Abstract: Multipath TCP (MPTCP) allows applications to transparently use all available network interfaces by creating a TCP subflow per interface. One critical component of MPTCP is the scheduler that decides which subflow to use for each packet. Existing schedulers typically use estimates of end-to-end path properties, such as delay and bandwidth, for making the scheduling decisions. In this work, we show that these scheduling decisions can be significantly improved by incorporating readily available local information from the device driver queues in the decision-making process. We propose QAware, a novel cross-layer approach for MPTCP scheduling. QAware combines end-to-end delay estimates with local queue buffer occupancy information and allows for a better and faster adaptation to the network conditions. This results in the more efficient use of the available resources and considerable gains in aggregate throughput. We present the design of QAware and evaluate its performance through simulations, and also through real experiments, comparing it to existing schedulers. Our results show that QAware performs significantly better than other available approaches for various use-cases and applications over internet and datacenter networks.
New publication: Tanya Shreedhar, Nitinder Mohan, Sanjit K. Kaul, Jussi Kangasharju, QAware: A Cross-Layer Approach to MPTCP Scheduling, In Proceedings of IFIP Networking 2018