You need to imagine: what happens if I can't access part of my data? Usually this could mean you go out of business! (You can't just rely on the hope that that small part that fails will not be the core data necessary for your company's operation, for example.)
You'll probably need:
- redundancy of data (ie, have your data on at least 2 separate piles of disks, or a version on disk and another on tapes, or a combination thereof)
- redundancy of location (ie, have a good off-site backup, in case you experience fire/theft/sabotage/flood/? on your primary location)
You can combine both, but off-site usually takes some time to retrieve (unless you also have off-site office, or access to it via some means, check with your off-site backup providers)
About the need for redundancy :
Hard Drives (actually, any storage medium) ages, and some have defects, and some gets bumped (for example, a little knock with a screwdriver or against another drive while storing them on a shelf could yeld enough G to scratch a disk's surface, and this will cause failure either now or within a short-to-not-long-enough timeframe). Which means: your backup media will fail, some within days, others within years, with no means to know when.
You can't know when something will go wrong, but you can assess the risks (not to have precise infos, but to get a sense of what can go wrong and why it's good to prevent it with redundancy and with regular transfer to other mediums before the prior mediums fails):
Imagine that you have similar disks and that you can say that the probability of 1 failing within 2 years is 5%
P(one disk fails within 2 years)=5%=0.05
P(that disk won't fail within 2 years)=0.95
P(you have n disk, 0 disk fails within 2 years)=(0.95)^n
P(at least 1 disk amongst n fails within 2 years)=1-(0.95)^n
if n=72 (you have 72 disks), P=1-0.95^72=1-0,024894281=0,975...=more than 97%
In english : if you have 72 disks, and if you can say that a single disk have a 5% probability of failing within 2 years, then you'll have more than 97% chance you have a disk failure within 2 years. This is not at all precise, but it should show you that things can go wrong, and faster than usually expected...
To "scare" you even more, you could say now that P(one disk fails within 20 years)=0.99 (20 years is a long time for a drive. And even if it works, the devices to read them may not be able to handle them)
P(that disk won't fail within 20 years)=0.01
P(no disk amongst n fails within 20 years)=(0.01)^n=0
P(at least one disk amongst n fails within 20 years)=1-(0.01)^n=1=100% (or so close to 100% that it doesn't matter)
And in addition to all this: if you buy all the drives at the same vendor, P(one of those disk fails after another failed) is very high, much higher than the P(one disk fails) : disks are built similarly, so if one fails, chances that the others in the same batch will fail around the same time is high (or very high) !
So I recommend:
- backup on at least 2 separate things (2 x 72 hd, from 2 separate vendors ? or hd and tapes? ).
- Have at least one (if possible, yet another one) backup stored in another town. Update this every n days (weekly at least?) so you can go back to the backup time in case the primary storage all fails (ie: you'll be able to go back to the data of last week in case the primary storage fails this week)
- check backups regularly (I mean, do check them, reading the whole thing byte per byte at least once a month. A good time to also copy to yet-another-backup media too.)
- ensure at least 2 people know about this and can help retrieving the data when the external backup is needed.