Moving off VMware

Cloud Exit migrates VMware environments to Kubernetes with KubeVirt, on the infrastructure you choose.

SAVINGS

€50,400 a year

VMware Cloud Foundation license fees on three hosts with 144 cores. Kubernetes and KubeVirt have no license fees.

YOUR VMS

Applications run unchanged

Each VM keeps its disks, operating system and network settings, including VLANs and IP addresses.

MIGRATION

Copied while running

Forklift copies each VM to KubeVirt while it runs on VMware. Switchover happens in an agreed window, and the original stays until you sign off.

RUNS ON

Your servers or rented bare metal

In-house server clusters, air-gapped environments, colocation and rented bare metal such as Hetzner.

Overview

Virtual machines on Kubernetes

KubeVirt runs virtual machines on Kubernetes, next to containers. Each VM keeps its disks, operating system and network settings, so applications run unchanged. We build the cluster where you need it and migrate the VMs in planned batches.

Migration flow

Assess

We export your vCenter inventory, map networks and storage to the new cluster and agree the order in which VMs move.

Copy

Forklift copies each VM’s disks to the cluster while the VM keeps running on VMware, and keeps them in sync until switchover.

Switch over

In an agreed window the VM is shut down on VMware, the remaining changes are copied and the VM starts on KubeVirt.

Verify

You test the application. The original VM stays on VMware, switched off, until you sign off.

VMware

vCenter and ESXi

VM running
Shut down
Switched off, kept until sign-off

KubeVirt

Kubernetes cluster

Networks and storage mapped
Disks copied, kept in sync
VM running on KubeVirt
AGREED WINDOW
  1. Assess

    We export your vCenter inventory, map networks and storage to the new cluster and agree the order in which VMs move.

    VMwareVM runningKubeVirtNetworks and storage mapped
  2. Copy

    Forklift copies each VM’s disks to the cluster while the VM keeps running on VMware, and keeps them in sync until switchover.

    VMwareVM runningKubeVirtDisks copied, kept in sync
  3. AGREED WINDOW

    Switch over

    In an agreed window the VM is shut down on VMware, the remaining changes are copied and the VM starts on KubeVirt.

    VMwareShut downKubeVirtVM running on KubeVirt
  4. Verify

    You test the application. The original VM stays on VMware, switched off, until you sign off.

    VMwareSwitched off, kept until sign-offKubeVirtVM running on KubeVirt

Savings

Stop paying VMware license fees

Kubernetes and KubeVirt have no license fees. For a typical estate of three hosts with two 24-core CPUs each, 144 cores in total, these are the VMware Cloud Foundation license fees you no longer pay. Most vSphere features have a direct equivalent; the feature parity table shows where they differ.

TODAY · VMWARE CLOUD FOUNDATION€50,400 per yearLicense fees for 144 cores
AFTER · KUBERNETES WITH KUBEVIRT€0 per yearNo license fees, open source
YOU SAVE€50,400saved every year
YOU SAVE€151,200saved over a three-year term
YOU SAVE€16,800saved every year for each extra 48-core host

What you keep

VMs keep their disks, operating systems and network settings, and applications run unchanged. The capabilities teams use every day carry over.

  • Live migration, like vMotion
  • High availability and restart
  • Snapshots, templates and clones
  • CPU and memory hot-add
  • Distributed storage
  • Existing SANs through CSI drivers
  • Existing VLANs and IP addresses
  • Veeam backups
  • GPU passthrough and vGPU
  • Monitoring and alerting
  • Roles and single sign-on
  • Automation with Terraform

Savings are VMware license fees only, based on the widely reported VCF list price of €350 per core per year; Broadcom does not publish a price list. Customers on vSphere Foundation, quoted in 2026 at €100 to €196 per core per year, save €14,400 to €28,224 a year on the same estate. Hardware and hosting are needed on either platform. Cloud Exit support and services are priced separately. Figures as of October 2026.

Engagement

What's included in the engagement

Assessment and plan

An inventory of your VMs, networks and storage, sizing of the new cluster and a migration schedule.

Kubernetes cluster

A highly available cluster with KubeVirt, CSI storage and Cilium networking. VLANs and IP addresses carry over.

