Uploaded June 2026 | Updated September 2026, 3 weeks ago
PolicyCache: Intra-flow Learning in Congestion Control
Han Tian, Han Wang, and Wenbo Li, University of Science and Technology of China; Xudong Liao, Decang Sun, and Wenxue Li, Hong Kong University of Science and Technology; Donghui Chen, Bin Huang, and Senbo Fu, Huawei Technologies Co., Ltd.; Junxue Zhang, University of Science and Technology of China; Dian Shen, Southeast University; Kai Chen, Hong Kong University of Science and Technology
TCP congestion control (CC) schemes must balance fast responsiveness, adaptability to diverse network conditions, and low computational overhead. Existing approaches fall short: heuristic-based algorithms are lightweight but brittle, learning-based schemes provide high responsiveness yet struggle with generalization, and exploration-based methods adapt well but converge slowly. We present PolicyCache, the first CC algorithm based on intra-flow learning, where both training and execution of the policy are confined to a single flow. Unlike prior inter-flow learning, this paradigm avoids cross-environment generalization pitfalls while maintaining high responsiveness. PolicyCache leverages a lightweight, non-parametric tree-based model coupled with online exploration and dynamic model switching to enable rapid and robust adaptation. We provide convergence analysis of PolicyCache and have built a fully functional Linux prototype. Extensive evaluations demonstrate that PolicyCache consistently achieves high throughput, low latency, and fairness across diverse emulated and real-world networks, while incurring minimal overhead. These results establish intra-flow learning as a practical and effective new direction for congestion control.
View the full NSDI '26 program at usenix.org/conference/nsdi26/technical-sessions
PolicyCache: Intra-flow Learning in Congestion Control
Han Tian, Han Wang, and Wenbo Li, University of Science and Technology of China; Xudong Liao, Decang Sun, and Wenxue Li, Hong Kong University of Science and Technology; Donghui Chen, Bin Huang, and Senbo Fu, Huawei Technologies Co., Ltd.; Junxue Zhang, University of Science and Technology of China; Dian Shen, Southeast University; Kai Chen, Hong Kong University of Science and Technology
TCP congestion control (CC) schemes must balance fast responsiveness, adaptability to diverse network conditions, and low computational overhead. Existing approaches fall short: heuristic-based algorithms are lightweight but brittle, learning-based schemes provide high responsiveness yet struggle with generalization, and exploration-based methods adapt well but converge slowly. We present PolicyCache, the first CC algorithm based on intra-flow learning, where both training and execution of the policy are confined to a single flow. Unlike prior inter-flow learning, this paradigm avoids cross-environment generalization pitfalls while maintaining high responsiveness. PolicyCache leverages a lightweight, non-parametric tree-based model coupled with online exploration and dynamic model switching to enable rapid and robust adaptation. We provide convergence analysis of PolicyCache and have built a fully functional Linux prototype. Extensive evaluations demonstrate that PolicyCache consistently achieves high throughput, low latency, and fairness across diverse emulated and real-world networks, while incurring minimal overhead. These results establish intra-flow learning as a practical and effective new direction for congestion control.
View the full NSDI '26 program at usenix.org/conference/nsdi26/technical-sessions
