Hello everyone,
We recently shared a new blog post on how to build a multi-tenant AI Factory on NVIDIA GB200 NVL4 with InfiniBand using OpenNebula. It covers network isolation (VXLAN/VLAN), NUMA-aware scheduling, and GPU passthrough with optional MIG-based sharing. Plus, there’s a recorded demo showing the full setup in action.
Scaling AI infrastructure requires more than high-performance GPUs. It requires efficient resource sharing, predictable performance, and strong tenant isolation. This blog post addresses these challenges by combining these key building blocks into a unified architecture.
Read more and check out the screencast here 
Multi-Tenant AI Factory on NVIDIA GB200 NVL4 with InfiniBand
Best regards,
The OpenNebula Team