Beyond TCP-friendliness: a new paradigm for end-to-end congestion control

With the success of the Internet comes the deployment of an increasing number of applications that do not use TCP as a transport protocol. These applications can often improve their own performance by not being TCP-friendly and severely penalizing TCP streams. Also, designing these new applications to be TCP-friendly is often a difficult task. For these reasons, we propose a new paradigm for end-to-end congestion control (the FS paradigm) that relies on a Fair Scheduler network and assumes only selfish and non-collaborative end users. The flow isolation property of the FS paradigm is commonly agreed by the network community, however the lack of formalism of the FS paradigm hides fundamental properties. We rigorously define the properties of an ideal congestion control protocol and show that the FS paradigm allows to devise end-to-end congestion control protocols that meet almost all the properties of an ideal congestion control protocol. The FS paradigm is fully compatible with the TCP flows. Moreover, we show that the incremental deployment of the FS paradigm is feasible per ISP and leads to immediate benefits for the TCP flows since their mean bandwidth is increased by up to 25%. Our main contribution is the formal statement of the congestion control problem as a whole that allows to rigorously prove the validity of the FS paradigm.