Abstract
For the last four years we’ve been collecting daily SMART statistics from the drives in our data centers. We have data from over 50 different models ranging from 1-8 TB in size and we’d like to share what we’ve observed. We’ll start with annual failure rates of the different drive models and how often do they fail by size, model and manufacturer. Then we’ll look at the failure curve over time and see if hard drives actually fail according to the “bathtub curve” as we expect. Next we’ll compare the failure rates of consumer versus enterprise hard drives. We’ll finish by examining whether or not SMART stats can be used to predict drive failure by reviewing the results of multiple research studies that have completed using the data we publish.