Engineering at Meta

JUNE 12, 2024

There are two leading choices in the industry that fit these requirements: RoCE and InfiniBand fabrics. On the other hand, Meta had built research clusters with InfiniBand as large as 16K GPUs. So we decided to build both: two 24k clusters , one with RoCE and another with InfiniBand. Both of these options had tradeoffs.

Infiniband 141

Infiniband Data Centers Topology Network 141

Engineering at Meta

AUGUST 5, 2024

We ensure that there is enough ingress bandwidth on the rack switch to not hinder the training workload. The BE is a specialized fabric that connects all RDMA NICs in a non-blocking architecture, providing high bandwidth, low latency, and lossless transport between any two GPUs in the cluster, regardless of their physical location.

Network 132

Network Networking Topology Data Centers 132

Engineering at Meta

MARCH 12, 2024

The other cluster features an NVIDIA Quantum2 InfiniBand fabric. Through careful co-design of the network, software, and model architectures, we have successfully used both RoCE and InfiniBand clusters for large, GenAI workloads (including our ongoing training of Llama 3 on our RoCE cluster) without any network bottlenecks.

Infiniband 145

Infiniband Data Centers Server Network 145

Router-switch

SEPTEMBER 25, 2024

In today’s high-speed networking world, optimizing and troubleshooting performance is crucial, especially with high-performance equipment like NVIDIA Infiniband switches. Whether you’re a data center admin or network engineer, mastering effective techniques is key.

Infiniband 40

Infiniband Network Networking Routers 40