More general questions about HD MTBF

[Dick Dunn]

Following on from where cpcahil at virtech.uucp (Conor P. Cahill) writes:
> dnb at fltk.UUCP (David Buonomo) writes:
> >drive.  I still have reservations about this decision due to this drive
> >having a 20,000 hour MTBF.  Should I be concerned?
> 
> This sounds low.  If you run your system 24 hrs a day, you can expect to
> have a drive failure within 3 years (1,000 days or so)...

Does anyone really know about MTBF ratings on these small drives?  I'd like
to know whether there's more than one approach being used for rating MTBF or
some drives are really that much more reliable than others.

Among the current Conner drives, the MTBF quoted on spec sheets goes like
this:
	80 Mb	50,000 hr
	100 Mb	30,000 hr
	200 Mb	20,000 hr

This factor of 2.5 over the range of these drives bothers me...but even
more, the Maxtor 7080A is a roughly 80 Mb drive spec'ed at 150,000 hr MTBF.
Is this factor of 3 real, or is it just due to some difference in analysis
of MTBF?  The 17-year MTBF of the Maxtor drive sounds like the sort of
margin I'd really like.

On the other hand, I don't much like the just-over-2-year MTBF of the large
Conner.  I'd like to have the MTBF be a lot larger than what I expect the
useful life to be--because MTBF is saying you're uncomfortably likely to
see a failure by that time.

> Three years can be considered a long time in the pc-unix industry (you
> will probably replace the drive by than anyway) so it may be no big deal.

Not necessarily...I think you're more likely to move the equipment down to
a system where performance is not so critical.  But with depreciation for
computer equipment at five years (ridiculous, but that's reality), it's
hard to justify just tossing it.  I've got a drive at home with > 35000
hours on it.
-- 
Dick Dunn     rcd at ico.isc.com -or- ico!rcd       Boulder, CO   (303)449-2870
   ...Worst-case analysis must never begin with "No one would ever want..."