Wed, 05 Mar 2008
For a long time I have been wondering how the so called "server-grade" hard disk drives differ from the consumer-grade hardware. I tend to think the difference is precisely zero, provided that the other specs (RPM and capacity) are the same.
There is a well-known paper from Google researchers (PDF warning), which supports my opinion. For example, in our IBM FAStT-600 array with 28 250 GB SATA drives we have replaced about half of the drives, and from the consumer-grade 250 GB P-ATA drives in our other servers the failure rate is about the same.
Last week we have replaced our backup server by a new hardware with newer and bigger drives, and I have looked at the old server whether some of the drives
would still be usable: I have found two 250 GB P-ATA drives which both had
zero errors logged in the SMART log, and both had over 36,000 hours of usage
(4 years and 1.5 month)! I remember they were previously installed in
ftp.linux.cz), where they were almost continuously busy for at least a year and half.
So the next time some salesdroid would try to persuade you about the different firmware and higher manufacturing standards of the "server-grade" drives, do not buy the snake oil!
1 replies for this story:
Spes wrote: RAID edition drives
BTW when I was buying some drives to fadmin.fi.muni.cz, I was interested in what is the difference between ordinary and RAID edition drivers by WD. I got answer that the RAID drives try to obfuscate SMART log as long as a (more) serious problem occurs.