Cloud bills creep. A test environment left running, logs retained forever, an oversized database, egress nobody measured. None of it is dramatic; together it doubles the invoice.
See it first
Turn on cost allocation tags by environment and service. One dashboard, reviewed monthly. You cannot manage a number you only see once a year.
The usual savings
- Right-size compute and databases to real utilisation, not the number you guessed at launch.
- Auto-stop non-production environments overnight and on weekends.
- Set retention on logs, metrics and backups. Ship old objects to cold storage.
- Put a CDN in front of static assets to cut egress and load.
- Commit to a savings plan only for the steady baseline you are certain of.
Keep it from coming back
A budget alert per environment. A line in code review for anything that provisions infrastructure. Cost becomes a normal design constraint, not an annual fire drill.