So I'm reading up on Amazon Web Services; S3 is the current target.
What's been tripping me up with S3 is that I keep wanting to treat it like a remote filesystem... I was having trouble finding a good "model" to use for backups, but I think I've settled on using something like rsync as the answer. Put stuff you want to backup into a folder, specify a bucket, and run a program that will periodically check the stats on all files in the dir and upload anything new or that has changed up to S3.
Gonna check to see if anyone's already put something together for it yet, but if not will probably whip up a utility based on the Right Scale AWS ruby libs.