Choose a plan
How to match Budget, Standard or Extreme to what you plan to run, and how to check whether a server you already have is too small or too big.
Our plans differ in CPU, memory, disk and monthly transfer. Everything else is the same: KVM, NVMe storage, and one IPv4 and one IPv6 address on every one. These are the planned sizes. They're fixed, and priced, when orders open. The plans page always has the current list.
| Plan | vCPU | RAM | NVMe | Transfer |
|---|---|---|---|---|
| Budget | 1 | 2 GB | 40 GB | 2 TB |
| Standard | 2 | 4 GB | 80 GB | 4 TB |
| Extreme | 4 | 8 GB | 160 GB | 8 TB |
Custom, where you pick CPU, RAM and disk yourself, comes after launch.
Start with memory
Small servers almost always run out of memory before anything else, so that's the number to pick by.
Budget, 2 GB, is plenty for a WireGuard VPN, a static site or a small blog behind nginx, a chat bot, a Git server for one, or a box to tinker on.
Standard, 4 GB, fits a typical web app with its database on the same machine, a handful of Docker containers, a small mail server, or a monitoring stack.
Extreme, 8 GB, is for busy sites, bigger databases, CI runners, game servers, and anything that compiles large projects.
If you're between two, take the smaller one and measure for a week. Add a swap file so a spike slows things down instead of killing them.
Then the rest
One vCPU is fine for a server that's mostly waiting on the network. You want more cores if you build software, encode media, or handle lots of requests at the same moment.
For disk, add up the system (a few GB), your data, logs, Docker images, and room for local backups before they're copied off. Then leave a fifth of it free; full disks fail in strange ways.
Transfer is the monthly allowance of traffic. A few terabytes is far more than most sites and VPNs ever use. Big downloads and video are what eat it.
Checking the server you have
On a server that's already running, three commands tell you a lot:
free -h
df -h /
uptimeIn free -h, look at available, not free (Linux uses spare RAM for cache, so "free" always looks low). If available is regularly under a tenth of the total, or swap is always in use, you need more memory.
In df -h /, a Use% over 80% means it's time to clean up or size up.
uptime shows load averages for the last 1, 5 and 15 minutes. If they sit above your number of vCPUs, work is queueing for the CPU.
Monitor your server goes into more detail.
Changing plans later
Buy for what you run today, not what you might run next year. We'll publish how moving between plans works at launch. Moving to a new server always works in the meantime: create the bigger one, rsync your data across, switch DNS, then cancel the old one.