The Impact of Reverse-Path Bandwidth Asymmetry on TCP Congestion Control: A Telemedicine-Motivated ns-3 Study

Abstract

Telemedicine services in underserved rural regions often rely on asymmetric access technologies. These include Low Earth Orbit (LEO) satellite broadband and 5G Fixed Wireless Access (FWA), which prioritize downstream over upstream bandwidth. This asymmetry can affect reliable transport protocols such as the Transmission Control Protocol (TCP), on which most telemedicine services depend. The impact of bandwidth asymmetry on TCP was studied extensively in the late 1990s and early 2000s, and recent studies have revisited it for modern congestion control algorithms (CCAs), mainly within specific access technologies. In this paper, we revisit the bandwidth asymmetry problem in TCP, isolating reverse-path bandwidth asymmetry as a single controlled variable across three CCAs, namely NewReno, CUBIC, and BBRv1, in the ns-3 simulator (ns-3.47). We also isolate the impact of asymmetry on the ACK-clock mechanism itself, using a fixed-window approach that decouples ACK-clock dynamics from CCA behavior. Our results show that the fixed-window baseline converges to the ACK-clock ceiling imposed by the reverse path, while the adaptive algorithms diverge sharply. We also find that CUBIC's time-based window growth makes it achieve the highest goodput under severe asymmetry, sustaining an average goodput of 7.75 Mbps on the 10 Mbps forward path at a 300:1 ratio. On the other hand, BBRv1 converges to the ACK-clock ceiling without a single packet loss, achieving the lowest average goodput among all configurations at 1.72 Mbps. The findings indicate that TCP performance in asymmetric access networks depends strongly on how each algorithm consumes the ACK stream.