We chose ECS over EKS: what we gained and what we gave up

Introduction

We chose ECS over EKS for a production platform, and I still think it was the right call.

Not because Kubernetes was too hard. I run Kubernetes workloads on EKS and on GKE. It was because that platform did not need what Kubernetes charges you for. Here is the honest ledger both columns.

What we gained

No control plane bill

ECS orchestration costs nothing; you pay only for the EC2 or capacity underneath it. EKS is $0.10 per cluster per hour, roughly $73 a month. That is per cluster, and cluster count is the number nobody says out loud every environment, in every region, plus the shared ones. Twelve clusters is $876 a month before a single container starts.

No version treadmill

An EKS Kubernetes version gets 14 months of standard support, then 12 months of extended support. Extended support is enabled by default and costs $0.60 per cluster per hour roughly $438 a month. Fall two upgrade windows behind and the same cluster costs six times more while doing exactly the same work. Those twelve clusters become $5,256 a month, and nobody had to approve anything for that to happen. ECS has no Kubernetes version for you to chase.

Fewer moving parts to secure

A task role per task, a security group per service, a target group on the load balancer. No CNI plugin decision, no admission controllers, no RBAC model layered on IAM. Every component you do not run is a component nobody has to review at audit time.

What we gave up

Portability

A task definition is an AWS artifact. A Kubernetes manifest is not.

The ecosystem

No Helm chart, no operator, no Argo CD, no CRD for the thing you need next quarter. You open an editor instead. That is the point where the trade-off stops being theoretical.

Scheduling

ECS placement strategies are good enough for most workloads. They are not the Kubernetes scheduler, and if you need topology-aware placement or a custom scheduler, you will find the ceiling.

Hiring signal

The market asks for Kubernetes. A job advert is not interested in your reasoning.

Let's Build Something Great Together

Ready to transform your idea into a powerful software solution? Talk to our experts and get a free consultation.

Contact Us

Run it on your own account

Clusters × 730 hours × $0.10 then add the extended-support rate for anything you already know you will not upgrade on time.

If that number is smaller than the cost of the Kubernetes expertise already on your team, EKS is the right call and you can stop reading comparisons. If it is larger, the question is not which service is better. It is whether you are buying capability you will use, or optionality you will never exercise.

Which way did you go and would you make the same call again?