Actual enterprise drive reliability meets expectations
One failure in a million hours? It’s claims like these that seem extreme to some people when they look at enterprise disk drives. Yet a study of 39,000 NetApp systems by a researcher have found that these drives fail at a 1% annual failure rate (AFR). Robin Harris summarizes the study in his blog.
The translation from AFR to MTBF (Mean Time Between Failures) is not exact, but this AFR number puts MTBF in the million-hour ballpark, showing that disk drive specifications do indeed portray actual reliability performance.
It’s hard to test for a failure every 1.6 million hours
This is not an exact science, because to prove that any one drive will only fail on average every 1.6 million hours (the spec for the Seagate Savvio drive), you’d have to run a whole bunch of drives a whole bunch of years. This study is a nice real-world validation!
Great study!
One other point you also could have highlighted was that while it does matter which disc drive is is used because it’s where the data is lost, it is also true that the enclosure design and integration around the drive are significant contributors to the reliability of the data.
So not only is it important to buy the right drive for the application, you should also make sure it is being housed correctly as well!
Thanks Tim. Another interesting thought: does drive and system reliability vary based on application? I would bet it does. NetApp systems have above average capacity I think. That means more drives per system and fewer cumulative IOPS per drive.