Guides
Estimate cloud costs before you build
A practical method to estimate the monthly cloud cost of an application before deploying it. The cost drivers, the assumptions that matter, and how to get a line-by-line estimate from official catalogs.
The four cost drivers
- Always-on compute: instance or task size × count × hours. Most applications pay mainly for this.
- Managed data: database size, high availability, storage and backups.
- Traffic: egress to the internet, NAT processing in private networks, load balancer capacity, CDN.
- Operations: logs, metrics, traces, alarms, CI minutes and the container registry.
The assumptions that matter
| Assumption | Why it matters |
|---|---|
| Peak requests per second | Sizes compute and load balancers |
| Share served by the CDN | Removes traffic from compute and egress |
| Average response size | Drives egress |
| Data volume | Drives storage and backup costs |
| Log retention | Log storage grows linearly with it |
| Environments | Staging and development add a fraction of production |
Brunel shows every assumption it uses, with its default, and re-prices instantly in your browser when you change one. See How prices are computed.
Common surprises
- NAT gateways in private networks charge per hour and per GB processed.
- High availability roughly doubles database instance costs.
- Idle environments: staging running around the clock.
- Logs kept longer than anyone reads them.
Get an estimate you can defend
Paste your repository in Brunel. Every line of the estimate shows quantity × unit price with the catalog date and source, from the AWS Price List or Scaleway's public catalog, for four tiers. Components without an official price say so instead of guessing.