D is correct. A dedicated Supervisor Cluster for each tenant establishes the strongest infrastructure and Kubernetes control-plane isolation. Each Supervisor has its own Kubernetes control plane and underlying vSphere infrastructure boundary, avoiding the shared control-plane dependency that would exist if multiple tenants used namespaces on the same Supervisor.

This design also addresses the requirement for independent failure domains . A Supervisor deployed on separate workload-domain infrastructure can be associated with distinct vSphere clusters, networking, storage, and lifecycle boundaries. Consequently, failure or maintenance affecting one tenant ' s Supervisor does not inherently affect another tenant ' s Kubernetes control plane.

The independent-upgrade requirement is also significant. Broadcom documents Supervisor and VKS lifecycle operations at the Supervisor level , including compatibility validation between the Supervisor, VKS service, and workload-cluster Kubernetes releases. VKS service upgrades are activated on the relevant Supervisor, reinforcing the Supervisor as an important lifecycle boundary.

A and B are useful security controls but do not isolate Kubernetes control planes or infrastructure failure domains. C provides namespace-level tenancy, but all namespaces still depend on the same Supervisor control plane and its lifecycle , so it cannot satisfy the stated isolation and independent-upgrade requirements.

Study Guide References/Topics: VCF VKS Architecture; Supervisor Clusters; Workload Domains; Tenant Isolation; Kubernetes Control-Plane Isolation; Failure Domains; Supervisor and VKS Lifecycle Management.