RIPE Atlas 的日常：网络测量中的运维洞察与应用 (Day in the Life of RIPE Atlas: Operational Insights and Applications in Network Measurements)

Network measurement platforms are increasingly popular among researchers and operators alike due to their distributed nature, simplifying measuring the remote parts of the Internet. RIPE Atlas boasts over 12.9K vantage points in 178 countries worldwide and serves as a vital tool for analyzing anycast deployment, network latency, and topology, to name a few. Despite generating over a terabyte of measurement results per day, there is limited understanding of the underlying processes. This paper delves into one day in the life of RIPE Atlas, encompassing 50.9K unique measurements and over 1.3 billion results. While most daily measurements are user-defined, it is built-ins and anchor meshes that account for 89% of produced results. We extensively examine how different probes and measurements contribute to the daily operations of RIPE Atlas and consider any bias they may introduce. Furthermore, we demonstrate how existing measurements can be leveraged to investigate censorship, traceroute symmetry, and the usage of reserved address blocks, among others. Finally, we curate a set of recommendations for researchers using the RIPE Atlas platform to foster transparency, reproducibility, and ethics.

翻译：网络测量平台因其分布式特性，简化了对互联网远程部分的测量，日益受到研究人员和运维人员的青睐。RIPE Atlas 在全球 178 个国家拥有超过 12,900 个观测点，是分析任播部署、网络延迟和拓扑等关键问题的重要工具。尽管每天产生超过 1 TB 的测量结果，但对其底层运行机制的理解仍有限。本文深入探讨了 RIPE Atlas 一天内的运行情况，涵盖 50,900 项独立测量和超过 13 亿条结果。虽然日常测量大多由用户定义，但内置测量和锚点网格贡献了 89% 的结果产出。我们详细分析了不同探测节点和测量任务如何支撑 RIPE Atlas 的日常运维，并评估了它们可能引入的偏差。此外，我们展示了如何利用现有测量数据研究网络审查、路由追踪对称性以及保留地址块的使用等问题。最后，我们为使用 RIPE Atlas 平台的研究人员提出了一系列建议，以促进透明度、可重复性和研究伦理。