I just read Brian Beach’s blog for Backblaze and am amazed: His comparison of Backblaze storage pods that use consumer drives with enterprise storage controllers that contain enterprise disk drives from EMC and Dell is like comparing apples and flying pigs.
EMC and Dell both have detailed information on which metrics are evaluated before they decide that a drive is going to fail. This is often referred to as sparing a drive by looking at the SMART data and it prompts admins to begin the rebuild process before the drive dies. There is nothing in the blog that discusses this. In addition Mr. Beach does not explain what SMART data he collects and how it is used. Is it the same that EMC and Dell have collected and researched for years, or just the basic information that is published – not the special, priveledged information that the drive vendors share with their partners?
Another issue I see is that RAID controllers will spare a drive if the performance is not up to specification over a certain period of time. For example I know of one RAID vendor that kills a drive if it does not respond in five seconds, far less than most drive timeout values. Also, are we comparing enterprise drives that are 3 years old with brand new consumer drives? What are the brands, what are the ages of the comparisons?
There was a study a few years ago that compared FC drives to SATA drives. The FC drives in question were from the early 2000s and have a known bearing issue and the SATA drives were from the mid-2000s (about 3 years later) and had no known issues. Guess what was better? Without a great deal more detailed information, I consider this study lacking technical rigor and it provides potentially very misleading information. I suspect that after today the major drive vendors will be quietly saying something. At least I hope so, as without more detailed information this study provides no new information.