Free — no signup required

Introduction: The "AWS Native" Orchestrator

2 min read

What is a container orchestrator?

Imagine you have ten restaurant locations, each needing a specific number of cooks, servers, and cashiers on shift at all times. If someone calls in sick, you need a manager to immediately find a replacement. A container orchestrator is that manager — it ensures the right number of containers are running, restarts them when they crash, and distributes traffic across healthy ones. Amazon ECS (Elastic Container Service) is AWS's built-in version of that manager.

Kubernetes isn't the only way

While EKS (Kubernetes) gets all the hype, Amazon ECS (Elastic Container Service) runs some of the largest workloads on the planet — including Amazon.com's own retail infrastructure. The two services represent fundamentally different philosophies:

  • EKS is Un-Opinionated: It gives you a blank canvas. You choose the networking plugin (CNI), the ingress controller, the logging agent, and the monitoring stack. It is powerful but requires significant assembly and ongoing maintenance.
  • ECS is Opinionated: It is built purely for AWS. It assumes you want CloudWatch for logs, ALBs for load balancing, IAM for security, and ECR for image storage. Every integration is pre-wired. You configure rather than assemble.

This is not a weakness — it is a deliberate design choice. ECS trades flexibility for operational simplicity. A team of three engineers can run a production ECS cluster with confidence. Running EKS at the same scale typically requires dedicated platform engineers.

The Core Hierarchy

ECS has three layers, each mapping to a familiar concept:

ECS Concept Equivalent Purpose
Task Definition Docker Compose file Blueprint: what to run and how
Task Running container (Pod) The actual running instance
Service ReplicaSet / Deployment Keeps N tasks alive, handles rollouts

Understanding this hierarchy is the foundation for everything else in ECS. A Service runs Tasks. Tasks are created from Task Definitions. You never skip a layer.

This is one of 18 chapters

Get every chapter — Kubernetes, Terraform, SRE, distributed systems, and more — with fast daily review built in.

See pricing