BELIA: bandwidth estimate‐based link increase algorithm for MPTCP

Abstract

As a set of extensions to the regular Transmission Control Protocol (TCP), Multipath TCP (MPTCP) enables the multi-homed devices to access one data stream through multiple paths. MPTCP congestion window control is the crucial part of maintaining the fair network resource allocation, especially in the coexist scene of MPTCP applications and regular TCP applications. However, the current MPTCP congestion window control is facing two major problems: (i) the lack of the lost-packet detections, and (ii) the lack of the analysis mechanisms on the through-load relationships. These two problems combined may affect the accuracy of the congestion window allocations and the performance of MPTCP. In this study, the authors propose a bandwidth estimate-based link increase algorithm (BELIA). With the analysis of the link congestion occurrence mechanisms, they utilise a batch filter to estimate the real bandwidth of the links and define the link condition set C r to demonstrate the possibility of the congestion occurrence. For every loss, BELIA decides whether to cut window size off referring to the bandwidth estimates and C r . Hence, they proved that BELIA satisfies three design goals of MPTCP. Moreover, the authors' simulations show that BELIA performs better than some typical MPTCP congestion control algorithms under the same circumstances.