TopolearnAt scale, a few architectural and purchasing decisions routinely save 30 to 60 percent of a cloud bill.
Cost optimization works in a loop: see spend, reduce waste, commit for discounts, and architect for efficiency.
Visibility: tag consistently and activate the tags, export detailed billing data (CUR 2.0 via Data Exports) to query with Athena, and enable Cost Anomaly Detection to get alerted when a service's spend jumps. Track unit costs, such as cost per order, not just the total.
Waste: idle and oversized resources. Compute Optimizer recommends instance, EBS, Lambda and ECS sizes from utilisation. Schedule non-production environments off at night, delete unattached volumes, old snapshots and AMIs, and migrate gp2 volumes to gp3.
Purchase options:
- Savings Plans: commit to a dollar amount per hour for 1 or 3 years. Compute Savings Plans (up to about 66% off) apply flexibly across EC2 families, Regions, Fargate and Lambda; EC2 Instance Savings Plans (up to about 72%) are tied to a family in a Region. Cover the steady baseline, not peaks.
- Reserved Instances remain the mechanism for RDS, ElastiCache, OpenSearch and Redshift.
- Spot capacity is up to 90% cheaper but can be reclaimed with a 2-minute warning. Use it for stateless, fault-tolerant work: CI, batch, containers. Diversify instance types and use capacity-optimised allocation.
Architecture: Graviton instances and arm64 Lambdas, S3 Intelligent-Tiering and lifecycle rules, gateway endpoints instead of NAT for S3 traffic, CloudFront in front of origins, same-AZ traffic where it is safe, right log retention, and serverless for spiky low-volume workloads.
Example
# Rightsizing findings from Compute Optimizer
aws compute-optimizer get-ec2-instance-recommendations \
--query 'instanceRecommendations[].[instanceArn,finding,recommendationOptions[0].instanceType]' \
--output text
# arn:aws:ec2:eu-west-1:111122223333:instance/i-0a1b... Overprovisioned m7g.large
# What Compute Savings Plan commitment would AWS suggest?
aws ce get-savings-plans-purchase-recommendation --savings-plans-type COMPUTE_SP \
--term-in-years ONE_YEAR --payment-option NO_UPFRONT --lookback-period-in-days THIRTY_DAYS \
--query 'SavingsPlansPurchaseRecommendation.SavingsPlansPurchaseRecommendationSummary.[HourlyCommitmentToPurchase,EstimatedSavingsPercentage]'
# gp2 -> gp3: same or better performance, lower price, online
for v in $(aws ec2 describe-volumes --filters Name=volume-type,Values=gp2 \
--query 'Volumes[].VolumeId' --output text); do
aws ec2 modify-volume --volume-id "$v" --volume-type gp3
done
# Alert on unusual spend per service
aws ce create-anomaly-monitor --anomaly-monitor \
'{"MonitorName": "per-service", "MonitorType": "DIMENSIONAL", "MonitorDimension": "SERVICE"}'Remember
- Loop: visibility, waste removal, commitments, architecture
- Savings Plans cover the steady baseline; Spot for interruptible work
- Rightsize with Compute Optimizer; move to Graviton and gp3
- Data transfer and NAT are architecture problems, not purchasing ones
Watch out for
- Buying commitments for peak usage, leaving them under-used when load drops.