Disasters and Recovery

From: Philip.Belben_at_pgen.com <(Philip.Belben_at_pgen.com)>
Date: Mon Jan 18 11:40:44 1999

>> I find this proposition very doubtful at best, although I don't have any
>> solid figures for EPROM data retention. You seem to be suggesting a
>> linear relationship between elapsed time and the number of bits that
remain
>> good. The actual curve is an inverse exponential (if memory serves).
>
> Wow! That would imply some sort of self-catalyzed reaction, and I'd be
> very interested in hearing how that works!


Eh? No it wouldn't. Never mind the mechanism, do a thought experiment.
Sorry Hans, I mean a Gedankenexperiment :-)

Assume you have a memory technology for which you expect half the bits to
get zapped in 20 years.

In another 20 years, half the bits will get zapped - but it won't be
selectively all the others. On average, expect half the remaining (intact)
bits to get zapped, and half the already zapped bits to get zapped. So
you're left with a quarter. And so on.

20 years is thus the half-life of an exponential decay.

Now with technologies like EPROM it's a bit more complicated, because the
zapping of bits is not a random (Poisson) process but depends on a decay of
charge within the device. But you get the idea (I hope).

Philip.
Received on Mon Jan 18 1999 - 11:40:44 GMT