For once a Gartner article that speaks sense, the key bit in the article is the size of organisation and complexity.
I deal with small organisation with a couple of VM on a single cluster to multiple clusters on one or two sites to the enterprise with hundreds of clusters with tens of thousands of VM spread across the globe. And the report is focused on the enterprise migrations and to be honest their figures are conservative from my experience. Enterprise you will often have hardware, networking, ESX, storage, etc all split across multiple teams never mind thousands of affected apps many with legacy configurations that IT doesn't want to support. Mix in a push for architecture modernisation, hardware and app refresh cycles and maintain business as normal and you have a recipe for a nightmare. For them giving a clear cut figure for ESX Exit isn't so straightforward and if you were to include duplicate hardware, app upgrades, training in new platforms, testing, migration disruption, etc, etc you can see the cost mounting up.
We have a 25,000 VM split across 200 clusters across the globe ranging from 3 host clusters to 100 host clusters, with 1,500 apps and realistic figure for migration is probably $50m to $100m over 4 years if you included everything. We looked at outsourcing the migration and found the bulk of the cost is things you can't easily outsource as it requires more business knowledge rather than technical skill or manpower. As you can guess, the ESX replacement licensing is a fraction of the cost of the project. And of course the vendor has a migration tool and tells us how wonderful is, that is great for newer VM with a automated build, then have a large number of servers that are hand script builds but at supported OS and then of course the legacy stuff which we don't want to migrate of ESX as OS is no longer supported. And speaking to other enterprise organisation, their challenges are just as bad. Its easy enough for hosting companies to say, yeah we migrated 10,000 VM in a weekend. Sure that little bit of picking up the VM from ESX to some alternate hypervisor is a piece of cake, they don't own the workloads and that is where the challenge is and where outsourcing may not be of great help.
Timing wise, we started project at beginning of 2023 with a view to finish by end 2026 and that is being optimistic to say the least. So anybody who hasn't started has a long road ahead of them.