Serve customer models in dedicated inference environments. vMetal provisions infrastructure, and vCluster assigns Private Nodes with separate tenant control planes and optional runtime hardening.
Shared inference infrastructure needs clear tenant boundaries across the control plane, worker capacity, runtime, and network.
Namespace-only designs share control-plane and node-level components between customers.
Shared worker nodes can expose inference workloads to cross-tenant resource contention.
Data residency depends on the placement and configuration of compute, storage, networking, and inference services, not on a single product setting.
vMetal supplies machines for inference workloads, and vCluster provides independent Kubernetes tenant environments with dedicated Private Nodes. Runtime and network controls complement application authentication, model access policies and workload readiness checks. vCluster Labs serves 50+ GPU clouds & Fortune 500s.
Single-tenant AI inference combines a separate tenant control plane, Private Nodes, and optional runtime and physical network isolation.
Private Nodes dedicate worker capacity to one tenant cluster, with its own CNI and storage configuration. For dedicated inference workloads, select physical servers or VMs according to the required infrastructure boundary.
vNode uses Linux user namespaces and seccomp to restrict workload privileges and system calls. It adds runtime hardening for dedicated inference workloads without a guest kernel or hypervisor, complementing dedicated workers rather than allocating GPUs.
Tenant-owned CNI configuration and supported Netris integrations provide separate network controls for dedicated inference workloads. Ethernet segmentation can use VLANs, VRFs and ACLs; InfiniBand isolation uses fabric partitions rather than Ethernet constructs.
vCluster Platform supports air-gapped deployments and FIPS features on supported plans for infrastructure operated in controlled environments.
Each inference tenant gets its own virtualized Kubernetes control plane, API server, and RBAC boundary.
This isn’t a side project. Behind every vCluster deployment is 5+ years of deep K8s engineering, security hardening, and battle-tested infrastructure work at massive scale.
Talk to our team about your stack
Deploy vCluster on your infra in minutes
Go live with a hyperscaler-grade tenant experience in days
Namespaces divide resources within a Kubernetes cluster, but share its API and cluster-wide administrative scope. RBAC, quotas and network policies remain useful controls. vCluster adds a separate tenant API and data store, with Private Nodes dedicating production workers. This reduces dependencies between tenant environments while platform administrators still manage the underlying infrastructure and its failure domains.
Residency depends on where you place the platform, tenant clusters, storage, networking and inference services, so the architecture must be set up for the required jurisdiction. vCluster Platform supports air-gapped deployments and FIPS features on supported plans, and Polarise uses it to enforce 100% data residency at the infrastructure layer.
vCluster gives each tenant an independent Kubernetes API, data store and RBAC scope. Private Nodes dedicate worker capacity to that tenant, including its own CNI and storage configuration. vMetal orchestrates the machines underneath. This separates tenant administration and worker placement; shared infrastructure, application permissions and network policies still require an explicit security design.
vNode uses Linux user namespaces and seccomp to restrict workload privileges and system calls, helping prevent container breakout. It uses the underlying Linux kernel rather than a separate guest kernel. Add it as runtime hardening alongside dedicated Private Nodes, patching and least-privilege policies. Validate kernel, container runtime, GPU access and workload compatibility; it is not a replacement for GPU allocation or application authorization.
vCluster Platform can coordinate tenant environments across supported clouds and data centers. Each Kubernetes tenant has its own control plane; a central fleet interface does not mean one tenant API or one failure domain. Plan worker placement, routing, storage and data residency for each location. An optional encrypted VPN can connect control planes and workers where appropriate for the deployment.
Provisioning a tenant control plane is one step in creating an inference environment. Private Node capacity, GPU drivers, model downloads, storage and serving endpoints must also be ready. vMetal can automate supported machine provisioning, while your serving stack manages model loading and endpoint health. Test authenticated requests and model readiness before treating the environment as available to customers.
See how vCluster Platform delivers single-tenant AI inference environments with hard tenant isolation.
