On 08/29/2014 06:15 PM, Chris Cappuccio wrote: > Evan Root [cellarr...@gmail.com] wrote: >> It seems that after reading the backblaze and google papers about drive >> reliability that there is no statistically obvious difference. It's too >> close to call. Both papers end with hedges and further questions. Even if >> enterprise drives are more reliable it isn't by more than a single percent >> and it isn't consistent enough to matter either. >> > The difference in firmware between consumer and enterprise drives can > be papered over with a good RAID controller. The kernel might need to timeout > or do something. But the problem should be attacked there instead of saying > "oh just buy the 5x cost drive". It depends a lot on how you intend to use them: how annoyed you get if you have to replace one, how annoyed you get if the array goes off-line or goes into degraded mode when a drive performs its most extensive bad block recovery, how annoyed you get when the RAID controller mistakenly takes a drive off line, and how much money and effort you're willing to put into the case in which you mount the drives.
You could set the RAID timer to wait for multiple minutes before declared the drive offline. That would effectively put the RAID in degraded mode during that time - writing would be disabled. You might be able to disable the drive's extended error recovery with an internal parameter setting, requiring the RAID controller to do bad block remapping but the drive will never take a long time to respond. That would minimize the time that the RAID was in degraded mode or off line. If you did neither, then when a drive did extended error recovery, the controller would declare the drive off-line. You would have to intervene and figure out whether the drive was really bad or that you just needed to reset the controller's status. The controller firmware is unlikely to do that for you. You could install the drives in a well-cooled vibration-isolating case. That would minimize seek errors caused by vibration. excess spindle bearing wear, and premature failure due to overheating. The consumer drives are not designed to read, write, and seek continuously. Consumer multiple-drive bays and boxes usually have no vibration isolation and poor cooling. Given the above precautions, a consumer drive could work very well. I had several very premature drive failures using consumer drives in a consumer multi-bay case. Since then I've always mounted drives with considerable space between them and have had no failures. YMMV Geoff Steckel
