Checklist
AWS cost quick-wins checklist
What to check first when an AWS bill grows faster than the product: idle resources, sizing, storage, network and non-production environments.
Visibility
- Cost allocation tags are enforced, and untagged spend is reviewed every month.
- Cost Explorer is grouped by service, account and tag, and someone looks at it on a schedule.
- A budget alert exists for each account, going to a person rather than a shared inbox.
Compute
- Instances are sized from observed CPU and memory, not from the original estimate.
- Non-production environments are shut down, or shared, outside working hours.
- Stopped instances, unattached volumes and old snapshots are listed and removed.
- Steady workloads are covered by savings plans or reserved capacity, after sizing, not before.
- Auto Scaling groups have scale-in policies, not only scale-out.
Storage and data
- S3 buckets have lifecycle rules, and incomplete multipart uploads expire.
- EBS volume types match the workload, and gp2 volumes are reviewed for gp3.
- Log retention is set on every log group, with a reason for the number chosen.
Network
- Cross-AZ and NAT gateway traffic is measured, and the biggest flows are explained.
- VPC endpoints serve the AWS services that carry the most traffic.
- Load balancers that serve no traffic are removed.
Want someone to run through this with you?
The AWS cost and reliability review does exactly that: 1,400 EUR, 5 days.