The renewal quote arrived with a VMware licensing cost increase that the finance team could not absorb. The same roughly 300 virtual machines would cost sharply more each year, with no added capacity. The client, a mid-sized Central and Eastern European media group running a broadcast archive and several publishing platforms, asked us for a VMware exit strategy, and it set one condition: the next platform must not tie it to another single US-controlled provider.

The group had no appetite for a rushed move. A broadcast archive that editors search every day and publishing platforms that carry breaking news do not tolerate a long freeze. The renewal date, though, was fixed.

The estate we assessed

We started with IT Consulting: a four-week assessment of the whole VMware estate before anyone proposed a target. We inventoried every VM, its owner, its storage, its network dependencies and its busiest hours. About 300 VMs fell into three groups: the archive and media-processing systems, the public publishing platforms, and internal tooling such as editorial back office, finance and identity.

Three findings shaped the plan. Around 40 VMs were idle or duplicated and could be retired, not moved. The archive systems carried the largest disks and the strictest rules about when they could be touched. A handful of VMs depended on VMware-specific features, mostly in the way snapshots and storage were handled, which we would have to replace with equivalents on the target.

What a VMware alternative in Europe had to deliver

The client’s criteria went beyond price. It wanted a platform whose provider has no parent company and no entity above it on which a foreign order could be served. It wanted the datacenters in the EU, operated by the provider and not resold from someone else’s cloud. And it wanted a way out of the new platform that was written down before the first VM moved.

DIVERSITY Cloud fitted that brief. We are privately held and owned by our founders and leadership, we operate our own datacenters, and every one of them is in the EU. The client chose Munich for the publishing platforms and Amsterdam for the archive systems, with the choice of site its own, as it is for every Cloud customer.

An exit plan you have never tested is a hope. The client wanted one it had run before it needed it.

Migrating VMware to cloud around the editorial calendar

Most migration plans are built around infrastructure dependencies. This one was built around the newsroom. We mapped the editorial calendar for the coming year: election nights, a major football tournament, budget announcements, the autumn broadcast schedule. Then we blocked every window in which a platform must not change.

What remained became six waves over roughly five months. Each wave moved a set of VMs that depended on each other, so that no application was ever split between the old and the new platform for longer than a weekend.

  • Wave 1: internal tooling and the retired-or-duplicated clean-up, to prove the process on low-risk systems.
  • Waves 2 and 3: the smaller publishing platforms, cut over at night in the lowest-traffic hours.
  • Wave 4: the main news platform, scheduled in a quiet fortnight between two major editorial events.
  • Waves 5 and 6: the broadcast archive and media-processing systems, moved last because they carried the largest disks and the longest copy times.

How we moved the VMs

For each wave we built the target first. We created the instances on our KVM-based Cloud, converted the VMware disk images to run there, and rebuilt the network layout the applications expected. Data was synchronised ahead of time, so the cutover window covered only the final delta, the DNS change and the checks.

Every cutover followed the same runbook, written with the client’s operations team and rehearsed on a copy of the VMs before the real window. The source VMs stayed powered off but intact for two weeks after each wave. That gave the client a rollback path without keeping the VMware licences running on the whole estate.

The archive waves took the most care. The disks were large, so we began copying weeks before the cutover and checked file counts and checksums against the source before the client’s archivists signed off. Editors kept searching and retrieving footage on the old platform until the last window.

Where a VM relied on a VMware-specific feature, we replaced the behaviour instead of copying it. Snapshots and backups moved to the Cloud snapshot and backup features, and the client’s team learned the new routine in the early waves, before the production platforms depended on it.

The exit plan for the new platform

The client’s second requirement was a documented exit from the platform it was moving to. We wrote it with them during the migration, using what we offer on exit. Cloud instances export as a raw disk image plus a JSON configuration, managed databases as a full dump, and Object Storage through its S3-compatible API. None of these exports carries switching or egress charges.

The plan names the owner of each step, the order in which the systems would leave, and where the exported images would land. To test it, the client’s team exported a non-production instance as a raw disk image with its JSON configuration, and booted it on a hypervisor outside our platform. The test showed which networking details the JSON did not carry, and the team added them to the runbook.

The plan covers data and instances. It does not remove commitments the client has chosen, such as a savings plan, which still runs to the end of its term.

What changed

The client’s infrastructure spend came out about a fifth lower than the previous year, measured by its own finance team on a full year of invoices. That figure is the client’s result and depends on its estate, not a guarantee of ours. The retired VMs contributed to it, and so did the end of a licensing model that billed for capacity the group no longer needed.

  • Licensing risk removed: the group no longer faces a renewal on a platform it does not control, and the next price change is not a forced decision.
  • Concentration reduced: its production estate runs with a provider that has no parent and no entity above it on which a foreign order could be served.
  • Exit documented and tested: the group holds a written plan for leaving the platform, with an export it has actually run.
  • Editorial uptime kept: no major publishing platform changed during a protected editorial window, and no wave needed a rollback.

The newsroom noticed little of this, which was the aim. What the finance director and the head of technology noticed was a lower bill and a plan on paper for the day they want to move again.

If you are facing a similar renewal, we can run the same assessment on your estate and show you the waves before you commit to anything. The details of this project are available under NDA.