Workload domain · 2 ESX hosts × 4 H100
VKS GPU worker (40 GB)Deep Learning VM (10 GB)Inference VM (20 GB)free
Two ESX hosts in a VCF workload domain, four H100 80GB GPUs each. Nine workloads want GPUs: four VKS GPU worker nodes, three Deep Learning VMs for data scientists and two inference VMs. Pick how the GPUs are presented to VMs, then put a host into maintenance mode, fail the other one, or start a noisy neighbour. Behaviour follows NVIDIA AI Enterprise on vSphere; check exact support for your vGPU and VCF versions.