On 27 Nov 2010 at 15:42, Eric Smith wrote:
The MTBF is a statistical measure or prediction of the
behavior of a
population of drives (not a single drive!) during the design life
(typically five years). The MTBF rating of drives you purchased 15-20
years ago means absolutely NOTHING at this late date. If the drives
keep working, that's great, but you don't have any particular basis
for expecting them to do so for any length of time.
For those wondering about "bathtub curves" see the following very
good paper:
http://db.usenix.org/events/fast07/tech/schroeder/schroeder_html/index
.html
Which essentially sheds a lot of light on the MTTF issue (which
appears to be part snake oil). The far end of the bathtub curve is
characterized as the "wearout" end, just as it would be with, say, an
automobile. It makes a lot of difference if you drive a 1957 Chevy
Belair with 5,000 miles on the clock and one with 750,000 miles.
True, components such as tires will fail due to their nature, but
which would you rather have?
So time spent spinning is quite germane in a disk drive.
But as I mentioned, MTTF on hard disks largely seems to be a
statistical trick not particularly observed by reality--and the
Schroeder-Gibson paper.
New drive or old, you pays your money and takes your chances. But my
money will be on the low-mileage units.
But the OP speculated that SCSI drives will become unobtainium. My
point was that the possibility is very remote, at least during my
lifetime--and that not all drives have been run to death.
Now, wanting a cheap substitute is a different matter--and we should
be candid about this. I paid about $1000 for 330MB CDC 5.25" SCSI
drive sometime around 1989 and I thought I was getting a heckuva
deal. If there were a strong demand for the same drive at the same
price today (adjusted for inflation), I'd imagine that an
enterprising soul could find a ready supply of working drives, even
if they had to be remanufactured.
--Chuck