With the increasing complexity of network topologies and architectures, adding intelligence to the network control plane through Artificial Intelligence and Machine Learning (AI&ML) is becoming a trend in network development. For large-scale geo-distributed systems, determining how to appropriately introduce intelligence in networking is the key to high-efficiency operation. In this treatise, we explore two deployment paradigms (centralized vs. distributed) for AI-based networking. To achieve the best results, we propose a hybrid ML paradigm that combines a distributed intelligence, based on units called "AI routers," with a centralized intelligence, called the "network mind", to support different network services. In the proposed paradigm, we deploy centralized AI control for connection-oriented tunneling-based routing protocols (such as multiprotocol label switching and segment routing) to guarantee a high QoS, whereas for hop-by-hop IP routing, we shift the intelligent control responsibility to each AI router to ease the overhead imposed by centralized control and use the network mind to improve the global convergence.

中文翻译：

---

AI 路由器和网络思维：数据包路由的混合机器学习范式

随着网络拓扑和架构的日益复杂，通过人工智能和机器学习（AI&ML）为网络控制平面增加智能正成为网络发展的趋势。对于大规模的地理分布式系统，确定如何在网络中适当引入智能是高效运行的关键。在这篇论文中，我们探讨了基于 AI 的网络的两种部署范式（集中式与分布式）。为了获得最佳结果，我们提出了一种混合 ML 范式，该范式将分布式智能（基于称为“AI 路由器”的单元）与称为“网络思维”的集中式智能相结合，以支持不同的网络服务。在提议的范式中，