Tag: #RoCE
Kubernetes, DevOps and MLOps, written from things we ran and measured ourselves · 1 posts
GPU Networks Read via RDMA and NCCL: InfiniBand, RoCE, Kubernetes Deployment
We break down communication layers and bandwidth differences inside and outside GPU servers, explain the differences between InfiniBand and RoCE v2, GPUDirect RDMA, NCCL operation and measurement methods, and common fail
2026-10-07 · 41 min read #RDMA#InfiniBand#RoCE#NCCL#GPUDirect