Mean time between failures

nomadicus · 2003-10-03, 13:36

In discussing hardware reliability, I was looking at disk drives. Suppose a disk has a 300,000 hours MTBF (Mean Time Between Failures). If we have 100 disks on a system (assume equal usage), does that say the next failure of any single disk will be 300,000/100 or 3,000 hours?
What if the 100 disks are spread equally between 4 systems? Does that change the probabilities?
Do probability mathematics come into play? (of which I know very little).

xtreme2k · 2003-10-03, 17:51

http://www.storagereview.com/guide20.../specMTBF.html

This should help you understand MTBF. After you have read it you will see that lot of us has misconceptions about it.

PageFault · 2003-10-03, 23:20

Yes they do, possibly the most tortuous of mathematics when I was taking those courses. Because it is "mean" time before failure, you must account for the variances and sample sizes. Now, when you add more units, you have not only more chances of failure but also a higher probability of a premature failure ... consider building a 1000 disk array, its MTBF is probably zero ... because one of those suckers is DOA ... even when running, it wouldn't last long ...

Quote:

Originally posted by nomadicus
What if the 100 disks are spread equally between 4 systems? Does that change the probabilities?
Do probability mathematics come into play? (of which I know very little).

nomadicus · 2003-10-06, 16:35

So now I should consider infant mortality of a group of disks I get, the service life of the disks within that group, the operational MTBF (if I can get it), balanced along with the MTBF as a guideline toward understanding when a group of disks would become more prone to failure.

Things are never as simple as they seem.

Great pointer.
Thanks!

2003-10-03, 13:36	#1
nomadicus Jan 2003 North Carolina 2·3·41 Posts	Mean time between failures In discussing hardware reliability, I was looking at disk drives. Suppose a disk has a 300,000 hours MTBF (Mean Time Between Failures). If we have 100 disks on a system (assume equal usage), does that say the next failure of any single disk will be 300,000/100 or 3,000 hours? What if the 100 disks are spread equally between 4 systems? Does that change the probabilities? Do probability mathematics come into play? (of which I know very little).

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Time to End	davar55	Lounge	4	2013-02-23 02:40
How much time	Unregistered	Information & Answers	4	2008-12-20 21:00
Hardware Failures at different FFT lengths?	CADavis	Hardware	8	2005-11-29 10:29
Question about what failures "mean"	Semm	Hardware	7	2005-02-06 23:26
Complex, but deceptively simple question about Prime95 failures?	halcion	Software	6	2004-12-16 20:10

2003-10-03, 17:51	#2
xtreme2k Aug 2002 2×3×29 Posts	http://www.storagereview.com/guide20.../specMTBF.html This should help you understand MTBF. After you have read it you will see that lot of us has misconceptions about it.

2003-10-06, 16:35	#4
nomadicus Jan 2003 North Carolina 366₈ Posts	So now I should consider infant mortality of a group of disks I get, the service life of the disks within that group, the operational MTBF (if I can get it), balanced along with the MTBF as a guideline toward understanding when a group of disks would become more prone to failure. Things are never as simple as they seem. Great pointer. Thanks!