Building a Modern Control Plane for Hyper-V (20 Years Later)

Halvantic — Sat, 26 Sep 2026 08:54:07 +0000

After 20 years in infrastructure, I finally built the Hyper-V management layer the market should have had years ago. Here's why I did it, how I did it, and what I learned.

The Problem

Hyper-V is solid. It's a genuinely capable hypervisor. But the management story around it has always been broken.

SCVMM feels like it was built in another era. It's clunky, expensive, and doesn't inspire confidence. Windows Admin Center is better, but it's single-host focused, not a fleet orchestrator. Most people either hand-roll PowerShell scripts, use DSC, or just accept that they don't have the centralised control they'd get with vSphere.

VMware/Broadcom's licensing changes forced me to seriously ask the question: could Hyper-V actually replace vSphere?

The honest answer: the hypervisor's fine. The management layer is the problem.

The Gap

vCenter treats vSphere as a first-class citizen. You get a control plane that knows vSphere inside and out. Declarative desired state. Continuous reconciliation. Fleet orchestration across multiple hosts and clusters.

Nothing does the same for Hyper-V.

I looked at existing open source options before building anything. Most FOSS orchestration in this space targets KVM, Proxmox, or oVirt. Not Hyper-V specifically. And the ones that touch Hyper-V treat it as a second-class citizen, not something worthy of a proper control plane.

So I built one.

What I Built

Ballast is a control plane and agent for Hyper-V hosts and failover clusters.

Each host runs an agent that pulls its desired configuration from the centre and keeps enforcing it continuously. If the centre goes offline, the host doesn't freeze or drift. It just keeps honouring whatever it was last told (network config, cluster membership, VM state, etc.) and catches back up automatically once the centre is reachable again.

Losing the centre never changes intent, only the ability to change it further.

It also handles cluster provisioning (SET-teamed switches, Storage Spaces Direct, Cluster Shared Volumes), VM lifecycle and templates, live migration, Hyper-V Replica-based disaster recovery, and VMware migration tooling that streams VMDKs directly in.

The agent, the piece running with elevated privileges on your hosts, is fully open source (Apache 2.0). You can audit every line. The control plane console is closed source. Community tier is free forever for one cluster of any size plus unlimited standalone hosts. A paid tier for multi-cluster, SSO, and audit logging is coming.

The Approach

I didn't build this alone, but I built it by myself.

Let me explain. I architected the entire system and drove every design decision. I tested every change against real Hyper-V infrastructure and debugged live issues. I own the product decisions, the architecture, the roadmap.

Claude Code, an AI coding agent, handled the implementation. It wrote the code, tests, and documentation under my direction. This let me move at a pace that solo development normally can't match. I was strict about testing, validation, and real-world constraints.

20 years of infrastructure design and architecture went into Ballast. The AI was the execution tool that let me move fast, not the designer.

Being upfront about that matters. Too many people hide their process. I'd rather be honest about using modern tooling to solve hard problems faster.

Why This Matters

Hyper-V users have been asking for this for over a decade. Not a feature-rich behemoth. Not something that does everything. Just a clean, modern control plane that treats Hyper-V like it deserves to be treated.

Simplicity matters. Engineers are tired of bloated management layers that get in the way. This is built to do the core job well, do it reliably, and stay out of the way.

And the agent is open source. That means you can see exactly what's running on your infrastructure with elevated privileges. You can fork it, modify it, audit it. That trust matters when we're talking about code that runs on your hosts.

Try It

Community tier is free forever. One cluster of any size, unlimited standalone hosts. That covers the overwhelming majority of homelab and small business setups.

Site: https://ballast.halvantic.com
GitHub (agent): https://github.com/halvantic/hyperv-control-plane
Product Hunt: https://www.producthunt.com/products/ballast

I'd genuinely love your feedback. Where does it break? What's missing? What looks wrong?

DEV Community: Halvantic

Building a Modern Control Plane for Hyper-V (20 Years Later)