A flurry of email—partially caused by my glossing over things a bit to glibly in my last post (since corrected)—indicates that it’s time to say this more clearly again:
RAID is not backup.
If you’re thinking about getting or have bought a RAID or a NAS so that you don’t have to make backups, reconsider. Reconsider now. While being able to mitigate a single disk failure is a strong point of a good array, they’re still subject to catastrophic failure of multiple disks or some other hardware component. They also can’t help you with accidental or purposeful deletion of data, lightning bolts, or a beverage spill.
For planning purposes, you should think of an array as a way to get a bigger or faster volume. And that volume needs to be backed up. David Magda wrote a great definition in an email exchange:
A backup is a coherent copy of the data on independent media
Back in the days of old, this was a pile of floppy drives that could recreate data on a hard drive. Then it became a smaller pile of DVDs. If you were big time, you used tape drives. These days, it’s easy to use a single disk to backup onto if your dataset is less than 4TB.
He went on to add that you have to be able to use that copy to restore from in order for it to be considered any good. Agreed.