Background
This engineering firm had run project systems, databases, and client delivery on VMware for years.
After Broadcom acquired VMware, licensing costs rose sharply. At the same time, the channel could not obtain pricing for many months.
Business Challenge
Storage was growing about 4–7TB a year—mostly project images, drawings, and related files, alongside databases that required consistent performance. They still had available capacity, but they were expanding toward 100TB to support new work when they hit VMware’s 64TB datastore limit. Licensing—not hardware—stopped the expansion.
Dataforge had secured them perpetual VMware licenses some time ago, which provided time to evaluate alternatives without a forced weekend cutover.
Solution
We already ran Proxmox in production. Licensing is per-CPU, predictable, and does not impose a datastore-size limitation.
Proxmox does not require a separate vCenter management layer—the cluster itself is the management system. The API supports clean integrations. Ceph provided redundant, scalable storage for databases and VMs. Project images and bulk file archives were placed on ZFS over NFS.
We manage the platform for them. They continue their engineering work; we operate the hypervisor and storage.
Implementation
They already used Veeam. Veeam is compatible with Proxmox, so backup through cutover remained in place without a parallel backup redesign.
Proxmox’s VMware import tools moved most VMs. Clonezilla covered a few older systems and edge cases.
They later moved from Veeam to Proxmox Backup Server. An onsite PBS handles local backup; PBS-to-PBS replication sends copies to Dataforge’s offsite backup. Operational overhead is lower, licensing is simpler, and backup runs in the same stack as the cluster.
Project Considerations
Performance in testing met requirements. Validation also covered failure scenarios: node loss, path interruption, and restores under load.
Leaving VMFS meant designing capacity by workload. Databases and VMs run on Ceph SSD; images and bulk project files run on ZFS.
Proxmox runs on standard Linux hosts, which allows scripts, tooling, and integrations to run directly on the hypervisor layer.
Environment Comparison
| Previous Environment | Current Environment |
|---|---|
| Broadcom-era VMware pricing increases; channel pricing unavailable for months | Per-CPU Proxmox licensing with predictable cost planning |
| Tried to expand to 100TB; blocked by VMware’s 64TB datastore limit | No datastore-size license limitation; capacity follows workload growth (~4–7TB/year) |
| vCenter required as a separate management layer | The Proxmox cluster is the management system |
| Shared datastore model for databases, VMs, and project files | Ceph SSD for databases and VMs; ZFS+NFS for project images and bulk files |
| Veeam on VMware as a separate backup layer | Onsite PBS for local backup; PBS-to-PBS replication to Dataforge offsite backup |
| Hypervisor host access limited for custom ops tooling | Standard Linux hosts supporting scripts, tooling, and integrations |
Key Outcomes
- Predictable licensing — per-CPU Proxmox licensing after Broadcom-era pricing increases and months without channel quotes
- Past the 64TB datastore limit — planned 100TB expansion no longer blocked by VMware licensing; ~4–7TB/year growth continues
- Simpler backup operations — onsite PBS for local backup; PBS-to-PBS replication to Dataforge offsite
- Managed platform, clear roles — Dataforge operates Proxmox—Ceph for databases, ZFS for images, Linux hosts, usable API
Long-Term Track Record
Proxmox has been stable and predictable in production. Licensing remains straightforward to plan and renew.
PBS is fully integrated and simpler to maintain than the previous backup arrangement. The API supports monitoring and reporting. Live maintenance on 3- and 4-node Ceph clusters proceeds without user impact.