Stanford AI Course Map 中文 GitHub ↗
← Back to the course map
Stanford · 2026—27 CS 349F ML Systems & Hardware ★★★★☆

Core course · ML Systems & Hardware

Fabric Architectures For AI Systems

中文:AI 系统的网络 Fabric 架构

The network fabric of AI clusters: fat trees, congestion control, and scheduling, aiming for ultra-low latency and near-100% goodput.

Draft translation: official facts are checked; wording may change after native review.

Offered in Autumn 2026–27; the official site is restricted, so only the ExploreCourses link is provided.

2 units Autumn 2026 Satisfactory/No Credit ★★★★☆
NetworkingClustersData centersSystems

Course positioning

CS 349F focuses on the interconnect networks of large GPU/CPU clusters and on time-sensitive systems such as financial trading and multiplayer games. It covers data-center fabric architectures, transport protocols, congestion control, load balancing, and scheduling, with industry experts discussing real cluster design and operations.

What it covers

  1. Data-center fabric: fat trees, leaf-spine, topologies
  2. Transport protocols: RDMA, RoCE, congestion control
  3. Load balancing: ECMP, packet spraying
  4. Scheduling: job scheduling, fabric scheduling, interaction
  5. Edge vs network-centric: two high-performance paths
  6. Industry cases: GPU clusters, trading systems

Prerequisites

CS 144/CS 140 or distributed-systems basics; EE 178 helps.

Assessment & workload

2 units, S/NC; discussion plus a project/case study. Check the current syllabus for details.

Who it's for

Networking, clusters, and data centers; anyone working on AI infrastructure.

Study advice

Read it with CS 349D/E to get the full single-machine, cluster, and network picture.

Related courses

Official course description

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.