On Fri, Jul 5, 2013 at 11:20 AM, Ben Scott <[email protected]> wrote:
> On Fri, Jul 5, 2013 at 1:37 PM, David Lum <[email protected]> wrote:
>>> Be careful of a SATA Raid array.  I've got a couple of these and in an
>>> effort to save money and I'm not doing it again.  The issue comes down to
>>> something I learned about the hard way.  "Array Puncture"...
>>
>> I have never heard this term before today, looking it up now.
>
>   Me neither.  This explains it:
>
> http://www.dell.com/support/troubleshooting/us/en/04/KCS/KcsArticles/ArticleView?c=us&l=en&s=bsd&docid=438291
>
>   Nothing to do with SATA, it appears.
>
>   It's just a fallback method for a double fault.  A double fault is
> when two blocks (stripes) are both found to be bad.  This is more than
> a RAID 5 can tolerate.  A "puncture" appears to just be an option to
> keep an array operating despite a double fault.  You've got two
> choices when you encounter a double fault, fail the entire array
> offline, or "puncture" it, rebuilding what you can, and returning
> error on those particular blocks.
>
>   The double fault problem has been known for decades, and is why
> better RAID implementations do regular consistency checks across all
> disk members.
>
> -- Ben

Two things to add to that:

o- RAID 6 is more resilient, but of course the controllers cost more,
and there is a performance penalty vs. other RAID categories, which
better ($) controllers can mitigate.

o- Larger arrays (whether made up of larger disks or larger numbers of
disks, or both) will have more errors, and therefore a higher
probability of simultaneous errors, than smaller arrays. They are also
slower to perform consistency scans against.

So, having said that, I'm of the opinion that a RAID 5 array of 16 x
300gb drives is probably OK for RAID 5, especially if one of them is a
dedicated hot spare. I haven't done the math to back that up, but in
the world of 3.5" drives it's been pretty reasonable. Don't know about
OPs situation, as he's considering 2.5" drives, though a quick search
of CDW reveals that 300gb, 2.5", 10k rpm SATA drives don't exist -
although they have them in SAS and SAS-2, in both 10k and 15k speeds.

Kurt


Reply via email to