Containers
Amazon ECS
Amazon ECS orchestrates containers using an AWS-native control plane. It fits custom GenAI APIs and background workers that need container control without Kubernetes compatibility requirements.
Key points
- Services maintain desired task count and integrate with load balancers
- Capacity providers connect placement to Fargate or EC2 scaling
- Task roles give each workload AWS permissions without node credentials
When to use it
- Run asynchronous embedding consumers from SQS
- Host a custom inference gateway with controlled rollouts
Exam tips
- Choose ECS for AWS-native container orchestration and EKS when Kubernetes APIs or ecosystem are required
- Scale on queue depth or request concurrency when CPU does not reflect token workload