DocsBest practices
Choosing the right plan
How to size RAM and CPU without overpaying.
Best practices 5 min read
Most projects need far less than people expect. The right approach is to start small, measure for a few days, then adjust. Changing plan takes seconds and never requires a redeploy.
Typical needs
| Project | Usual memory | Suggested start |
|---|---|---|
| Small Discord bot (1–50 servers) | 80–200 MB | Free or Starter |
| Discord bot with music / Lavalink | 400 MB – 1 GB | Pro |
| Small REST API | 100–300 MB | Hobby or Launch |
| API with heavy traffic | 500 MB – 2 GB | Growth, with 2 instances |
| Queue worker / cron jobs | 100–400 MB | Free or Worker |
Signs you should upgrade
- Memory stays above 80% of your limit for long periods.
- Your service is restarted with an
Out of memorymessage in the logs. - CPU is pinned at 100% and response times climb.
- You need a feature only available on higher plans, such as custom domains or multiple instances.
Signs you can downgrade
If memory stays under 40% and CPU under 20% for a week, you are very likely on a plan that is too large. Drop one level and keep an eye on the Metrics tab.
Check the Metrics tab over a full week before deciding. Weekend and evening traffic often looks very different from a Tuesday morning.
Was this article helpful?