VPS and servers · 4 min
How to choose a VPS plan
The rule of thumb: start with a plan that covers what you need today and add resources once real load shows up. Moving to a larger plan doesn't require reinstalling the system.
What the specs mean
- vCPU — virtual processor cores. One core handles a site with moderate traffic; databases, builds and parallel jobs benefit from 2–4 cores.
- RAM — memory is the most common bottleneck. A modern Linux system with a web server uses 300–500 MB; a database needs anywhere from a few hundred megabytes to several gigabytes.
- NVMe — disk space for the system, site files, databases and logs. Keep 20–30 % free: a full disk is a frequent cause of outages.
- Traffic — the data a server sends and receives. On dedisrv VPS it isn't limited; the ceiling is the 1 Gbit/s port, which shares the host's uplink evenly with other busy servers.
Plans by workload
| Workload | Plan |
|---|---|
| Telegram bot, VPN, WordPress, CRM or wiki for a team | Pulse — 2 vCPU, 2 GB |
| Online store, several sites, Docker | Drive — 2 vCPU, 4 GB |
| Databases, APIs, CI runner | Forge — 4 vCPU, 8 GB |
| High-load services, large databases | Titan — 8 vCPU, 12 GB, 30 GB |
How to tell you're running short
A few commands show the server's current state:
free -h # memory and swap
df -h # disk space
uptime # average CPU load
Signs it's time for a larger plan:
free -hshows swap constantly in use and less than 10 % of RAM available (theavailablecolumn);- the load average in
uptimestays above the number of cores for long periods; - less than 15–20 % of the disk is free.
When you need a dedicated server
Consider a dedicated server if you need more than 12 GB of memory, consistent disk performance for a large database, or your own virtualization. The whole machine is yours — no neighbours sharing resources.
Not sure which to pick? Describe your workload in a request and we'll recommend a configuration.