Elastic Compute Service (ECS) Operations
Instance lifecycle
ECS instances are defined by instance type (vCPU/memory ratio), image, system disk, network type, and security group. Know the lifecycle: create → start → stop → restart → release. Stopping a subscription instance retains it; stopping a pay-as-you-go instance may still incur storage charges. Releasing destroys the instance and, by default, its system disk. Enable release protection for critical instances.
Images and snapshots
A custom image captures the OS and data of a system disk for repeatable deployment. A snapshot is a point-in-time backup of a disk; you can create a custom image from a snapshot. Snapshots are incremental after the first full snapshot. Traps: assuming a snapshot restores individual files (it restores a disk), or that an image includes data disks (it does not by default).
Access and bootstrapping
Use SSH key pairs (preferred) or passwords for Linux; User Data scripts run at first boot for automation. Instance metadata provides instance identity without hardcoding credentials.
Scaling and placement
Auto Scaling groups use scaling rules and scheduled or metric-based triggers. Deployment sets control placement for latency or availability. Dedicated hosts satisfy licensing or compliance isolation needs.
Worked rule of thumb
For a stateless web tier: put instances behind a load balancer, spread them across zones via a scaling group, bake configuration into a custom image, and pass environment specifics through User Data. This makes replacement instances identical and fast to launch.
Exam framing
Distinguish stop from release, image from snapshot, and scaling group from deployment set - these pairings drive most operational questions.