VM migration

Every VM moved with Forklift in batches scheduled with your team, including conversion to VirtIO drivers.

Monitoring and alerting

Prometheus, Grafana and Alertmanager cover the cluster, storage and every VM, with alerts sent to your team.

Backups and Git

VM definitions are kept in Git and applied automatically. Velero backs up the VMs and the cluster state.

Handover and support

Documentation, training for your administrators, and ongoing platform support after the last VM has moved.

Architecture

Architecture compared with vSphere

KubeVirt adds virtual machines to Kubernetes. Each VM runs in a pod as a QEMU process on KVM, so VMs and containers share one scheduler, storage layer and network.

VMware vSphere

vCenter manages ESXi hosts

TODAY

vCenter Server

Inventory, HA, DRS, vMotion

esxi-01
web-01
db-01
ESXi
esxi-02
app-01
file-01
ESXi
esxi-03
db-02
erp-01
ESXi

Distributed switch, NSX

Port groups on VLANs

vSAN or VMFS

Datastores shared by hosts

OPERATIONSAria OperationsVADP backup
ForkliftVMs copied while running
ForkliftVMs copied while running

Kubernetes with KubeVirt

Kubernetes manages Linux nodes, KubeVirt adds VMs

AFTER MIGRATION

Git repository

VMs and networks as YAML

GitOps

Applies each change

Control plane

Highly available on three nodes

API serveretcdschedulervirt-apivirt-controller
node-01
web-01
pod
db-01
pod
apicontainer
virt-handler
Linux and KVM
node-02
app-01
pod
file-01
pod
workercontainer
virt-handler
Linux and KVM
node-03
db-02
pod
erp-01
pod
frontendcontainer
virt-handler
Linux and KVM

Cilium network

VMs keep their VLANs and IPs

Replicated storage

Local NVMe across nodes

Existing SAN or NAS

Vendor CSI driver

OPERATIONSPrometheus and GrafanaVelero or Veeam KastenForklift

The same x86 servers

Both platforms run on standard hardware

In-house clustersAir-gapped sitesColocationRented bare metal, such as Hetzner
Virtual machineContainerKubernetes podKubernetes and KubeVirt components

Platform

Why Kubernetes

On vSphere, people operate infrastructure step by step. On Kubernetes you describe what you want, your VMs and applications, and the platform does the work of making reality match it, continuously.

Intent instead of instructions

Declare the intent

Every VM, network and policy is written down in Git as the state you want. No click-paths, no unrecorded changes.

Every change on record

Every change is a reviewed commit: who changed what, when and why. Rolling back means reverting that commit.

Self-healing

Kubernetes keeps reality matching your intent. Failed VMs restart elsewhere and VMs move off hosts being patched.

No license fees

Kubernetes and KubeVirt are open-source CNCF projects under Apache 2.0. Nothing to renew, and it runs anywhere.

When you are ready

Containers

Applications move from VMs to containers one at a time, on the same cluster and the same network.

Scaling and rollouts

Containerised applications scale with demand and deploy with rolling updates and a quick rollback.

Managed services

Kubernetes operators run databases such as PostgreSQL with automated failover, backups and upgrades.

How a VM runs on KubeVirt

  1. 01

    Defined in Git

    The VM's CPU, memory, disks and networks are described in a manifest, stored in Git.

  2. 02

    Applied to the cluster

    GitOps applies the manifest to the Kubernetes API, which records the VM.

  3. 03

    Scheduled

    virt-controller creates a pod for the VM, and the scheduler places it on a node with capacity.

  4. 04

    Started

    virt-handler on that node starts the VM with QEMU and KVM, attached to its volumes and VLANs.

Parity

Feature parity with vSphere

Everything marked equivalent is set up as part of the migration.

