Multipath congestion control with network assistance

Abstract

The use of multiple transport flows over distinct, if possible, paths, is a well-known technique for enhancing the performance and stability of data transfer. Multipath TCP (MPTCP), the most popular multipath transport protocol in-use, allows a single receiver to exploit multiple paths from a single sender. Nevertheless, MPTCP cannot fully exploit the potential gains of multipath connectivity, as it must fairly share resources with regular, single-path TCP, without knowing whether the available paths are distinct or share bottleneck links, due to IP’s design choices. We introduce a hybrid congestion control algorithm for multipath transport that enables higher bandwidth utilization compared to MPTCP, while remaining friendly to TCP-like flows. Our solution employs (i) Normalized Multiflow Congestion Control (NMCC), a novel end-to-end congestion control algorithm and (ii) an in-network module that exposes routing information to the end-users in order to support the greedy friendliness technique. The end-to-end NMCC is architecture-independent and can be seamlessly integrated with MPTCP. The in-network module has been implemented for the PSI Information-Centric Networking architecture, but it can also be integrated with Multi-Protocol Label Switching (MPLS) and Software Defined Networking (SDN). Using an actual protocol implementation deployed on our testbed, as well as on a comprehensive packet-level simulator, we obtain experimental results which demonstrate clear gains for our design in terms of throughput and friendliness to other flows.