VMware to Microsoft: The Expert Panel
Seven Hyper-V and Azure Stack HCI guests on one call that ran long enough to become two videos. What it's like getting a group of people who love this stuff talking at once.
Every other episode of VMware to Microsoft has been one guest, one topic, a structured conversation start to finish. This one wasn’t that. I got seven people who’d each already been on the podcast, Bronson, Carl, Ben, Jaromir, Ernie, Daniel, and Philip, on one call at the same time and asked them a hard question. We just kept talking. It ran long enough that I split the recording into two videos instead of publishing one two-hour drop, but it was one sitting: seven people who clearly don’t get enough chances to argue shop talk with each other, making the most of it.
The opening question started a fight
I asked at what scale Azure Stack HCI stops making sense, where you should just be running plain Hyper-V on Windows Server instead. Bronson took it first, and he came in swinging: Microsoft markets Azure Stack HCI as an edge solution, not a data center solution, and he put a hard ceiling on it at 16 nodes. Jaromir pushed back that the ceiling doesn’t matter much anyway, because scaling isn’t linear. Double the nodes doesn’t get you double the performance, so there’s a practical sweet spot well under 16 regardless. Then Daniel went further than either of them: run a 16-node cluster in production and you can only lose two nodes before you’re rebuilding from backup, which he framed as a resiliency problem nobody should accept just to chase a bigger number.
That set the tone for the rest of the conversation. Five to seven people who’ve each run this in the field, agreeing on the shape of the problem and disagreeing hard on the specifics.
Nobody wants to get attached to a cluster
Partway into the scoped-volumes discussion, Daniel brought back the old line about servers being cattle, not pets, this time aimed at Azure Stack HCI clusters instead of individual VMs. Don’t get attached to one, don’t hand-name it, and if a node dies you replace it and move on. It resurfaced again, unprompted, an hour later once we’d rolled into what became the second video, still needling the same people who like to babysit infrastructure by hand. A joke that keeps finding its way back into two hours of unscripted conversation, without anyone planning it, is usually a sign the room actually enjoyed each other.
The real argument was AVD versus RDS
The best tangent, by a wide margin, happened in the second half. The question was whether Azure Virtual Desktop is worth it over running RDS on your own Hyper-V or Azure Stack HCI cluster. The cloud side made the agility case: a few clicks and a business has a few thousand working desktops the next day, and Microsoft owns the control plane so you don’t have to.
Philip didn’t let that stand. He walked through what it costs to stand up a genuinely highly available RDS farm on Hyper-V, broker, gateway, session hosts, a separate NPS box for MFA, and argued that the “agility” you’re buying evaporates the moment a Microsoft-side outage puts you in a support queue behind a hundred thousand other tickets instead of calling a person by name. He’d made a version of this same point earlier in the panel, running the numbers on a 60-month amortization for a small accounting firm client where an hour of downtime during tax season costs real money. Somebody on the AVD side offered to settle it with an actual race: stand up a two-node Azure Stack HCI cluster and an AVD deployment from bare metal at the same time and see who finishes first. Nobody’s run that race yet. I’d watch it.
For the fun of it
That’s the real reason I wanted this format. Not a debate stage, just a group of people who all enjoy this tech and have built businesses and careers on top of it. It brought out a lot of interesting insights as well as a lot of fun.
This post is part of the VMware to Microsoft podcast series.
Subscribe for more of these.