OpenKube · AI infrastructure walkthrough

How a GPU-backed AI workload finds capacity

Follow an AI workload request from the OpenKube control plane to a partner datacenter with available NVIDIA GPU capacity—using existing infrastructure and Kubernetes-based orchestration.

Customer AI inference request: 1 NVIDIA GPU storage: nvme-ssd OpenKube · AI control plane AI workload request openkube-orders ns GPU-aware placement multi-tenant ProvisioningJob → site-<id> ns fulfilled ✓ Site registry 3 sites live Read-only service Observability aggregator Operator dashboard Sites live: 3/3 Tenants: 3 auto-rolled up · live in-west · Mumbai Site A GPU: none agent · polling Utilization 78% tenant-012 running reserved capacity in-north · Delhi Site B GPU: 4× NVIDIA agent · polling Utilization 41% tenant-031 running reserved capacity in-south · Bengaluru Site C GPU: 2× NVIDIA agent · polling Utilization 63% tenant-026 running reserved capacity

↔ swipe to pan the diagram

1 / 8
SITE AGENTS

Registering AI-ready capacity

Loading…

Illustrative AI infrastructure flow using OpenKube's Kubernetes control plane: sites report capacity, the hub evaluates placement, and the partner site provisions locally. KubeVirt and NVIDIA GPU Operators bring virtualized workloads and GPU resources into the platform.