Predicting disk drive failure using condition based monitoring
Paul H. Franklin · 2017
This experience paper describes a model for identifying consumer devices at risk of failure due hard drive vulnerability. A model for estimating time to replace at risk equipment is proposed. This model is based on an empirical observation that reallocated sector count, a metric recorded by the disk drive, increases prior to failure This metric provides a pre-failure signature that can be monitored. Also, the probability of an individual device being removed from service can be determined from the value of reallocated sector count. This model provides useful guidance for setting corporate policy for managing returns.