I'm again evaluating the use of Amazon S3 for storing backups of images. I've dabbled with it before and liked it quite a bit, but never got into the habit of using it regularly. I guess the idea of incremental pricing concerns me.
I'm less concerned about that nowadays as I've been using reverse proxies, compression and caching to reduce bandwidth consumption, as well as combining shared resources into a single domain instead of serving the same data over different domains.
My plan is to use Amazon S3 as a static image backup and fail-over, not a direct provider of service. I buy bandwidth from other network providers and use it for a variety of services beyond REST / HTTP content deliver. Besides, the bandwidth on S3 can get pricey pretty quick!
I've chosen a few tools to use and made some preliminary notes over at Docunext: