devops24x7

Services

Reducing AWS compute costs without stopping services

When you need it

The compute bill grows faster than the load, nodes are requested "with headroom", and the last attempt to enable autoscaling or consolidation restarted services during working hours.

What you get

An audit: requested versus actually used resources per cluster, spot candidates, Kubernetes versions paying the outdated-version surcharge, image and artifact storage without retention. Then changes as code: Karpenter with consolidation in an agreed window, separate pools for batch jobs, spot pools that service owners opt into, retention policies, a Savings Plans coverage plan. Rolled out cluster by cluster with verification after each. Deliverables: code in your repository, a before/after report on nodes and cost, and rules for your team on what to put on spot and how.

Typical duration

Audit: 1–2 weeks. Implementation across several clusters: 3–6 weeks. Estimate after a first call.

Related work

Work examples 1 and 4; image retention is part of the CI solution.

Contact