false disk failures, an interesting article @ techspot...

Discussion in 'Storage & Backup' started by tiro_uspsss, Mar 29, 2013.

  1. tiro_uspsss

    tiro_uspsss Member

    Joined:
    May 10, 2007
    Messages:
    2,442
  2. SLATYE

    SLATYE SLATYE, not SLAYTE

    Joined:
    Nov 11, 2002
    Messages:
    26,858
    Location:
    Canberra
    Am I missing something here? SATA HDDs are less likely to actually fail?

    300K SAS drives and 15 - 30 failures per day; call it 22.5 failures per day on average. One failure per 13333 drives per day.

    1.2M SATA drives and 60 - 80 failures per day; call it 70 failures per day on average. One failure per 17143 drives per day.
     
  3. OldnBold

    OldnBold Member

    Joined:
    Oct 5, 2004
    Messages:
    4,261
    Location:
    Sunshine Coast
    You are missing something:

    "What’s interesting is the relative failure rate of SAS drives vs. SATA. It’s about an order of magnitude worse in SATA drives than SAS. "

    SATA drives = greater failure rate according to him. Meanwhile we have not seen any stats or research on this. So who cares what he says .. and, oh by the way, he does have an interest in SAS drives (LSI guy) .. so meh.
     
    Last edited: Mar 29, 2013
  4. elvis

    elvis Old school old fool

    Joined:
    Jun 27, 2001
    Messages:
    45,349
    Location:
    Brisbane
  5. OP
    OP
    tiro_uspsss

    tiro_uspsss Member

    Joined:
    May 10, 2007
    Messages:
    2,442
    LSI sells controllers that support both SAS & SATA, so I don't see a bias in that. His words have more weight than your opinion.

    Well done on the misquote & misinformation. :rolleyes: :thumbdn: :thumbdn: :thumbdn:
    I just read the 13 page PDF of that study:

    So thats SATA & PATA, no SCSI or SAS & no enterprise grade included. :tired:
     
  6. rowan194

    rowan194 Member

    Joined:
    Jan 5, 2009
    Messages:
    2,047
    I've had one of those drives. When put into a RAID1 array it starts clicking away, then is disconnected a short period later due to errors and timeouts. The strange thing is that it's quite happy to run as a standalone drive... no problems at all. A zero fill, which is basically functionally the same as an array rebuild, succeeds without any errors.

    In both instances it's sitting in exactly the same computer, same SATA port... it's just something about the RAID1 rebuild that makes it play up.
     
  7. elvis

    elvis Old school old fool

    Joined:
    Jun 27, 2001
    Messages:
    45,349
    Location:
    Brisbane
    Wow, aren't you a cranky guy then? From the PDF, page 11:

    Enterprise or not, drives fail, and fail often. There are lies, there are damned lies, and then there are statistics. MTBF is a nice statistic, but I've bought million dollar SANs full of enterprise (SCSI, SAS and FC across two decades) disks where I've had drives die in the first 24 hours of operation.

    Drives die. Spend all the money you like, they still die. Buy the most expensive ones you like, they still die (and do so before the advertised MTBF, because it's a mean, not an absolute).

    This is why we have RAID, and this is why we have backups on top of RAID.

    And finally:

    When you ask people for their thoughts and comments, you get their thoughts and comments. Being a prick in return like this:
    ...doesn't do much to make people want to contribute to your threads.
     
  8. OldnBold

    OldnBold Member

    Joined:
    Oct 5, 2004
    Messages:
    4,261
    Location:
    Sunshine Coast

    Could not have said it better myself :thumbup:
     
  9. OP
    OP
    tiro_uspsss

    tiro_uspsss Member

    Joined:
    May 10, 2007
    Messages:
    2,442
    Your reading comprehension sucks. :thumbdn: You stated Google did a study which included SAS & SCSI drives. They didn't - read your what you quoted again. Talagala and Patterson did a study, albeit on a small sample, of 368 SCSI drives, sure, but not Google like you quoted.

    never disagreed :thumbup:

    If simply disagreeing with someone & giving weight to one individuals words more than another makes me prickly, by all means feel free, the both of you, to man the heck up :thumbup:
    Bringing your 'feelings' into this isn't a contribution :thumbdn:
    If you want to disagree with the LSI guy, no problemo! :)
    Post something meaningful, maybe you've done your own study, maybe you know of a study; but simply calling him (LSI guy) bias with no prove etc is lame :thumbdn: Even yourself, elvis, could have some 'weighty' perspective since you deal with 800TB arrays ;) :thumbup:
     
  10. led_blind

    led_blind Member

    Joined:
    Oct 23, 2006
    Messages:
    646
    sound like an end to this discussion....

    Next...
     
  11. Iceman

    Iceman Member

    Joined:
    Jun 27, 2001
    Messages:
    6,647
    Location:
    Brisbane (nth), Australia
    All true, doesn't matter how much you pay, drives die.

    One caveat to note: Any time you're reading about anyone (but especially "san vendor's") reporting on failed disks you need to keep in mind that the metric for a "failed disk" on say your windows desktop (reporting via SMART) is different to your SME ZFS array is different again to your enterprise array.

    This is the crux of the article. Yes a drive might lose its marbles and then be fine. When you're paying a couple million for your SAN, it isn't worth trusting that drive any longer. When it's in your desktop and that $200 drive is 1/3rd the cost of your system and the cost of reinstalling is next to nothing, then it's worth continuing to trust it.


    This, a hundred times this. I'm getting a bit of " This user is on your Ignore List." lately. Some people are no longer worth it.
     
  12. Luke212

    Luke212 Member

    Joined:
    Feb 26, 2003
    Messages:
    10,224
    Location:
    Sydney
    sas failed 0.01% per day
    sata failed 0.007% per day.

    sas false failed 0.0045% per day
    sata false failed 0.0036% per day

    it implies both types are similar. then he says sata is order of magnitude worse?! im with slate, this does not make sense to me.
     
  13. OP
    OP
    tiro_uspsss

    tiro_uspsss Member

    Joined:
    May 10, 2007
    Messages:
    2,442
    not sure how you arrived at that the figure in bold :confused:

    maybe the statement re: SAS vs SATA is done with the fail rates minus the false fail rates? *shrug*
     

Share This Page

Advertisement: