ECS – Auto Scaling (for both Fargate and Ec2 instance type)

Automatically increases or decreases the number of running ECS tasks based on demand.

More traffic → more tasks. Traffic drops → tasks removed. Cost and performance managed automatically.

What Triggers Scaling

ECS watches these CloudWatch metrics:

Strategy	How It Works	Best For
Target Tracking	Set a target metric value, ECS maintains it automatically	Simplest — set and forget
Step Scaling	CloudWatch Alarm crosses threshold → add/remove specific number of tasks	More control over each step
Scheduled Scaling	Scale at a specific date and time	Predictable traffic patterns

Examples:

Target Tracking: Keep average CPU at 60% — add tasks if above, remove if below
Step Scaling: CPU > 80% → add 2 tasks. CPU > 90% → add 5 tasks
Scheduled: Every Friday 6 PM traffic doubles → schedule 5 extra tasks at 5:55 PM

This is a common confusion point.

	ECS Service Auto Scaling	EC2 Auto Scaling
What scales	Number of ECS Tasks (containers)	Number of EC2 instances (servers)
Level	Task level	Infrastructure level
Configured where	When creating the Service	During Cluster creation

They are completely separate. Adding more tasks does not add more EC2 instances, and vice versa.