AWS cost engineering notes
What we learn running AWS cost pipelines in production. Billing data, Kubernetes, networking, storage, and whatever else the bill throws up. Every figure is checked against AWS documentation and linked, including the parts that do not work.
-
EKS / SCAD / Kubernetes
EKS split cost allocation data: the two choices that change your numbers (2026)
How to enable split cost allocation data, which of the three EKS measurement options to pick and why it matters, and the Fargate-derived 9:1 weighting AWS uses to price your EC2 instance.
-
EKS / Kubernetes / Discovery
How to find every Kubernetes cluster in an AWS account (2026)
aws eks list-clusters only returns EKS clusters. The three populations you actually have, the API flag that surfaces connected GKE and AKS clusters, the EC2 tag scan for self-managed ones, and the dedup step that stops you double-counting.
-
CUR 2.0 / FOCUS / EKS
CUR 2.0 vs FOCUS: why only one of them can cost your pods (2026)
A field-level comparison of the two AWS Data Exports formats, taken from the AWS column dictionaries: what CUR 2.0 has that FOCUS does not, every split cost allocation column, and the four gotchas that bite when you build on them.
-
Network cost / VPC
What a NAT Gateway actually costs, and what regional mode changed (2026)
What AWS charges for a NAT Gateway, sourced from the VPC pricing page, why one zonal gateway across three AZs costs more, and how the regional availability mode added in November 2025 changes it.