[TriLUG] Dieing hard drive?

[Chip Turner]

Jason Tower <jason at cerient.net> writes:

>> Sorry, but I have to disagree on this one.  Hard drives have too many
>> moving parts to trust after being that old.  Especially if they've
>> sat around unused.  Time kills things with moving parts :)  Plus, old
>> drives may not provide SMART data, which helps you detect errors
>> sooner.
>
> as a mechanical engineer, i have to disagree with your disagreement :-)
>
> time doesn't kill things with moving parts, cycles kill things with 
> moving parts.  movement cycles, temperature cycles, whatever.  a hard 
> disk that's been sitting idle for a couple of years is no closer to 
> death now than it was two years ago, assuming it was stored reasonably.

Largely true; I didn't take your original mail to mean the old drives
you were speaking of hadn't been in service, though.  If the drives
were bought new and shelved, then that's not too bad; if they saw
three years of life before, though, then that certainly impacts their
longevity in whatever future use someone intends for them.  But time
can kill even inactive hard drives.  The lubricant used in drives can
and does undergo changes over time, in some cases clumping and
changing viscosity.  One of the reasons the "freezer trick" to revive
old, dead hard drives sometimes works is because the
expansion/contraction helps break the static friction of a slightly
congealed lubricant that was holding a drive's platters in place.

> modern disks have -substantially- higher storage densities than their 
> older siblings, which require much tighter physical tolerances in order 
> to operate.  as such, a trifling deviation that doesn't adversely 
> affect an older disk will render a new disk worthless.  you can only 
> push the cost/capacity envelope so far without affecting quality and 
> reliability, and unfortunately people keep voting (with their wallets) 
> for cheap rather than reliable.

While what you state is true about tolerances, densities, etc, I think
drawing the conclusion that it makes newer drives less reliable (even
if it fits anectdotal evidence any of us may have seen) isn't kosher.
I would argue many of today's drives are -more- reliable than they
once were.  Technology advances improve in many ways (remember having
to park hard drives?  also, error correction used today is more
advanced.  not to mention improvements in manufacturing in general).
I'd be curious to see what hard data the industry has on this kind of
thing, besides volunteer surveys (like below).

Personally I'd much rather have a new drive with a 3-5 year warranty
than hope an older drive (esp if it has seen many active hours of
service).  Storage Review has a nice user survey of drive reliability
with a ton of data that also offers interesting insight into the
reliability of particular manufacturers, brands, and models.

Just my own $0.02.  There's room for disagreement among friends :)

Chip

-- 
Chip Turner                   cturner at pattern.net