EquivalentPartialDifferent approach

Compute and availability

FEATUREVSPHEREKUBEVIRT
HypervisorESXiKVM
Central managementvCenter ServerAPI and Git; web console optional
Live migrationvMotionBetween nodes on shared storage
Storage live migrationStorage vMotionVolume migration while the VM runs
High availabilityvSphere HAVMs restart on a healthy node
Workload balancingDRSDescheduler rebalances periodically
Fault tolerancevSphere FTQEMU COLO; experimental, not orchestrated
CPU and memory hot-add
Templates and clonesGolden images and instance types
SnapshotsThrough the storage layer
Guest toolsVMware ToolsQEMU guest agent and VirtIO drivers
Full GPU / PCI passthroughDirectPath I/OSupported through VFIO, with IOMMU and device allowlisting.
Mediated vGPU assignmentNVIDIA vGPUSupported, including declarative creation of mediated devices when the vendor driver is installed.
SR-IOV vGPU assignmentSR-IOV passthroughSupported; VF creation and profile configuration require vendor tooling.

Storage, networking and operations

FEATUREVSPHEREKUBEVIRT
Distributed storagevSANReplicated block storage
External SAN and NASVMFS, NFS datastoresCSI drivers: NetApp, Pure, Dell, HPE
Virtual networkingDistributed switchCilium, existing VLANs kept
Micro-segmentationNSXCilium network policies
Resource pools and quotasNamespaces with quotas and limits
Roles and single sign-onRBAC with OpenID Connect
BackupVeeam, VADP toolsVelero, Veeam Kasten or Veeam plug-in
MonitoringAria OperationsPrometheus and Grafana, included
AutomationPowerCLI and TerraformKubernetes API, GitOps and Terraform
Image libraryContent libraryGolden images imported with CDI
Disaster recoverySite Recovery ManagerSecond-site replication; manual failover
ContainersKubernetes ServiceVMs and containers on one cluster

Where support is partial

Partial means the main use case needs a different tool or design. If you depend on Fault Tolerance, NSX routing and load balancing, or orchestrated disaster recovery, raise it during the assessment so the scope reflects it.

Deployment

Runs where you need it

The platform is the same wherever it runs. We deploy on your own servers, in isolated networks, or on rented bare metal, and manage every cluster the same way.

In-house server clusters

The cluster runs in your data center or colocation, on the existing VMware hosts after migration or on new servers.

Servers
Existing hosts or new x86 servers
Storage
Existing SAN or NAS, or local NVMe
Network
Existing VLANs and IP ranges

Air-gapped environments

For sites with no internet connection. Container images, VM images and updates are mirrored to a registry inside the network, and the cluster is installed and upgraded from there.

Images
Internal registry and mirror
Updates
Imported and applied offline
Git
Self-hosted inside the network

Rented bare metal

Dedicated servers rented monthly with no long-term contract, from Hetzner or another bare-metal provider.

Example
Hetzner AX162-2: AMD EPYC 9454P, 48 cores, 256 GB ECC RAM, 2 × 1.92 TB NVMe
Cost
€842.30 per server per month, plus setup
Locations
Falkenstein and Nuremberg (DE), Helsinki (FI)

Storage

On local NVMe, VMs use storage replicated across the nodes. With an existing SAN or NAS, KubeVirt uses the vendor's CSI driver, over Fibre Channel, iSCSI, NVMe/TCP or NFS.

  • NetApp ONTAPTrident
  • Pure Storage FlashArrayPortworx
  • Dell PowerStore, PowerMax, PowerFlexDell CSI drivers
  • HPE Alletra, Primera, NimbleHPE CSI Driver
  • IBM FlashSystemIBM block storage CSI driver
  • Hitachi VSPHitachi Storage Plug-in for Containers

Partner

Why Cloud Exit

Built on 10+ years of experience with Kubernetes, virtualization and containerization.

10+ years

Experience with Kubernetes, virtualization and containerization

Dedicated engineers who love their craft

Options for 24x7 SLAs

Cost effective

Clear up-front pricing structure and agreements

Start your migration today

contact@cloud-exit.com
  1. 01

    Send us your inventory

    An RVTools export or a list of hosts and VMs is enough to start. If you have a renewal quote, send that too.

  2. 02

    Get a scope and price

    We size the cluster and quote the migration at a fixed price.

  3. 03

    Migrate in batches

    VMs move on a schedule agreed with you. The originals stay on VMware until you sign off.