Re: What is the point of RAID?

On Wed, Nov 12, 2008 at 09:44:47AM -0500, Henning Follmann <hfollmann@xxxxxxxxxxxxxxx> was heard to say:
On Wed, Nov 12, 2008 at 08:53:46AM -0500, Jeff Soules wrote:
On Wed, Nov 12, 2008 at 3:44 AM, lee <lee@xxxxxxxxxxxxxxx> wrote:
Do you mean it is more likely that any one drive in the array fails when
you have more drives, or do you mean that it is more likely for a drive
in the array to fail when you have more drives? If drives fail more
often when being used in an array with more drives, what makes them
fail more often under those conditions?

It's purely a statistical property, not related to being in a RAID
array. But if there's (say) a 5% chance for a given drive to fail on
a given day, there's a 95% chance it won't fail.
If you have two drives, the chance *both* won't fail is the chance of
one not failing, times the chance of the other not failing -- 95%
times 95%, or 90.25%.

With 24, the chance of all the drives not failing is .95^24 or 29.2%.

Of course I just made the rates up, the survival chances of individual
drives are higher. But logic holds; the more drives you're watching,
the more lucky you'd have to be for none of them to be a dud.

Jeff,
you math is off - way off.

P(one fails) != 5/100

P(two drives fail at the same time) = P(one fails) * P(one fails)

Yes, but he was calculating

P(drive 1 fails OR drive 2 fails) =
1 - P(neither drive 1 NOR drive 2 fails) =
1 - P(drive 1 doesn't fail) * P(drive 2 doesn't fail)

and he already said he made the rates up for illustrative purposes.
Besides, the failure rate is meaningless unless we know the interval --
over a sufficiently long time, I expect that drives will have a failure
rate of 5%.

Daniel


--
To UNSUBSCRIBE, email to debian-user-REQUEST@xxxxxxxxxxxxxxxx
with a subject of "unsubscribe". Trouble? Contact listmaster@xxxxxxxxxxxxxxxx

Relevant Pages

  • Re: You Cant Spell "Stupid..."
    ... fail one of the two disks in my RAID1 array was when the box locked up ... > PATHETIC ONES I WOULD'VE GOTTEN IF I'D PLUGGED THE DRIVES STRAIGHT ... > INTO THE FUCKING MOTHERBOARD & INSTALLED XP SOFTWARE RAID 5 ON THEM, ...
    (alt.sysadmin.recovery)
  • Re: RAID 5 drive replacement schedule
    ... The real answer to your question is that it is very, very improbable that all of the drives in the array will fail at the same time. ... Netragard Whitepaper Downloads: ...
    (Security-Basics)
  • Re: Click of death (Quantum XP32150 takes a powder)
    ... week in an array. ... The drive *might* be good in a single-drive environment with optimizes / forced cooling, but matter of fact is, that it tends to fail in arrays. ... They particularly disliked the MCA Raid controllers, but tend to fail as well with the Streaming Raid PCI. ... From the 6 drives I had connected to the one in the S-520 *none* is working any longer. ...
    (comp.sys.ibm.ps2.hardware)
  • Re: Information Store service terminated unexpectedly
    ... "But why would it fail only during the ... The IO load on an undersized array of a non-optimal type. ... The RAID 5 write penalty is just too high. ... the performance of a mirror (2 drives), you'll need 4 or more drives in RAID ...
    (microsoft.public.exchange.admin)
  • Re: XP-VISTA Ultimate upgrade UAC Problem
    ... Should that fail, ... Attempt to turn off System Restore on all drives, run sfc/scannow from an elevated prompt. ... I think it is an issue with UAC / security. ...
    (microsoft.public.windows.vista.security)