[SATLUG] NAS HD recommendation
brad at shub-internet.org
Tue May 21 22:39:27 CDT 2013
On May 21, 2013, at 5:22 PM, Alex Bartonek <bartonekdragracing at yahoo.com> wrote:
> I'm looking at finally putting together my home NAS. Its a 1U rack mount, 6 SATA ports, 1 will have a 2.5" 320gb boot drive, then 4 slots for 3TB xxxx brand hard drives.
> I was thinking the "WD Red NAS" drives. I'm going to leave it running 24/7 so I didnt want to run regular SATA drives since I at least want some sort of durable drive and it will be in a RAID config running CentOS.
I have personally had bad experiences with Western Digital drives, so I wouldn't buy them again. I've also had failures with Seagate drives, but not to anywhere close to the same failure ratio.
If you're Google or Amazon, then you can take advantage of the fact that "consumer" grade drives don't appear to have any detectable large-scale failure rate that is statistically greater than the "enterprise" grade drives. Just buy a few gazillion drives and make your storage systems sufficiently redundant -- Google used to use a "triple store" mechanism years ago, but I don't know what they're using now.
Of course, the key to building a really reliable storage system is to ensure that you never actually have to "rebuild" the array, because that is when the array will be under the greatest level of stress and you are most likely to have a secondary failure. Just mark the bad drive as bad, use your existing additional copies of the data to re-generate new good copies to make up for the loss in redundancy, and then when the bad drive is replaced and a new drive is added, you let nature take its course.
Of course, you'll want to thin-provision the storage you provide from that array, too.
If you're not Google or Amazon, then you've got a harder task ahead of you, because the "second failure while rebuilding" problem is highly likely to bite you when you do ultimately have a drive failure.
I wouldn't get the full "triple store" redundancy that Google was doing years back, but with ZFS and RAID-Z over the zspools, I could get the same basic effect as a traditional RAID array, but integrated directly into the filesystem in such a way that rebuilds are less painful and more easily managed.
Brad Knowles <brad at shub-internet.org>
LinkedIn Profile: <http://tinyurl.com/y8kpxu>
More information about the SATLUG