AWS CDK Stack Cost Estimator

CDK stack cost estimates are approximations based on public AWS list prices. Actual costs depend on your specific usage patterns, data transfer between services, free tier eligibility, and any existing Reserved Instances or Savings Plans. For production planning, always validate against the AWS Pricing Calculator (calculator.aws) and enable AWS Cost Explorer after deployment to track real spending. This tool uses simplified pricing models — for example, EC2 pricing omits EBS volume costs, and RDS excludes I/O charges for Multi-AZ failover operations.

The highest-impact CDK optimizations are: (1) Reserved Instances or Savings Plans for EC2/RDS — save 30–70% on predictable workloads; (2) Right-sizing instances — many teams over-provision by 2–4× at launch; (3) S3 Intelligent-Tiering for infrequently accessed data; (4) Lambda vs. always-on EC2 for event-driven workloads — Lambda at low invocations is dramatically cheaper; (5) Use CloudFront to cache S3/ALB responses and reduce origin load; (6) DynamoDB on-demand pricing for unpredictable traffic, provisioned for steady loads. Enable AWS Trusted Advisor and Cost Anomaly Detection to catch unexpected spending automatically.

Multi-AZ RDS roughly doubles your instance and storage costs but provides automatic failover (typically < 60 seconds) and synchronous replication to a standby in a different Availability Zone. It makes sense when: (1) downtime cost exceeds the Multi-AZ premium — calculate this as (hourly revenue × expected failure hours × failure probability); (2) your SLA requires 99.95%+ availability; (3) you have compliance requirements for data durability. For dev/staging environments, single-AZ is almost always sufficient. Consider Aurora instead of Multi-AZ RDS for production — Aurora offers Multi-AZ by default with better performance at similar cost.

Reserved Instances (RIs) and Compute Savings Plans commit you to a specific instance type/region (RIs) or compute spend (Savings Plans) in exchange for 30–72% discounts versus On-Demand. 1-year No Upfront RIs save ~30–40%; 3-year All Upfront save 50–70%. For CDK stacks, identify your baseline compute (EC2 instances that run 24/7, RDS instances, Fargate tasks) and apply RIs to those. Use Convertible RIs for flexibility to change instance families as your CDK stack evolves. Spot Instances (not modeled here) offer 70–90% savings for fault-tolerant, interruptible workloads like batch processing or CI/CD runners.

ECS Fargate charges per vCPU-hour ($0.04048) and per GB-hour ($0.004445) with no idle infrastructure. EC2 charges for the entire instance whether it's 10% or 100% utilized. Fargate wins for: bursty workloads with high peak-to-average ratios, small services that don't fill a full EC2 instance, or teams that want to eliminate EC2 patching/management overhead. EC2 wins for: steady-state high-utilization workloads (>70% CPU), workloads needing GPU, large containers that efficiently fill an instance, and cost-sensitive production at scale with Reserved Instances. A common pattern: Fargate Spot for dev/staging (70% savings), Fargate On-Demand for small prod services, EC2 + Savings Plans for large production fleets.