What Is an AI Fabric?

The networking sense, not "data fabric"

Two different fields use the word "fabric". A data fabric is a software architecture for unifying and governing data across an enterprise. An AIファブリック, the subject of this page, is a physical network: the wires, switches, and transport that move RDMA traffic between GPUs. When an AI infrastructure team says "the fabric was the bottleneck", they mean this network, the one that decided how long the training run took.

How the fabric is built: rails and Clos

An AI fabric is a leaf-spine (Clos) network tuned for one traffic pattern: many GPUs sending large flows to each other at once. The common building block is rail-optimized wiring, where each of a GPU server's 8 NICs connects to its own dedicated rail leaf, so the dominant same-rail AllReduce traffic stays on one leaf and never touches the spine. As the cluster grows past a single pod, the design extends to a 3ステージClos (leaf, spine, super-spine). On radix-64 Tomahawk 5 switches, a 2-tier fabric reaches roughly 2,048 GPUs at 1:1 non-blocking, and a 3-stage Clos scales to 16,000+ GPUs. See the AIファブリックトポロジー and rail-optimized network pages for the port maths.

Why it has to be lossless

GPUs move data with RDMA, which performs poorly when packets are dropped. So the AI fabric is engineered to be lossless: Priority Flow Control (PFC) stops a switch queue from overflowing, and Explicit Congestion Notification (ECN) marks packets early so the sending NIC slows down before loss happens. That combination is what carries RoCEv2, the transport GPUs use over Ethernet. OcNOS-DC ships this fabric with PFC (including over L3), ECN, Dynamic ECN, WRED, and dynamic load balancing to keep every path evenly loaded.