The Samba-Bugzilla – Bug 9624
rsync p2p extensions
Last modified: 2013-02-01 05:06:02 UTC
Add p2p enhancements to rsync to allow allow the rsync client to source the file from multiple endpoints. The rsync server would track and nominate a list of valid mirrors.
(Incredibly) large internet archive sites are moving towards phasing out anonymous rsync and FTP in favour of BitTorrent P2P systems which offload the majority of the transfer load to other sites. With the use of torrents then P2P clients can also help with the traffic offloading. P2P also allows continued functionality in the cases where network traffic isn't available to the original server, due to network outages or attacks.
The torrent approach is being investigated now (see http://www.terasaur.org/browse/category/software ) and the issues being ironed out ( https://docs.google.com/a/ibiblio.org/document/d/1d2gJYsesalyXkYvrnY9AOornHYeGlwcLX1gjR8VMFlw/edit )
The problem is when torrents are working adequately the intent is to phase out rsync and FTP. This causes issues in the unix world with auto-mirroring from cron, access via a cli, fetching files using regexps, accessing torrents from behind corporate etc firewalls and reinventing the agility of rsync into the bittorrent clients.
rsync needs to extend its protocol to support more P2P functions.
- mirrors could nominate themselves when fetching content from the
central site, possibly with timeouts for caching their existance
- mirrors could download the mirror list using the P2P system and
re-share it to P2P clients
- clients attach to the server and negotiate P2P fetching. They
would then switch from classic rsync behaviour to P2P behaviour and
spread the traffic load across a number of mirrors, possibly client
peers in the future.
Searching google for "rsync p2p" returns a number of older projects which have attempted this. It's possible a lot of those efforts could be leveraged to reduce the development burden.
Adding these and other P2P enhancements will future proof rsync so that archive sites can utilize rsync more without worrying that excessive loads will continue on the central site. The main traffic transfer cost can be shared with sites around the world. rsync's value will remain high and those of us who rely on it will continue to be offered it.