[geeks] Fwd: [IP] Interesting speculation on the tech behind gmail

[Phil Stracchino]

On Thu, Apr 08, 2004 at 03:40:52PM +0000, Lionel Peterson wrote:
> I don't follow your math... NOT saying you are wrong, but further insight
> appreciated.
> 
> If a drive has a MTBF of 100K hours, that comes out to what, about 400 24 hour
> days, right?

Um ...  no.  Try about *4000* 24 days.  About 4166 days, actually, which
is about 11 years.  At a rough approximation, assume 10k hours = 1 year.


> Now, some drives may fail immediately (DOA+1 hour), other drives
> may last for 200K hours (giving a 100K hour mean), correct?

Correct.


> Rather than a perfect linear failure rate (3 per day, for 400 days), the
> presumption is that the failures would tend to cluster around the 400 day
> mark, with very few failing in either the first 200 days or the last 200
> days... Right?

Actually, it tends to be a bathtub curve.  You tend to see a cluster of
initial "infant mortality" failures in the first few hours or days, then
very few failures for 10 years or so, then they start dropping off.

> I would suspect that if the MB relies on a moving part (fan) it is likely to
> fail when the fan fails. If it does not rely on moving parts, it *could* run
> forever (seemingly)...

No, because semiconductors fail too.  Electrons are moving parts, and
atomic-scale migration can eventually kill a transistor.  The smaller
they become, the more of a problem this is.  And then there's damage
from stray cosmic rays, ...  there's a variety of "wear" mechanisms that
affect nanoscale semiconductor devices.


-- 
 .*********  Fight Back!  It may not be just YOUR life at risk.  *********.
 : phil stracchino : unix ronin : renaissance man : mystic zen biker geek :
 :  alaric at caerllewys.net|phil-stracchino at earthlink.net|phil at novylen.net  :
 :   2000 CBR929RR, 1991 VFR750F3 (foully murdered), 1986 VF500F (sold)   :
 :    Linux Now!   ...Because friends don't let friends use Microsoft.    :