Hard drive Spin down or not?

[ccsmoke]

On my server I always leave them spinning, on my other computers I power down. It's really a double edge sword, as intial shock of start-up and loading to resting postion can indead flex the gimble assembly. Likewise, spinning constantly will produce heat which the metal of the suspension can warp as all metal does when heated. Had 3 harddrive failures over 15 years, 2 of which were 10k raptors (first versions) in which I had them foolishly stacked close together in harddrive cage.

I think the quality of harddrives and heat is the biggest reasons for harddrive failures. One of the reasons I use full size cases and clean out cases every 3 months with water free compressed air. Every manufactor has good designs and bad designs and basically come down to 3 types of failures.

1) circuit board on the drive fail
2) squealing...ball bearing issue where disk is not or can not maintain speed.
3) clicking...arm or read/write head failure.

20 years ago I started work for a company, Hutchinson Technology, in which I worked there for 10 years. In their prime, they did have 75% of the market in read/write suspensions. By no means am I an expert in harddrives, but do have a good knowledge base on the suspension part of the hardrives.

The role suspension I usually describe it as a needle on a record player. It doesn't come in contact with the disk itself but rather maintains a height above the disk called fly height. The flexure, where the head is mounted, is ultra thin, thinner than a human hair.

http://physics.gac.edu/~huber/Presen...#slide0061.htm

http://www.eetimes.com/design/embedd...ata-and-drives

[Bizarroterl]

I work in IT and do a lot of disaster planning. One of the things we've seen is where an array (typically raid 5) is up for years and a drive finally fails. If the array is powered down there's a increased probability that one of the other drives won't spin up when the array is powered back up. We now will always install a replacement drive, rebuild, then go though a power down/restart.

If you spin your drives continuously it would be a good idea to reboot once or twice a year to lessen the chance you don't lose more than one when you finally do reboot. This is important for raid5 arrays, not so much for single drives or parity based arrays.