Ceph, StarWind, and the VMware Blues: What Happens When Broadcom’s vSAN Is Off the Table

Actually, it starts the same way most small-to-medium IT stories do: with a manager saying no.

A system admin we’ll call T., sitting on a pile of refurbished enterprise servers, was setting up a three-node vSphere 8 cluster. Everything was ready — the networking, the disks, the licenses. The only missing piece? Shared storage. VMware’s own vSAN was the obvious solution. But after Broadcom’s acquisition of VMware, the licensing model changed, and the cost jumped to a level his manager couldn’t justify.

So T. did what any good sysadmin does when told “no”: he went searching for a workaround.

His mission was simple — find something that behaves like vSAN, but doesn’t come with Broadcom’s price tag.

That’s how he ended up comparing StarWind, Ceph, MinIO, and a handful of open-source storage systems that promise vSAN-like magic — the kind of magic that lets virtual machines migrate seamlessly, survive host failures, and share storage without an expensive SAN in sight.

But as the community around him quickly reminded him, not all storage systems are created equal.

The Search for a vSAN Alternative

In a typical VMware setup, vSAN acts as the glue that makes hyperconverged infrastructure (HCI) possible. It pools the local disks of multiple hosts into one shared datastore — simple, elegant, and tightly integrated.

When you lose vSAN, that glue disappears. Suddenly, you’re left piecing together NFS shares, iSCSI targets, or full-blown software-defined storage stacks to get the same functionality.

That’s where the names StarWind, Ceph, and MinIO start flying around.

T. had heard good things about StarWind vSAN, even got a quote for it, but his manager didn’t want to spend. That’s when he turned to the internet to ask: can Ceph or MinIO do the job?

The answer, from nearly every experienced admin who chimed in, was a mix of sympathy and technical realism.

“No, you can’t use S3 as a datastore type,” one VMware engineer replied flatly when asked about MinIO. “Ceph can technically be set up for VMware, but it’s not officially supported outside IBM’s version. And it’s not cheap.”

MinIO, as several people pointed out, is for object storage — great for Kubernetes, backups, and S3-compatible apps, but completely the wrong tool for virtual machines. Ceph, on the other hand, can expose iSCSI or NFS targets, but it’s a beast to deploy, especially for a small 3-node setup.

“There’s nothing that does what you want that’s totally free and reliable,” one veteran engineer summarized. “StarWind Free probably won’t eat data, but it’s CLI-only. DIY with DRBD or Ceph? Good way to lose your job when something breaks.”

That’s the heart of the VMware blues right now — the growing frustration among small teams that still want HCI-like performance and resilience but can’t afford, or justify, the enterprise-level licensing now tied to Broadcom’s ecosystem.

Broadcom’s New VMware World

To understand why admins like T. are suddenly scrambling for vSAN alternatives, you have to look at what happened after Broadcom bought VMware in 2023.

Broadcom didn’t just buy the company — it tore down the product lineup and rebuilt it around VMware Cloud Foundation (VCF) and VMware vSphere Foundation (VVF) bundles. The standalone vSphere licenses many small businesses relied on? Phased out.

Even vSAN, which was once sold as an add-on, became part of these bundled subscriptions. If you didn’t buy into the full Foundation suite, you didn’t get vSAN — or NSX, or Aria Operations.

As one commenter dryly noted:

“How long do you think you can run pure vSphere? Until it hits end of life in 2027.”

That date matters. VMware has confirmed that vSphere 8 — the last version you can run outside VCF/VVF — reaches end-of-life in October 2027. After that, anyone still holding onto a standalone cluster will be running unsupported infrastructure.

In short: Broadcom’s VMware isn’t built for the small cluster crowd anymore.

So if you’re a small IT team with three hand-me-down servers and a modest budget, you’re faced with an ugly choice.

Either:

• Pay up for VCF or VVF,
• Switch to another hypervisor entirely,
• Or build a DIY shared storage solution and pray it holds.

That’s where the “Ceph or StarWind?” debate gets interesting.

StarWind: The Familiar Favorite

StarWind isn’t new to this space. For years, it’s been the go-to alternative for teams priced out of VMware’s vSAN or Microsoft’s Storage Spaces Direct.

The concept is simple: you install StarWind Virtual SAN (VSAN) on your hosts, and it mirrors local storage between them over the network. The result is a shared iSCSI target that vSphere sees as a single datastore.

Admins like it because it’s easy to understand, integrates neatly with VMware, and just works once it’s configured correctly. But there’s a catch — the paid version isn’t cheap either.

One user put it bluntly:

“I set up StarWind Free and it runs great. But setting up networking and storage through their PowerShell library at the end is a headache.”

