Public Cloud
Instead of renting one pre-built VM, you provision infrastructure — compute instances, block storage volumes, object storage, and private networks — as separate building blocks through an OpenStack API or dashboard. Infomaniak offers this as their Public Cloud product, in the same Swiss, steward-owned, green data centres as their VPS.
| Single VPS | Public Cloud | |
|---|---|---|
| Scaling | Resize the whole VM (often with downtime) | Resize compute and storage independently, on the fly |
| Storage | OS and data share a fixed disk | Block-storage volumes attached and grown separately from compute |
| Object storage | External bucket, wired in manually | Native S3-compatible object storage in the same tenant |
| Networking | Single public IP, OS-level firewall | Private networks, floating IPs, security groups from the control panel |
| Reproducibility | Manual setup, snapshots | Infrastructure-as-code via OpenStack API / Terraform |
| Billing | Fixed monthly per plan | Pay-per-use per resource |
| Complexity | Low | Higher — you assemble the pieces yourself |
Provisioning
Infomaniak Public Cloud is standard OpenStack, so it can be driven three ways:
- Horizon dashboard — the web UI, for clicking through instance and volume creation.
- OpenStack CLI —
openstack server create,openstack volume create, etc., using credentials downloaded from the dashboard. - Terraform — the
openstackprovider, to keep the whole setup in version control and rebuild it reproducibly.
For a self-hosting stack, Terraform is the recommended path: it captures the instance flavour, the data volume, the network, and the security groups as code, so the infrastructure can be torn down and recreated without clicking through the dashboard each time.
Cost
Public Cloud is billed per resource, per hour.
Compute
| Flavour | vCPU | RAM | € / hour | ~ € / month |
|---|---|---|---|---|
| a2-ram4 | 2 | 4 GB | €0.00725 | €5.29 |
| a4-ram8 | 4 | 8 GB | €0.01450 | €10.59 |
| a4-ram16 | 4 | 16 GB | €0.02018 | €14.73 |
Storage
| € / GB / hour | ~ € / GB / month | |
|---|---|---|
| Block (Perf1) | €0.00011 | €0.08 |
| Object | €0.000013 | €0.01 |
Object storage is one-tenth the price of block storage — the right fit for shared data that accumulates over time: Nextcloud files, Penpot assets, and backups.
Network
| € / hour | ~ € / month | |
|---|---|---|
| Reserved IPv4 | €0.00411 | €3.00 |
Traffic is included up to 10 Gbps across the entire cloud tenant; costs apply above that threshold.
Why Public Cloud
Public Cloud makes this setup more reliable, cheaper to run, and more secure — at the cost of more knowledge required to operate it.
More reliable. A single VM shares CPU and memory across everything. A misbehaving app — a Wordpress site under load, a runaway cron job — can starve the rest. Separate instances contain the blast radius: a problem on one instance stays on that instance.
Cost savings. While a bit more complex to set up, if the knowledge and skills are available it can save significantly in the long run.
- Right tool for the job. Not everything needs the same storage. Block storage is fast and close to the instance — right for the OS and active container data. Object storage is cheap and scales automatically — right for things that accumulate over time: Nextcloud files, Penpot assets, backups. On a single VM these get conflated onto one disk; here they go where they actually belong.
- Precise resource allocation. Each instance is sized for its workload. A maintenance server running Coolify and monitoring doesn't need the same CPU and RAM as a server running ten user-facing services. You pay for what each instance actually needs, not for a single machine large enough to cover the worst case across all of them.
- Scaling when needed. Compute and storage scale independently. Adding a volume, resizing an instance, or spinning up a new server doesn't touch anything else. Growth is incremental rather than a forced migration to a larger plan.
More secure. Coolify assigns servers to teams, so not every user sees the full infrastructure. For that model to work cleanly, the underlying compute needs to be separate instances — not a single shared VM where all teams' containers run side by side.
Easier to maintain. Everything — instances, volumes, object storage buckets, networks — lives in one tenant and is managed through one interface. On a single VPS, ancillary services like object storage get wired in from external providers and tracked separately. And because the infrastructure is code, it can be maintained programmatically — including with the aid of AI.
But it comes at the cost of knowledge. You assemble the pieces yourself rather than getting a pre-built machine. That means understanding OpenStack primitives — instances, volumes, networks, security groups — and managing them through Terraform rather than a simple control panel.