• A Great friend to the HardForum with a great kid that he is trying to get a scholorship to continue his schooling. Please give hime a vote! Only 24 hours left! Thanks.
    If you have an VOTE FOR KEENAN!

RAID question.

piako

[H]ard|Gawd
Joined
Apr 5, 2006
Messages
1,613
How often does a RAID-5 array (small size ~6 disks) fail during rebuild (another disk dies during rebuild = total data loss)?
 
How often does a RAID-5 array (small size ~6 disks) fail during rebuild (another disk dies during rebuild = total data loss)?

If you assume that the chance of every individual disk is 33% per year, then the chance of fail is 4.5716763e-5 per hour.

If your rebuild takes 5 hours, then the chance (per disk) is 2.2858381e-4.

If you have an array with 6 disks live + one hot spare (or one inserted right after failure), then there are 5 live disks, but the new spare can fail, too, during those 5 hours, so your risk applies to 6 disks.

That puts the total risk at 0.001370728 or roughly 1/1000.

%%

The problem is, of course, that disks often break not out of random chance, but that some external event damaged them, such as:
- power fluctuations in A/C support
- power fluctuations due to broken PSU
- temperature fluctuations
- fan failure and overheat
- somebody kicks the case
- other causes of mechanical impact such as loose screw on disk cage or earthquake

Then the chance of a second drive failure right after the first is several orders of magnitude higher.

But of course, if that happens the chance of a triple failure is also high, which makes a three-disk raid-0 or a raid-6 almost equally useless.
 
ty uOpt, those probabilities seem pretty low. Care to share the equations?
 
ty uOpt, those probabilities seem pretty low. Care to share the equations?

This is normal probably multiplication. I didn't keep the cut'n'paste from the above run.

Here's it backwards for proof, single disk over 5 hours converted to pre-year chance of failure:
Code:
(defun prob* (prob n)
  (- 1.0 (expt (- 1.0 prob) n)))

:user - Yes, Master? (cra:prob* 2.2858381e-4 (/ (* 24 365) 5))
0.3300315
:user - Yes, Master?
 
Back
Top