Receiver-Controlled Congestion Control for End-to-End Transport over the Internet Protocol: Implications and Performance Evaluation of a NewReno Adaptation

Over the years, TCP congestion control research has advanced through extensive algorithmic refinements. In a parallel line of research, transport architecture has been reconsidered from the perspective of sender-state reduction, receiver assistance, and full receiver control, with the latter concentrated largely in Information-Centric Networking architectures. Building on these findings, in this study, we investigate how an established TCP congestion control algorithm behaves when adapted for full receiver control in a conventional end-to-end byte-stream transport over IP. We evaluate Rx-NewReno, an adaptation implemented over a receiver-driven ARQ mechanism established in prior work and operating over UDP. The established NewReno AIMD’s rules and well-understood recovery behavior provide a clear basis for evaluating the effects of relocating the congestion control loop. Using ns-3, we conducted three experiments assessing goodput, fairness, and recovery behavior, with each lossy configuration evaluated over 30 independent runs and the results reported using 95% confidence intervals. Under forward-path loss, Rx-NewReno achieves goodput closely matching that of SACK-enabled TCP NewReno and follows the same loss-dependent trend as QUIC-NewReno while recording no spurious retransmissions. At a 5% packet error rate, it averages 1.7 RTO expirations, compared with 9.9 and 23.1 for TCP NewReno with and without SACK, respectively. The average Jain’s fairness index was 0.996 in self-pairing, 0.996 against TCP NewReno, and 0.986 against TCP CUBIC. The reverse-path results further show that Rx-NewReno’s performance depends on the reverse-path request rate and loss handling in the underlying reliability layer. Overall, the results show that Rx-NewReno preserves NewReno’s increase and reduction behavior while re-expressing its congestion control and recovery procedures using receiver-side state and observations.