to a dataset consisting of 49,056 hard drives from Backblaze's data centers. The proposed method has improved the mean and median quadratic errors in 28.3% and 17.6% respectively in comparison with a baseline model.