Why it matters
Databases are often sized for a launch peak or a worst case that never comes, and the extra capacity is billed every hour.How nOps finds it
This recommendation comes from AWS Compute Optimizer through Cost Optimization Hub, not from an nOps rule. AWS looks at the instance’s CloudWatch metrics over its lookback period (14 days by default) and recommends a class that would still meet the workload’s needs. The saving is AWS’s estimate, including your Reserved Instance and Savings Plans pricing.Before you start
AWS Compute Optimizer and Cost Optimization Hub enabled. Permission to modify the instance (rds:ModifyDBInstance).
How to implement
- Review the recommended class and check peaks outside the lookback window, such as month-end jobs and reports.
- Check memory headroom: database caches fill available memory, so watch FreeableMemory and read latency after the change.
- Change the class in a maintenance window (RDS console → Modify → DB instance class, or
aws rds modify-db-instance --db-instance-identifier DB_ID --db-instance-class NEW_CLASS). - For Aurora, resize readers first and fail over, so the writer changes last.
- Update Terraform or CloudFormation so a later deployment does not change it back.
Risks and considerations
- Changing the class restarts the database. With Multi-AZ the outage is usually a short failover.
- A smaller class has less memory for cache, which can raise read I/O and latency.
- To roll back, modify the instance back to its original class.
FAQs
Will a smaller instance slow my database down?
Will a smaller instance slow my database down?
Compute Optimizer recommends a class that covers the usage it measured. Check peaks it may not have seen, and watch latency after the change.