课程定位
CS 349F 关注大规模 GPU/CPU 集群的互连网络,以及金融交易、多人游戏等 time-sensitive 系统。课程覆盖数据中心 fabric 架构、传输协议、拥塞控制、负载均衡和调度,并邀请工业界专家讲真实集群的设计与运维。
主要内容
- 数据中心 fabric:fat tree、leaf-spine、拓扑
- 传输协议:RDMA、RoCE、拥塞控制
- 负载均衡:ECMP、packet spraying
- 调度:作业调度、fabric 调度、交互
- 边缘 vs 网络中心:两种高性能路线
- 工业案例:GPU 集群、交易系统
先修
CS 144/CS 140 或分布式系统基础;EE 178 有帮助。
考核与工作量
2 unit S/NC;讨论 + 项目/案例,具体以当季官网为准。
适合谁
网络、集群、数据中心方向;想做 AI 基础设施的人。
学习建议
和 CS 349D/E 一起看,形成“单机-集群-网络”的完整视角。
相关课程
官方课程描述(英文原文)
The course is concerned with the design and operation of the network fabrics which interconnect large-scale compute and storage nodes in modern AI GPU clusters, cloud computing systems, and "time-sensitive systems" like financial trading platforms and massive multi-player games. We will consider architectures, protocols and algorithms which enable these network fabrics to deliver deterministic and ultra-low latency at near-100% goodput. Topics include data center fabric architectures - the fat tree topology, transport protocols - congestion control and load balancing, and scheduling algorithms - job scheduling, fabric scheduling and their interaction. A particular focus will be the contrast and synergy between the edge- and network-centric approaches to building high-performance network fabrics. Students will hear from industry experts who design, operate and use large-scale GPU and CPU clusters. Recommended: Knowledge of basic Networking, OS, or Distributed Systems (CS 144, 140, or equivalent), as well as basic EE courses (EE 178) will be useful.