将集体操作搬进网络，加速AI训练。
In-Network Collective Operations: Game Changer or Challenge for AI Workloads?
- 在节点或交换机中实现集体操作，减少通信延迟。
- 可提升AI训练性能，但面临六项关键技术挑战。
- 适合关注AI系统优化与网络协同的工程师与研究者。
本文总结了在网络中实施集体操作（INC）在加速人工智能（AI）工作负载中的潜力。我们提供了充分细节，使非AI或网络领域的研究人员也能理解这一重要领域，促进两个学科间的交流。考虑两种INC类型：边缘级INC（Edge-INC），在节点层面实现；核心级INC（Core-INC），嵌入网络交换机中。我们梳理了其潜在性能优势，并指出了在边缘和核心两种场景下阻碍其采用的六大关键障碍。最后，我们对未来的发展与应用提出了若干预测。
原文摘要 · Abstract (English)
This paper summarizes the opportunities of in-network collective operations (INC) for accelerated collective operations in AI workloads. We provide sufficient detail to make this important field accessible to non-experts in AI or networking, fostering a connection between these communities. Consider two types of INC: Edge-INC, where the system is implemented at the node level, and Core-INC, where the system is embedded within network switches. We outline the potential performance benefits as well as six key obstacles in the context of both Edge-INC and Core-INC that may hinder their adoption. Finally, we present a set of predictions for the future development and application of INC.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。