zboss:
Someone should make a fire/flood/tornado/quake proof computer. case.. that way it could survive anywhere. I'm sure the military makes them!
I think the best bet so far is storing the files remotely on external and secure servers. You can even encrypt them so even if the provider gets broken into it wouldn't do them any good.
Fire is pretty hard to protect against. You do not actually need to 'burn up' a drive to destroy the data on it. In fact it does not even need to be scorched, all that needs to happen is the temp reaches the curie temprature which for iorn is about 1000K (1300f). The temp for iron oxide is not well documented, but may be below 500K (440 f).
CDs (and backup tapes) suffer damage at similar (or lower) temps. Tapes often fail because the acitate backing hardens and becomes brittle.
The only real solution is off site storage. Fire safes and containers are allowed to reach about 460 f, after 1 hour of fire (in a typical residential construction) and 'pass', if UL 1 hour rated. Unrated safes are often less robust. providing only 10 or 15 minute rateings. The fire safe mentioned earlier is rated as 1 hour 350 degrees which is pretty good if located away from heavy construction, which would retain high temps or conbustion for over 1 hour.
'Fire proof' consturction often results in hotter fires if it starts to burn. Combustable liquids and plastics often burn hotter than wook (about 650 f) and are outside the rating of the safe. Do not locate fire safes in a basement or any location where burnt or burning materials may collect.
Pelican cases are good for water protection, but not for shock (tornados) or fire (the case burns at moderate temps). Cases lined with foam pockets may provide some shock protection, at the cost of a more easily ignited case, with more fuel present.
HD drives may last 3 to 5 years, but are unlikely to last much longer. Running the drive hot (above 145 f) will reduce the drive life conserdably (perhaps to minutes). Voltage fluctuations also are problematic.
Many modern drives have S.M.A.R.T technology build in. This is a method for the drive to communicate internal conditions to the external computer system. S.M.A.R.T. errors include recovered retries (over some limit), high temps, low voltages, slow spin or drive starts and other bits. S.M.A.R.T. errors are significant and any drive getting them should be replaced. The only problem is getting access to them...
I use a 3Ware hardware raid controller card. It reports them to me via email or a built in web server. S.M.A.R.T. errors generally precede failure by 1 to 2 weeks (except in the face of catastrophic destruction). Hardware raid is better than software raid, in that you can reboot a system that has a boot drive failure with hardware raid, but can not with software raid (its really part of the driver or OS).
I have had 3 failures since I started using it (including a system board failure) and have not lost any data. MB raid does not survive a MB failure... S.M.A.R.T. errors alerted me to another drive failure before it happened. I swapped the drive and rebuilt the cluster, without loosing data.
NB: not all raid technologies provide any data security at all some are optimized to increase drive throughput/access speed only. I believe raid 1, 10 and X are best for small clusters, and 5 is OK for bigger ones.