Haven't read back through the whole thread, but I'm
going to comment anyway.
SMART isn't very smart. If you plan on relying on this
metric alone to keep your data safe, then you are
going to be losing data all too frequently. It's a
well known fact in the industry that SMART doesn't
really work a *large* percentage of the time. If
anyone wants to challenge me on that, I'll google it
and dig up some statistics and white papers and hard
evidence. Unfortunately, the best hard evidence is
hidden behind the closed doors of the makers. I've had
drives last for decades and others 1 day. Fast/low,
cheap/expensive all the same. Also don't think that
going out and buying an expensive SCSI will shield you
from this. It's been proven that SCSI/IDE/SATA are all
pretty much equivalent, and that buying an expensive
one isn't going to boost your longevity either. Want
reliable harddrives, then your best bet is to buy a
disk array from one of the top packagers. But really
the best bet is to buy a pair or three, do frequent
backups, don't run 24/7, occasionally mirror to a
spare.
As far as stress testing HDs, I think that's
foolhardy. It's a good way to destroy HDs. I used to
do this for a living for a company that made blood
analsying equipment. Burn the drives in and stress
them over a period of weeks, discard the ones that
failed, and installed the ones that didn't. This
weeded out the ones from bad lots, but I can't help
but wonder if we didn't reduce the life of those
"good" drives by doing this. Drives were different
back then too, not as dense and a whole lot more
expensive (oh yeah and less storage).
--- Oren Beck <orenbeck(a)gmail.com> wrote:
> In regard to the concept of drive degradation being
> "hidden" by controllers.
> It seems to be a needful evil for the current
> applied art and science of
> rotating magnetic media at these price points. Yes,
> I suspect we "could"
> make media, heads, positioning methods et all that
> would be "true" non
> contact under most practical conditions.
>
> The game is a paraphrasing of an older tech rubric-
> Cheap- Fast- Stable=pick
> any 2.
>
> So the SMART ancestral concept was actually
> devised as an end run around
> those rules.
P.S. SMART was really just a marketing tool.