Network Specification Mining With High Fidelity, Scalability, and Readability

作者：Ning Kang, Peng Zhang, Hao Li, Sisi Wen, Chaoyang Ji, Yongqiang Yang · 发表于：IEEE Transactions on Networking · 年份：2025 · DOI：10.1109/ton.2025.3607440 · 被引用次数：1 · 研究领域：Software System Performance and Reliability、Software Testing and Debugging Techniques

Network specification, which describes what an existing network is designed for, can help operators better understand and manage their networks, and is a critical pre-condition for network verification and synthesis tools to work. Existing tools for specification mining either cannot scale to large networks, or scale by sacrificing fidelity. Moreover, the specification contains a huge number of low-level intents (e.g., tens of thousands of pairwise reachability), making it hard for operators to read. To this end, this paper presentsNetMiner, which can mine specification from network configurations, with high scalability, fidelity, and easier to read. The key idea ofNetMineris to faithfully simulate the network routing and forwarding behaviors with control plane simulators and data plane verifiers, so as to achieve high fidelity. Meanwhile,NetMinerimproves the scalability by identifying relevant failure scenarios, and aggregating them to significantly reduce the number of needed simulations. Moreover,NetMinerclusters similar low-level intents into a high-level intent, to make the specification more concise and easier to read. Experiments using real configurations from a large cloud service provider and synthetic configurations show thatNetMinercan mine specification$10\times $faster, and reduce the number of intents by$100\times $, compared to state-of-the-art tools.