Some of the Internet may be Heading Towards BBR Dominance: An Experimental Study

Low latency TCP congestion control (CC) is a key enabler of delay-sensitive applications such as cloud gaming, remote driving, and virtual/augmented reality(VR/AR). In recent years, TCP BBR has emerged as a popular choice for low latency CC, with it already having widespread adoption. However, a critical question that needs to be studied is how BBR will coexist with the current most dominant CC protocol on the internet, i.e., TCP Cubic. A recently published study proposed a mathematical model for the coexistence of TCP BBR and TCP Cubic. This study on BBR dominance predicted that the internet will reach a stable mixed distribution of BBR and Cubic flows, resulting in a Nash Equilibrium where no traffic will have any incentive to switch between BBR and Cubic. This work presents a fully reproducible experiment on an open-access networking testbed that replicates the major experiments from the BBR dominance study. In addition to reproducing the original results, we further show that the predictions of the proposed model do not hold up in specific scenarios that are highly relevant to today's low-latency networks. Our work motivates future work to improve the analysis of BBR and Cubic's coexistence. We also share all artifacts necessary to reproduce our experiment to help other researchers easily replicate this study and to further use it as a baseline for their research on low latency CC.