A conventional five-node Nutanix HCI cluster has five Controller VMs—one CVM running on each node. AOS operates as a distributed storage and data-services system rather than relying on a single centralized storage controller. The per-node CVM architecture provides local access to the node’s storage devices while cooperating with the CVMs on the other nodes to form the distributed AOS storage fabric.
The CVMs collectively provide services associated with data placement, replication, metadata, caching, compression, deduplication, snapshots, and cluster storage management. If a CVM becomes unavailable, the host can access storage services through another CVM, subject to cluster health and connectivity. The remaining CVMs also maintain the distributed cluster services required for continued operation.
The number of CVMs therefore normally matches the number of HCI nodes: three nodes produce three CVMs, four nodes produce four, and five nodes produce five. The count does not represent the replication factor or the number of simultaneously tolerated failures.
Compute-only nodes are a specific exception because they do not run a CVM. However, the question explicitly describes a five-node HCI cluster, so each node contributes storage and runs a Controller VM.
NHCF/NCA reference: Nutanix HCI architecture, AOS, nodes, and Controller VMs. Official references: Infrastructure Resilience and Compute-Only Nodes .
