[svlug] archive media of choice for 10~100GB
shaeffer at neuralscape.com
Thu Jan 1 11:54:31 PST 2004
On Thu, Jan 01, 2004 at 01:24:58AM -0000, John Conover wrote:
> If the data is not accessed on a regular basis, either a HD or CDROM
> would be adequate. The dominant failure mechanism in HD is seek
> errors, (the mechanics of the head servos wear out, meaning a head
> crash,) so the MTBF is roughly proportional to cumulated seeks, (if
> the HD is unplugged and stored off line, it will last as long as the
> grease in the bearings-about a quarter of a century.) If CDROMs are
> not read, the MTBF is proportional to the stability of the
> medium-about a century.
These extrapolations are not valid. And whoever is publishing them must know
it. There is an entropy effect that is in play with both magnetic media and
solid state structures. I am not aware of the numbers for magnetic media,
but it is not nearly as long as folks might assume. I did attend an IEEE
magnetics meeting once where a leading researcher was discussing magnetic
tapes. (Thin film technology) He was joking about the entropy effect on
archived data, suggesting that 5 years was a reasonable limit to assume you
could "always" recover your data. This assumed room temperature storage.
Now, I do know that solid state structures decompose at a well known rate.
Virtually all solid state devices are designed to last 7 or 8 years, without
any entropy related failures. Again this assumes you store the electronics
at room temperatures. At that time, the probability of entropy related
failure becomes non zero, with a nonlinear rate of increased probability of
failure going forward. The 10-18 year window of time is where most solid
state devices will fail due to decomposition.
> As with stuff electronic, keep things cool and dry-the MTBF gets cut
> in half for every 10C increase in temperature.
Neuralscape, Palo Alto, Ca. 94306
shaeffer at neuralscape.com http://www.neuralscape.com
More information about the svlug