Public Cloud overview
Technical stack, zones, account and project organisation, CPU modes, and default limits.
Platform
Public Cloud is infrastructure as a service (IaaS): hostd operates the hardware, virtualization, storage, and network. The guest OS, applications, and in-VM configuration are your responsibility.
Technical stack
| Layer | Technology |
|---|---|
| Virtualization | Proxmox VE — enterprise repository, stable branch |
| Servers | Dell PowerEdge only, AMD EPYC CPUs |
| Network | Dell PowerSwitch, Juniper |
| Backup | Proxmox Backup Server (PBS) — separate PBS datastore per project |
| VM storage | distributed all-flash NVMe (SDS), RoCEv2, aio io_uring; triple replication in the cluster |
When you create a project, you pick a zone — one cluster: shared compute, storage, backup, and pricing.
Available zones:
- WAW-AMD1 — Warsaw; Atman WAW-2 and Equinix WA2 data centres (Tier III+). Data and backups stay physically in Poland.
VM storage
Typical latency for a single random 4K write with fsync (iodepth=1) is about 0.11 ms — measured at the hypervisor layer, not inside the guest filesystem or database. This is not an in-VM application SLA.
How resources are organised
Resources have two levels: account and projects. The account is your login and billing; a project groups infrastructure in a chosen zone.
| Level | What it holds |
|---|---|
| Account | Login, billing, vCPU limits (shared across all projects) |
| Project | VMs, networks, public IPs, Cloud Gateway, PBS backups (own datastore), team permissions, hourly costs |
Public Cloud bills hourly in PLN. There are two models:
- Prepaid (default) — account balance; usage is debited during the month (top-up, auto-reload).
- Postpaid — one VAT invoice on the 1st of the month for the previous month; requires approval (apply from a prepaid account).
Details: Billing.
CPU modes
When you create or resize a VM, you choose a CPU mode and vCPU count. vCPU is how many cores the guest OS sees. The mode sets a total CPU-time cap on the host for the whole VM — load can be uneven across cores.
One fully busy vCPU equals 100% of one physical core’s time. A VM with 4 vCPU could theoretically use 400% in total if all four cores run flat out in parallel.
| Mode | Total CPU cap |
|---|---|
| Performance | Up to 100% per vCPU (e.g. 4 vCPU → up to 400%) |
| Eco | 50% × vCPU count (e.g. 4 vCPU → 200%) |
| Slim | 25% × vCPU count (e.g. 4 vCPU → 100%) |
Example — Eco, 4 vCPU: cap 200%. Either four cores at about 50%, or two at 100% — no more than 200% combined. For Slim with 4 vCPU the cap is 100% (e.g. one core at 100% or four at ~25%).
CPU mode also affects how many vCPU fit within your account budget — see account limits below.
Limits (defaults)
The numbers below are starting limits — hostd can raise them on request, contact support. Automatic limit increases in the portal are planned.
Account
| Limit | Default |
|---|---|
| Projects per zone | 5 |
| vCPU budget on the account | 16 (Performance-equivalent units) |
| vCPU per VM | 16 |
The vCPU budget is shared across all projects and counted in Performance units. Eco uses 0.5 units per vCPU, Slim 0.25:
| CPU mode | Budget used per vCPU |
|---|---|
| Performance | 1 |
| Eco | 0.5 |
| Slim | 0.25 |
At budget 16: up to 16 Performance vCPU, 32 Eco, 64 Slim, or a mix. The per-VM cap is a plain vCPU count — CPU mode does not change it.
Project
| Limit | Default |
|---|---|
| Private networks | 1 |
| Backup jobs | 1 |
| RAM (sum of VMs) | account vCPU budget × 6 GiB (96 GiB at default) |
| NVMe OS disk (sum of VMs) | account vCPU budget × 100 GiB (1600 GiB at default) |
VM
| Resource | Value |
|---|---|
| Minimum vCPU | 2, step 2 |
| RAM | vCPU × 1 … vCPU × 6 GiB |
| OS disk | vCPU × 10 … vCPU × 100 GiB (step ×10) |
| User snapshot | 1 per VM |
| LAN interfaces | up to 8 (each needs a private network) |
| SSH keys at create | up to 5 |
Lowering vCPU, RAM, or CPU mode: at most once per 23 hours per VM; lowering vCPU and RAM requires a stopped VM. See Resize a VM.
Networking
| Topic | Limit / rule |
|---|---|
| Public IPv4 | From the hostd pool; PTR (reverse DNS) in the portal |
| IPv6 | Not available (roadmap) |
| Private networks | Default 1 per project; IPv4 only |
| External VLAN | Visible in the UI; enablement via support |
| WAN speed | 100 / 200 / 300 / 400 Mbps at 2–7 / 8–15 / 16–23 / 24+ vCPU |
| LAN speed | 2 / 3 / 4 / 5 Gbps at the same vCPU tiers (labels in the VM form) |
| VM firewall | Public (WAN) interface only; CIDR in a rule — max /26 |
| Forward DNS | hostd does not host customer DNS zones — PTR only for our IPv4 |
Cloud Gateway, port forwarding, WireGuard, and reverse proxy — Cloud Gateway.
Next steps
- Getting started — first project, VM, and backup job.
- Create a VM
- Private network
- Capabilities and roadmap