I'm a fan of the "multiple copies" route, and my data gets subjected to different levels of redundancy according to how important I think it is. In order of priority:
- The bulk of the data I currently have (in GB) is not that important, and a total loss would be annoying but no more than that. For example, I have a lot of bits downloaded from the internet, and MP3s ripped from my own CDs, so I could always re-rip them. This data gets replicated between two disks in the same Windows PC, using Microsoft SyncToy.
- More important data gets replicated between two PCs, a desktop and a notebook, across the home network. I used to use SyncToy for this too, but the notebook is now running Linux, so I use a program called Unison. This is a bit tricky to set up, but extremely powerful once you do e.g. it can replicate files across the Internet using SSH Copy (SCP).
- For a small subset of really important files,I use a free service called Dropbox. This runs on both my machines, and continuously replicates changes between them and to the Dropbox server. It has rollback / recovery facilities via their website, so even if I damage or delete a file, and that gets replicated, I can roll back the changes to an older version of the file. It's like an offsite backup service that also does replication between machines. Plus, I replicate those files to a uSD card in my phone, so the subset includes files I might need on my phone.
I know it sounds complicated, but it all starts with a basic principle: that that any data you want to keep should exist in multiple copies. Now that disk space is so cheap, the limiting factors are time and network/internet speed, since two disks in the same machine is not much better than one if you want real data redundancy. It's a setup that has evolved over years, all batch-scripted and works pretty well unless I fiddle about too much.
