Improving Inter-Protocol Fairness Based on Estimated Behavior of Competing Flows

Abstract

Although the openness of the Internet is an excellent characteristic, the quality of service that users gain is not always fair due to the characteristic. Several versions of congestion control algorithms are used on the Internet. In recent years, CUBIC and BBR have taken a large share as congestion control algorithms, but throughput fairness among flows is not maintained when CUBIC and BBR flows compete in a bottleneck link. One possible method of improving throughput fairness is to regulate BBR flows based on the estimation of the average packet sending rate of competing CUBIC flows. Existing analytical models for estimating the performance of CUBIC flows cannot be used for improving throughput fairness because certain parameters in the models cannot be accurately obtained by competing flows. In this paper, we first present a novel analytical model to estimate the packet sending rate of competing CUBIC flows. We then propose a modification of BBR, named BBR-FCA, to improve throughput fairness with competing CUBIC flows. We evaluate the accuracy of our analytical model and performance of BBR-FCA through extensive simulations. The results indicate that the fairness index among CUBIC and BBR flows can be improved by up to 95.4%.