Vendor Sheet
NVIDIA AI Data Platform: Reference Architecture Addendum for AI on Supermicro SYS-212GB-FNR with NeuralMesh by WEKA
NVIDIA AI Data Platform: Reference Architecture Addendum for AI on Supermicro SYS-212GB-FNR with NeuralMesh by WEKA
This addendum extends the Supermicro SYS-212GB-FNR AI reference architecture with NeuralMesh by WEKA as scalable Kubernetes-based storage for NVIDIA AI Data Platform deployments. Dedicated Supermicro storage servers use NVMe drives, BlueField-3 or ConnectX-7 400G adapters, Spectrum-X Ethernet, RoCE, Ubuntu, Kubernetes, Cilium, and the WEKA Operator. NeuralMesh provides software-defined storage with microsecond latency and support for NFS, SMB, POSIX, S3, and NVIDIA GPUDirect; the document cites 10× faster model training, 93% GPU utilization, and 90% faster data pipelines. PaletteAI and Canonical MAAS automate the AI and storage clusters, while WEKA custom resources handle drive discovery, cluster initialization, clients, and CSI integration. The architecture supplies sizing, networking, st
