VMware to Microsoft: Let's Talk About Hardware

Everyone moving off VMware asks the software question first: Hyper-V or Azure Stack HCI? That’s the wrong first question. The hardware you already own decides a lot of this for you, and in episode 3 of VMware to Microsoft, Philip Elder walked through exactly how.

Philip runs MPECS Inc. with his wife. They’ve been building Hyper-V and clustering solutions since Server 2008, and before that, Virtual Server 2005. He’s also a longtime Microsoft MVP. So when he talks about hardware planning for a hypervisor migration, he’s talking from two decades of racking the gear.

Your Hyper-V license might already be paid for

Here’s the point Philip kept coming back to. If your VMware cluster is running Windows Server Datacenter as the guest OS on those VMs, you’re already licensed for Hyper-V. Datacenter edition includes unlimited Hyper-V rights on that host. You paid for it. You’re just not using it.

That changes the math on a migration. Philip walked through rough numbers on the podcast: a Windows Server Datacenter license runs somewhere around $7,000 to $8,000 per node. Across a six node cluster, that’s licensing you don’t have to buy again if you go to Hyper-V, because it’s already covering you. Meanwhile VMware’s vSAN and vMotion licensing on that same cluster could run well into six figures at renewal. Philip’s example put it around $150,000 for the VMware side versus roughly $100,000 to build out a solid Storage Spaces Direct or Azure Stack HCI cluster from scratch, once you’re not also paying for VMware.

That’s the reframe. The Hyper-V license isn’t a new cost. The VMware renewal is the new cost, and it’s the one you can avoid.

Three real paths off VMware

Philip laid out the options he presents to clients:

Backup and restore. Take your existing backup product, one that supports hardware independent restore, flatten the box, install Windows Server, and restore the VMs onto Hyper-V. There’s downtime and some network reconfiguration, but you land in the same place you started, just off VMware.
New hardware at the natural refresh point. Most clusters are built for a 5 year run. If you’re already near that mark, you’re going to buy hardware anyway. Time the buy to coincide with the OS move, and the money you’re not spending on VMware licensing helps fund it.
Move it to the cloud. Not covered in depth in this episode, but Philip flagged it as a legitimate third option next to the other two.

Reusing hardware for Storage Spaces Direct isn’t a given

If you’re hoping to keep your current servers and just switch hypervisors, Philip’s advice was blunt: send him the bill of materials before you assume anything.

Storage Spaces Direct and Azure Stack HCI both rely on RDMA for node to node storage traffic. That traffic has to be lossless, and iSCSI doesn’t get you there the way RDMA does. Philip’s preference is Mellanox (Nvidia) network adapters, specifically because they just work. He’s seen Broadcom adapters come and go, and he called out real problems with Intel’s 700 series and some issues on the 800 series too.

The practical filter: your existing network adapters need to support at least iWARP, one of the RDMA implementations, at minimum. If they don’t, that’s a hardware gap you need to close before Storage Spaces Direct is even on the table, regardless of how good your CPUs and storage are.

Philip’s rule for these builds is to eliminate every single point of failure. He’s blunt about it: he’s watched contractors talk clients into “good enough” builds that skip redundancy, and he doesn’t work that way. If the storage layer is latent, it doesn’t matter how much CPU you threw at the cluster. Latency there kills the whole design.

Ecosystem is the real differentiator

Philip also made a case for Hyper-V and Azure Stack HCI over other alternative hypervisors, and it wasn’t really about features. It was about who’s answering when something breaks.

He described checking out one hyperconverged competitor and hitting two red flags immediately: customers weren’t allowed to talk publicly about the product, and there was nothing in the vendor’s forums about real problems people were hitting. Compare that to Hyper-V or VMware, where there’s a global community of MVPs, bloggers, and forum posts covering the good, the bad, and the ugly. When you’re stuck at 2am, that’s what gets you unstuck, not a glossy feature comparison.

The takeaway

Before you pick a hypervisor, get your bill of materials together and check what your current network adapters support. Check whether your VMs are already running Datacenter edition, because that Hyper-V license may already be sitting there unused. Then decide: restore onto new infrastructure, ride out your hardware to its natural refresh cycle, or move to the cloud. All three are legitimate. Which one wins depends on your hardware, not on which hypervisor has the nicer UI.

This post is part of the VMware to Microsoft podcast series.

Subscribe for more of these.