隨著現代網絡應用對高效數據傳輸的需求日益增長，如何在高丟包（Loss）與高延遲（Delay）的環境下維持穩定且高效的數據轉送，並避免較差線路影響整體傳輸效率，成為網絡架構設計中的關鍵挑戰。 為因應此挑戰，本研究提出 XCCTP（eXtensible Congestion-Control Transparent Proxy），一套建構於 Linux 核心空間的可調式 TCP Proxy 系統，專為高延遲、高丟包環境下的穩定傳輸所設計。XCCTP 系統設計基於以下四個核心原則：1. 緩衝式封包中繼（Buffer-Based Relay）：利用 TCP Socket 的讀寫緩衝區暫存封包，實現耐受丟包與傳輸不對稱的中繼機制。2. 資料快速橋接機制（Efficient Data Bridging Mechanism）：避免使用 User Space 的封包搬移，改以 Kernel 空間中快速橋接 RCV 與 SND socket，以減少延遲與系統呼叫開銷。3. 多核心執行緒排程架構（Kernel Thread Processing Architecture）：使用多執行緒與 Polling 模式處理高併發連線，提升封包處理吞吐量與系統可擴展性。4. 自適應壅塞控制（Adaptive Congestion Control）：依據網路延遲與丟包情況，動態切換 TCP 壅塞控制演算法（如 Cubic、BBR、Reno），實現最佳化傳輸行為。 綜合而言，本研究驗證了 XCCTP 在不同網絡條件下的最佳化策略，並提出動態緩衝區調整、Congestion Control機制切換等未來優化方向。本研究結果可作為未來高效能網絡代理與流量管理技術的重要參考，並為網絡架構的優化提供實證依據。 ;As modern network applications increasingly demand efficient data transmission, ensuring stable and high-performance delivery under high packet loss and latency conditions—while preventing poor links from degrading overall throughput—has become a critical challenge in network architecture design. To address this issue, this study proposes XCCTP (eXtensible Congestion-Control Transparent Proxy), a tunable TCP proxy system built within the Linux kernel space, specifically designed for reliable transmission in high-delay and high-loss environments. The XCCTP system is built upon four core principles: (1) Buffer-Based Relay: Utilizes TCP socket read/write buffers to temporarily store packets, enabling resilience against packet loss and transmission asymmetry. (2) Efficient Data Bridging Mechanism: Avoids unnecessary data copying between user and kernel space by directly bridging RCV and SND sockets in kernel space, reducing latency and system call overhead. (3) Kernel Thread Processing Architecture: Employs multi-threaded and polling-based processing to handle concurrent connections, improving throughput and system scalability. (4) Adaptive Congestion Control: Dynamically switches between TCP congestion control algorithms (e.g., Cubic, BBR, Reno) based on real-time network delay and loss conditions to achieve optimal transmission behavior. In summary, this work demonstrates XCCTP’s effectiveness across varying network conditions and presents strategies such as dynamic buffer tuning and adaptive congestion control as promising directions for further optimization. The results serve as a practical reference for future high-performance network proxy systems and provide empirical foundations for optimizing modern network architectures.
