Flex-Cubic: A Runtime-Adaptive Loss-Tolerant TCP Cubic

• Pedro P. P. Filho UFES / IFES
• Moises R. N. Ribeiro UFES
• Magnos Martinello UFES
• Guilherme F. Miguel UFES

DOI: https://doi.org/10.5753/sbrc.2026.19741

Abstract

Traditional TCP still lacks the ability to differentiate between losses primarily caused by congestion from those caused by physical layer errors. This may severely impair performance, especially in data-intensive science applications over high-capacity and long-distance dynamically reconfigurable transparent optical networks. This work proposes and experimentally implements a variant of TCP Cubic designed for reconfigurable networks to turn transport layer tolerant to non-congestion losses and variable RTT exploiting available bandwidth more efficiently. This is done by designing a congestion window (cwnd) reduction mechanism that conditions loss reactions on evidence of congestion, given by RTT measurements. In addition, Flex-Cubic aims to supports dynamic parameter tuning and higher-precision timing, resulting in greater stability and improved bandwidth utilization. Thus, eBPF has been used as a platform for TCP congestion control algorithm (CCA) implementation, enabling new algorithms to be loaded into the kernel via JIT at runtime, without recompilation. Through struct ops and maps, eBPF allows per-flow instrumentation and dynamically CCA adaptation.

References

Antelmi, A. and Carlini, E. (2026). Large-scale hpc approaches and applications on highly distributed platforms.

Borges, E. S., Ribeiro, M. R., Guimaraes, R. S., Xavier, B. M., Pecolo, P., Dominicini, C. K., Martinello, M., and Rufini, M. (2024). Fso-based reconfigurable optical networks: Source routing for decoupling multilayer te.

Chen, X. et al. (2020). Measuring tcp round-trip time in the data plane. In Proceedings of the Workshop on Secure Programmable Network Infrastructure, SPIN ’20, page 35–41, New York, NY, USA. Association for Computing Machinery.

Chen, Z., Meng, Q., Lao, C., Liu, Y., Ren, F., Yu, M., and Zhou, Y. (2025). {eTran}: Extensible kernel transport with {eBPF}. In 22nd USENIX Symposium on Networked Systems Design and Implementation (NSDI 25), pages 407–425.

Chou, J. and Chung, W.-C. (2024). Cloud computing and high performance computing (hpc) advances for next generation internet.

Ha, S., Rhee, I., and Xu, L. (2008). Cubic: A new tcp-friendly high-speed tcp variant. SIGOPS Oper. Syst. Rev., 42(5):64–74.

Hinz, J.-T., Addanki, V., Györgyi, C., Jepsen, T., and Schmid, S. (2023). Tcp’s third eye: Leveraging ebpf for telemetry-powered congestion control. In Proceedings of the 1st Workshop on eBPF and Kernel Extensions, pages 1–7.

Jadin, M. et al. (2022). Leveraging ebpf to make tcp path-aware. IEEE Transactions on Network and Service Management, 19(3):2827–2838.

Lai, Z. et al. (2025). Leocc: Making internet congestion control robust to leo satellite dynamics. In Proceedings of the ACM SIGCOMM 2025 Conference, pages 129–146.

Magnani, S., Risso, F., and Siracusa, D. (2022). A control plane enabling automated and fully adaptive network traffic monitoring with ebpf. IEEE Access, 10:90778–90791.

Mathis, M., Semke, J., Mahdavi, J., and Ott, T. (1997). The macroscopic behavior of the tcp congestion avoidance algorithm. SIGCOMM Comput. Commun. Rev., 27(3):67–82.

Miano, S., Bertrone, M., Risso, F., Tumolo, M., and Bernal, M. V. (2018). Creating complex network services with ebpf: Experience and lessons learned. In 2018 IEEE 19th International Conference on High Performance Switching and Routing (HPSR), pages 1–8. IEEE.

Pan, W., Xu, Y., Wang, C., and Wu, J. (2024). An ebpf-empowered congestion control system with delay requirements. In 2024 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pages 3875–3880. IEEE.

Pecolo F, P. P., Ribeiro, M. R. N., Borges, E. S., and Martinello, M. (2025). On the interplay of congestion and osnr degradation for tcp over optical networks. In 2025 SB-MO/IEEE MTT-S International Microwave and Optoelectronics Conference (IMOC), pages 650–654.

Pedro Filho, P., Togneri, A. P., Ribeiro, M. R., Segatto, M. E., Borges, E. S., and Martinello, M. (2025). Tcp/ipodwdm: A case for adaptive congestion control in optical networks under physical layer impairments. In Workshop de Pesquisa Experimental da Internet do Futuro (WPEIF), pages 25–32. SBC.

Song, L. and Li, J. (2024). ebpf: Pioneering kernel programmability and system observability-past, present, and future insights. In 2024 3rd International Conference on Artificial Intelligence and Computer Information Technology (AICIT), pages 1–10. IEEE.

Wibowo E.and Bi, B. (2025). Evaluating ebpf as a platform for congestion control algorithm implementation. [link].

Yang, S., Tang, Y., Pan, W., Wang, H., Rong, D., and Zhang, Z. (2023). Optimization of bbr congestion control algorithm based on pacing gain model. Sensors, 23(9):4431.

Zadeh, S. A. et al. (2023). On augmenting tcp/ip stack via ebpf. In Proceedings of the 1st Workshop on eBPF and Kernel Extensions, pages 15–20.

Zanotelli, V. F., Pontes, E. C., Martinello, M., Ros-Giralt, J., Borges, E. S., Comarela, G., Ribeiro, M. R. N., and Newman, H. (2025). Transport efficiency for data-intensive science: Deployment experiences and bottleneck analysis. Annals of Telecommunications, 80(9):793–805.