I use a multistage backup process, for much this reason.
The lowest level is an rsync from my main work machine to a separate fileserver so 30 seconds after I save anything, it's replicated to a master copy on the server. That lets me get back something quickly should there be an immediate problem.
Then above that, the various systems all back up data from the live working directories to a separate backup drive on the machine in question every evening, using standard backup software doing incremental file addition. Those machines dump their new backups onto the main raid-5 backup server every night, and
that is backed up onto tape on a longer schedule, two copies being produced, one of which is then stored out of the house. So the end result is that if I have a disk failure in a given machine I can replace it and restore from a backup resident on that machine, if the entire machine bites it I can restore off the fileserver onto a new one, and if the house gets nuked I can restore from the offsite backup. If
that goes too, well, I'm likely gone along with it
It's not as comprehensive a disaster recovery method as a large company would probably have, but it works, it's been tested, and it wasn't all that expensive. I've got around 100TB of local storage, and backing that up to anything
other than tape would be rather expensive and awkward. Surplus LTO-5 tape drives are cheap and readily available, as are tapes, and while it's several generations old it's got enough capacity to be more than useful for the sort of load I have. One issue I had initially is that the tape drive is much faster than 1Gbit ethernet, and you really want to saturate the tape interface to get it streaming flat out rather than having to constantly stop and rewind. So I had to update the ethernet between the computers being backed up and the server running the tape drive to 2.5Gbit so I could get it to sustain 240 megabytes a second or so and keep the tape buffer filled. I've fine tuned the backup scripts so they pretty much run continuously, which took a lot of experimentation.
The bottleneck at the moment is the raid-5 drive array, actually. It won't read more than about 200MB/s sustained, with peaks higher than that, because it's running mechanical hard drives absolutely flat out and the processor isn't super fast either. In the longer run I want to build a new one around SSDs but they're still too expensive so that'll have to do for now.
There's also some critical stuff stored on a cloud service, but it's mostly a limited amount of things that would be bad to lose if all else went to shit, and might need to be accessed from somewhere else on short notice. Current work data, and of course my writing.
That is backed up all over the fucking place
Yes, I am serious about backups. A USB stick and xcopy won't cut it