Amazon ECS Deployment-Model Design and Selection

Choose the right Amazon ECS compute/launch model for a workload, architect the cluster + services around it, and plan the transition when a customer is moving off an older topology. This is the anchor ECS skill — the decision framework every other ECS skill leans on. It answers two coupled questions:

1. Selection — Which of Fargate, ECS on EC2, ECS Managed Instances, ECS Express Mode, or ECS Anywhere/External fits this workload, and what capacity-provider strategy backs it?
2. Design + migration — How is the task sized, how is the network laid out (awsvpc/ENI density), what service parameters are set, and — if an estate already exists — how does it move from EC2 launch type to capacity providers / Managed Instances, or from Service Discovery to Service Connect?

When to Use

• Picking a compute model for a new containerized workload on ECS ("Fargate or EC2?", "should I use Managed Instances?", "is Express Mode right for this?").
• Designing a capacity-provider strategy (Fargate / FARGATE_SPOT / EC2 ASG / Managed Instances mixes) and base/weight ratios.
• Sizing tasks (CPU/memory combinations, ephemeral storage) and planning awsvpc ENI density on EC2.
• Choosing a networking model (awsvpc, task ENI, load-balancer placement) and core service parameters (min/max healthy percent, health-check grace period, placement).
• Planning a launch-type or topology migration: EC2 launch type → capacity providers or Managed Instances; Service Discovery (Cloud Map) → Service Connect.
• Answering "which model + how to architect it, by criteria" for hybrid/edge (ECS Anywhere) or air-gapped constraints.

Don't Use