Red Hat OpenShift Virtualization

Three nodes (compact cluster where control plane and workers share nodes); ODF needs at least three storage nodes.

Red Hat publishes tested, not absolute, maximums: 500 compute nodes tested in 4.20 and control-plane tests at 2,000 nodes; 10,000 VMs per cluster cited in 2026.

Single-node OpenShift can run VMs, without HA or live migration and without ODF-based HCI.

Two-node OpenShift with a local arbiter (4.20) or with fencing (4.21) targets edge sites; confirm ODF and virtualization support for these topologies with Red Hat.

ODF stretch cluster over two zones plus an arbiter, and Metro-DR with ACM (synchronous, RPO 0) for sites within about 5-10 ms RTT.

Yes. Worker nodes need not host storage; ODF can run on dedicated infrastructure nodes or as an external Ceph cluster.

Certified CSI storage from partners (e.g. NetApp, Pure, Dell, Hitachi, IBM) over iSCSI, FC or NFS, with or without ODF.

All-NVMe bare-metal nodes supported for ODF/Ceph; no NVMe-specific storage architecture.

ODF pool compression is available; deduplication only for Multicloud Object Gateway object data, not VM disks.

ODF internal mode uses 3-way replication for VM block storage; erasure coding needs an external Ceph design.

Three replicas by default with host, rack or availability-zone failure domains (3-zone HA) and stretch clusters.

VM snapshots and clones through CSI snapshots, with smart cloning on Ceph RBD.

ODF cluster-wide encryption in Essentials; KMS-backed encryption in ODF Advanced. RHCOS disk encryption on nodes.

Kubernetes CPU and memory requests and limits; storage IOPS limits depend on the CSI driver, not ODF.

CephFS (RWX) volumes for workloads inside OpenShift; no SMB file service.

S3-compatible object storage via ODF (Multicloud Object Gateway / Ceph RGW).

Not covered: using ODF RBD or CephFS for platforms outside OpenShift is unsupported; external-mode ODF may serve S3 to outside apps.

Regional-DR: asynchronous Ceph RBD mirroring between clusters on a schedule, orchestrated by ACM (ODF Advanced).

Metro-DR: synchronous replication over a stretched Ceph cluster, RPO 0 (ODF Advanced + ACM).

ACM with ODF DR operators fails over and relocates protected applications/VMs between clusters; no vSphere-SRM-style runbook UI.

OpenShift API for Data Protection (OADP) backs up and restores VMs and namespaces to S3; incremental CBT-based backup is a 2026 roadmap item.

No native immutable snapshot vault; relies on object-locked S3 targets and partner tools.

Veeam Kasten, Trilio, Storware, Commvault, Cohesity and Rubrik support OpenShift Virtualization.

OpenShift Virtualization is GA on AWS, Azure, Google Cloud, Oracle OCI and IBM Cloud bare metal, so a cloud cluster can be the recovery site.

OVN-Kubernetes SDN with User-Defined Networks (isolated networks within or across namespaces), localnet/VLAN secondary networks and load balancing.

Kubernetes NetworkPolicy and multi-network policies for VM traffic; Advanced Cluster Security in Platform Plus.

Kubernetes RBAC with namespaces; OAuth with LDAP, OpenID Connect and other identity providers.

Immutable RHCOS with SELinux; Compliance Operator scans and remediates to DISA STIG and CIS profiles; official CIS Benchmark for OpenShift Virtualization (2026).

DISA STIG for OpenShift 4; FIPS mode uses RHEL cryptographic modules submitted for FIPS 140-2/140-3 validation (x86_64).

Red Hat Advanced Cluster Management (in Platform Plus) or ACM for Virtualization (for OVE clusters): fleet VM views, policies, GitOps, observability.

Over-the-air cluster updates of OS, Kubernetes and operators with rolling node drains and VM live migration; server firmware stays with the OEM.

Declarative Kubernetes API for everything, OpenShift GitOps (Argo CD), Ansible Automation Platform VM inventory, and CLI.

Built-in Prometheus-based monitoring, alerts and virtualization dashboards; ACM fleet observability. No capacity forecasting.

Projects with quotas, VM templates and a developer catalog; no built-in chargeback.

Live migration, node drain, automatic restart after node failure with remediation operators; cross-cluster and storage live migration (2026); load-aware descheduling in preview.

Native: VMs and containers share one Kubernetes platform (not in the VM-only OVE edition). ODF supplies RWO/RWX/S3 storage for pods.

Same platform self-managed or as ROSA (AWS), Azure Red Hat OpenShift, OpenShift Dedicated, IBM Cloud and OCI.

Windows VMs with vTPM and Secure Boot are supported, but Red Hat lists no validated VDI broker.

GPU passthrough and NVIDIA vGPU via the GPU Operator; Red Hat OpenShift AI on the same clusters.

No built-in DBaaS; database operators from partners via OperatorHub.

Assisted Installer (with a virtualization operator bundle), agent-based or installer-provisioned bare-metal installs. Typically a day per cluster with experienced staff.

Certified bare-metal servers from the Red Hat Ecosystem Catalog, plus supported cloud bare-metal instances; no Red Hat appliance.

Single-node, two-node and compact three-node topologies with ACM zero-touch provisioning; larger footprint than hypervisor-only stacks.

Supported: disconnected installation and updates from a mirrored registry.

Migration Toolkit for Virtualization (included): cold and warm migration from vSphere, RHV, OpenStack and OVA; storage-offload migrations up to 10x faster with certified arrays.

No free supported edition; product trials only. Upstream KubeVirt and Ceph are free community projects.

Subscriptions are hardware-vendor neutral; managed cloud editions are billed by the cloud provider.

Software only; server hardware is supported by the OEM.

Premium: 1-hour initial response for Severity 1, 24x7.

Large certified ecosystem (storage, backup, networking, system integrators) listed in the Red Hat Ecosystem Catalog, and broad OpenShift/Kubernetes skills.

Open-source KubeVirt/KVM and Ceph with standard disk images, but operations and tooling are OpenShift-specific.