4 ms·
MPTCP and MPRTP to the rescue... these also have inbuilt algorithms to pick the best interface. http://tools.ietf.org/html/draft-singh-avtcore-mprtp http://too
by vr000m 13y ago
MPTCP and MPRTP to the rescue... these also have inbuilt algorithms to pick the best interface.
http://tools.ietf.org/html/draft-singh-avtcore-mprtp http://tools.ietf.org/html/draft-singh-avtcore-mprtp (for skype and multimedia)
https://tools.ietf.org/wg/mptcp/ https://tools.ietf.org/wg/mptcp/ (for everything else, even Siri)
- windexh8er 13y agoIt was interesting to see everyone rage about how amazing Apple was to be leveraging MPTCP in the latest version of iOS. However most don't realize the implications and risks that these types of things bring to generalized use cases and operational models of networking and paid for data buckets. Regardless it looks like these guys are just taking the Commotion firmware and re-branding it as a service. It also doesn't solve the local node problem. Most people don't understand how cable (HFC), fiber and copper (DSL) are built out to the neighborhood or 'node'. It's all oversubscribed and shared. So this type of solution really does nothing for the local congestion in a node. To me, this is a pure marketing gimmick.
- vr000m 13y agoAt the networking stack level, it depends on how one implements congestion control. AFAIR, in the case of Apple/MPTCP: they were sending the same packet on both the interfaces and hoping for the best delay (a really bad idea) unless you are in a managed network else you are just creating unnecessary congestion. OTOH, if the last mile is oversubscribed, there is very little one can do.