← Back to articles
News· 3 min read

The OpenNebula Model for Migrating VMware Workloads with OneSwap

Tino Vazquez, COO at OpenNebula Systems, published an article on June 17, 2026 explaining how to move VMware workloads onto a KVM-based cloud managed by OpenNebula. The piece makes one thing clear up front: migration is not just converting a disk format. It’s a process worth planning and running in stages, especially if you have a large fleet of virtual machines built up over years of vSphere investment.

Three phases instead of a blind jump

The method splits into three phases. The first is validation and configuration: you assess the source VMware environment and prepare the target OpenNebula cloud. Many companies start with a small “pivot” cloud to test things before touching production.

The second phase develops the migration process itself. You take representative VMs from different OS families and workload types, migrate them as a pilot, and document what adjustments the guest OS needs. That way the procedures are tuned before the bulk of the work.

The third phase is execution, done in controlled waves following the validated steps: VM preparation, image conversion, metadata transfer, resource creation, network mapping, and final validation.

OneSwap and its transfer methods

The tool that moves the data is OneSwap. It supports several methods depending on your setup:

  • Hybrid transfer, which downloads the image locally before importing it into OpenNebula.
  • VDDK-based transfer, using VMware’s Virtual Disk Development Kit.
  • Direct SSH copy from the ESXi host.
  • Transfer through the vCenter API, which the article flags as the slowest option.

It also handles OVA files and standalone VMDK disks, in case you need to import appliances or individual disks.

Delta migration: from hours of downtime to minutes

The most useful part for anyone running services they can’t stop is delta migration. Most of the data transfers while the source VM keeps running. Once the initial sync finishes, only the disk blocks that changed during the short cutover are copied. That brings service downtime from hours down to a few minutes.

As a concrete reference, a Windows Server 2022 VM with a 20 GB disk converted in about 12 minutes over a 1 Gbps network with local storage.

What to expect in time and cost

OneSwap reached a 90% automatic VM conversion rate in real migration projects. The article notes this is a field metric, not a guarantee: the outcome depends on the guest OS version, VM configuration, storage, networking, and each application’s constraints.

On timelines, a 50 to 100 VM environment moves in 4 to 8 weeks; 100 to 250 VMs in 2 to 4 months; 250 to 1000 VMs in 3 to 9 months. For a 250-VM environment, professional services typically run from around €25,000 to €50,000 depending on scope.

If you want a better grasp of the target stack’s parts, our walkthrough of KVM, QEMU and libvirt on RHEL is a good companion read.

Who this matters to: teams looking to cut infrastructure costs, reduce dependence on a single vendor, or lay the groundwork for a private, sovereign, or hybrid cloud without throwing away years of VMware investment.

Source

Original article from OpenNebula: The OpenNebula Model for VMware Workload Migration, by Tino Vazquez (June 17, 2026). Aggregated through QEMU as a virtualization news source.