Interesting question! I don't think you can save storage, because you may need to distribute every file to every possible site depending on what box has more traffic available. You can distribute files using whatever you want (FTP, HTTP, RSYNC), preferably something that can resume interrupted transfers, to avoid waste of traffic.
On the other side, you must account also interrupted users' transfers, and this may be quite complex because you have to analyze logs, update traffic accordingly and give this informations back to your central repository.

There are a couple of projects that you may find interesting: BitTorrent and Open Content Network.

Ciao, Valerio