What's worse then the sound of one hard drive going "click, click?"

What’s worse then the sound of one hard drive going “click, click?” Why, two drives going “click, click, click” in the same RAID 5 array, of course.

I’m not very happy with my little Infrant NAS box right now. I think I’ve had it with RAID 5–if I’m going to pile my life onto a disk array, then I really want something that can survive a 2-drive failure without croaking, and that’s basically impossible in a 4-drive enclosure.

I’m seriously considering replacing the Infrant with an OpenSolaris box running ZFS over RAID-Z2 with 6–10 drives; that should live through 2-drive failures, right? Anyone feel the need to talk me out of it?

Posted by Scott Laird Sun, 14 Oct 2007 12:47:00 GMT

Read full article 9 comments


Comments

  1. Boris Mann about 6 hours later:

    First: oh, look, a Scott blog! Yay!

    Next: oh, sh*t, that doesn’t sound like fun…

    We totally need someone to write up cool hacks around storage that everyone can follow. I still haven’t made the jump to really solve my home storage needs.

    Oh, and of course: Tim Bray was just wanting this solved, too: http://www.tbray.org/ongoing/When/200x/2007/10/14/Basement-Storage

  2. Bryan Zug about 10 hours later:

    man that sucks – now you’ve got me thinking about how piddly my superduper mirror setup really is.

  3. Bob Aman about 12 hours later:

    Nope, you’re not gonna get me to talk you out of it. I’ve been planning to do exactly the same for some time. Either that or, maybe now that Leopard’s getting ZFS, getting an XServe RAID box and just setting it up for ZFS instead.

  4. Ben about 13 hours later:

    Have you considered something like Amazon’s S3 and let them deal with the inevitable disk failures!

  5. Scott Laird about 13 hours later:

    In this case, I don’t think that S3 is really a solution. It’s going to be too slow to use via my piddly little DSL line. It might make sense for backups, though.

  6. Chris Walker about 14 hours later:

    A friend of mine is using Nexenta in a VM attached to a RAID-Z2 and is quite pleased with it. It might not get the performance you want. If you have a remote zfs setup as well, zfs send/receive can be handy for remote backups.

    As for Leopard, I don’t think there will be r/w ZFS in 10.5.0, so you might be waiting a bit. This is good, because zfs isn’t all that speedy in the latest seed.

  7. Scott Laird about 17 hours later:

    Chris, compared to the Infrant, I have a hard time seeing how ZFS could be that slow :-). The Infrant usually hit the wall around 20 MBps. If I can get that, maybe with better caching performance, then I’ll be happy. I’d be surprised if it can’t get at least 100 MBps reading big files, but anything over 20 is fine.

  8. John Whitley 1 day later:

    It’s worth checking out Bianca Schroder’s FAST ‘07 paper, “Disk failures in the real world: What does an MTTF of 1,000,000 hours mean to you?”:http://www.cs.cmu.edu/~bianca/fast07.pdf

    It’s quite a nice head-shaping regarding the realities of drive reliability and the implications for RAID vs. multiple-backup strategies, etc. Won the best paper award, too.

  9. Kelly Jones 4 days later:

    Nice post. I’ve been bitten by the “two drives have died on my raid5” too. ZFS looks good except for the learning curve of Solaris. But, if it’s just a storage box, then that’s not a problem. The other problem besides dead drives is accidental deleting of stuff. ZFS snapshots and cron should help with that.