A multiple-path TCP congestion control algorithm based on subflow correlation matrix

Abstract

Due to the connection with multipath transport control protocol (MPTCP) cannot utilize network resources sufficiently over the unshared bottleneck link, this paper proposes a congestion control algorithm based on the subflow correlation matrix (established by packet loss and RTT information) which is used to detect the status of bottleneck link among subflows by correlation coefficient. Then it performs different congestion control strategies over shared and unshared bottleneck links by adopting fairness definition based on bottleneck link. The results show that the proposed subflow correlation matrix can be used to detect shared bottleneck subflows more accurate and improve overall throughput of MPTCP connection.