Another added:

“You could do it for free, but it’s strictly CLI. Support is awesome, though.”

StarWind’s free edition does exist, and for lab or test environments it’s solid. But production workloads? That’s where you hit limits — no management GUI, fewer features, and no official support.

Still, compared to something like Ceph, StarWind feels downright friendly.

“Ceph is viable but more complex,” another commenter advised. “If your manager won’t approve StarWind, go simple — NFS or iSCSI from a NAS box.”

That’s the recurring theme here: everyone loves the idea of open-source HCI until they actually have to manage it.

Ceph: The Open-Source Giant That’s Just Too Big

On paper, Ceph is the ultimate answer. It’s open source, massively scalable, and powers some of the largest storage clusters on Earth.

In practice, for a small VMware environment, it’s like bringing a freight train to a go-kart race.

Ceph’s beauty lies in its distributed architecture — data is replicated across multiple nodes, self-healing, and fault-tolerant. It can serve object, block, and file storage all in one. But the learning curve is brutal.

Admins in the thread warned that running Ceph for VMware workloads means dealing with:

• Complex configuration for iSCSI gateways
• Manual maintenance of OSDs and monitors
• Strict requirements for quorum and latency
• Painful recovery when things go wrong

And even if you get it working, it’s not officially supported by VMware except in IBM’s productized form, which uses custom NVMe/TCP layers — and costs as much as vSAN anyway.

As one long-time engineer joked, “DIY Ceph is a good way to lose your job when someone makes a woops.”

The Other Contenders: NFS, iSCSI, and NAS Boxes

For admins who just need something that works, the most pragmatic advice wasn’t about software-defined anything — it was about keeping it simple.

Several people suggested NFS or iSCSI exports from a NAS appliance, like a Synology or QNAP.

“Synology FS2500–12 bays, dual PSUs, dual 10G — $3,400,” one user wrote. “Add SSDs and you’re under $6K for 5TB usable. If your boss balks at that, they’re just not serious.”

Of course, others were quick to point out that by the time you fully outfit a Synology or QNAP with enterprise SSDs and redundancy, you’re nearing the price of a Dell PowerVault or HPE MSA — both of which offer better support and reliability.

In the end, these “budget SAN” routes might not save much money, but they offer one huge advantage over DIY Ceph setups: predictability.

A NAS won’t surprise you with a split-brain cluster or a misconfigured quorum that tanks your data. It just… works.

The Emotional Layer: When the Cost of VMware Isn’t Just Money

The most revealing part of these discussions isn’t the tech itself — it’s the tone.

You can hear the fatigue in the admins’ words.

Many of them have been running VMware for a decade or more. They’ve built entire infrastructures around vSphere and vSAN, and suddenly they’re being told the world they knew is ending.

“How long can you run pure vSphere? Until it dies.”

“We are hardware-rich but software-poor.”

“We didn’t design up front and used up our one budget request.”

It’s frustration mixed with resignation. VMware was once the safe, stable choice — the boring option that always worked. Now it feels uncertain, corporate, and distant.

As one engineer put it: “When your manager says no to vSAN, he’s really saying no to HA, vMotion, and uptime.”

The Bigger Picture: Broadcom’s Gamble

Broadcom’s strategy isn’t random — it’s aiming squarely at the enterprise market. By bundling everything into subscription-based suites like VCF, it ensures predictable revenue and tighter control over the VMware ecosystem.

For Fortune 500 companies, that’s fine. They can absorb the cost.

But for smaller IT shops, MSPs, labs, and home-grown clusters, it’s a gut punch.

That’s why communities are suddenly buzzing with talk of Proxmox, TrueNAS Scale, Hyper-V, and XCP-ng — platforms that promise freedom from the Broadcom lock-in cycle.

One commenter even joked, “Tell your manager to prepare the VMware exit — and buy Broadcom stock.”

So What Happens Now?

For T. and countless others like him, the road forward is uncertain. Maybe he’ll convince his manager to spring for StarWind. Maybe he’ll build a small NFS-based cluster. Or maybe, by 2027, he’ll migrate to Proxmox and never look back.

But the bigger story isn’t about a single admin’s setup — it’s about a shift in how small IT environments think about infrastructure.

The era of “just use VMware” is ending.

The future looks more fragmented, more open-source, and yes, more complicated.

StarWind, Ceph, TrueNAS, Synology — none of them are perfect replacements for vSAN. But they represent a new mindset: if the giants won’t cater to small teams anymore, small teams will build their own ecosystems.

And maybe that’s not such a bad thing.

In the end, Broadcom didn’t kill vSAN — it just reminded everyone why open infrastructure still matters.