There is a real conversation happening in small business IT right now: AWS and Azure bills keep climbing, and the same workloads can run on open-source cloud platforms for a fraction of the cost. Sometimes that is true. Sometimes it is a trap. This post is the honest version.
What people mean by "open-source cloud"
There is no single project called "open-source cloud." It is a stack:
- Hypervisor / cluster manager — Proxmox VE, OpenStack, Harvester
- Storage — Ceph, MinIO (S3-compatible), Longhorn
- Networking — built-in or layered (e.g. SDN, MetalLB for Kubernetes)
- Backup — PBS (Proxmox Backup Server), Restic, Velero
- Orchestration (optional) — Kubernetes (k3s, RKE2), or just plain VMs
For small business, the most common practical setup is Proxmox + Ceph + PBS running on three physical nodes in a colo or office rack. That gives you VMs, shared storage, snapshots, and offsite backup — the core of what a hyperscaler IaaS account gives you, just on hardware you own.
Where it actually saves money
The honest savings come from a specific shape of workload:
- Predictable, steady-state compute. A database, a couple of internal apps, file shares, a CI runner. Things that run 24×7 whether you are using them or not.
- High-egress workloads. AWS charges roughly $0.09/GB for egress to the internet. If you serve video, downloads, or backups, that line item alone can pay for an entire on-prem cluster.
- Storage-heavy workloads. S3 is cheap until you have 50 TB and you are accessing it a lot. Ceph or MinIO on local disk is dramatically cheaper at that scale.
- Dev and test environments that you do not need to scale to zero.
If your workload looks like that, the math is real. We have seen genuine 60–80% reductions in monthly cloud spend.
Where it costs more than it saves
The math falls apart in these cases:
- Bursty, unpredictable traffic. Auto-scaling is what hyperscalers are great at. If your load varies 50× between weekday and weekend, AWS earns its margin.
- Tiny teams. If you have one IT person and they already have a full-time job, the operational cost of running a cluster (patching, capacity planning, hardware replacement) often eats the savings.
- Workloads that depend on managed services. RDS, DynamoDB, Lambda, Cognito — replacing these with self-hosted equivalents is doable but takes engineering time you may not have.
- Compliance with strict residency or audit requirements. Sometimes the easiest path to a SOC 2 or HIPAA audit is a managed service with a turnkey BAA, not a stack you have to document yourself.
A realistic small business setup
For a 20–100 person business that wants to start moving workloads off hyperscalers, a sane starter setup looks roughly like this:
- Three Proxmox nodes (mid-range refurbished servers — 64–128 GB RAM, 8–16 cores, NVMe + a few TB of SSD each). Enough for HA and Ceph quorum.
- Proxmox Backup Server on a fourth, cheaper machine, ideally in a different building.
- A small managed firewall / router (OPNsense, MikroTik, or a Fortinet if you need vendor support).
- MinIO for S3-compatible object storage, behind your apps.
- Tailscale or WireGuard for secure admin access without exposing management interfaces to the internet.
Total hardware cost: typically $8k–$20k up front. Monthly run-rate: power and bandwidth, often $200–$600/month depending on colo or office. Compare that to a comparable AWS bill at $4k–$10k/month and the payback is months, not years.
The honest tradeoffs
You will pay back the savings in:
- Operational responsibility. When a disk fails, you swap it. When the cluster has a quorum issue at 2 a.m., that is yours.
- Capacity planning. You cannot just provision more. You buy and install it.
- Skills. Someone on the team needs to actually understand how Ceph behaves under pressure, or you need a partner who does.
This is where the "best of both worlds" pattern usually wins: keep workloads that benefit from managed services on a hyperscaler, move the steady-state and storage-heavy workloads to your own cluster, and have a real backup story between them.
Where F09 Tech fits
We are actively exploring open-source cloud infrastructure as part of our build-out — both to power our own SOC and managed services and to offer it to clients who want a cheaper, more controllable alternative to hyperscaler bills. If you are thinking about this for your business, we would be glad to walk through the math with you — including the cases where staying on AWS is the right call.