ECS Service Auto Scaling

• Dynamically adjusts the desired number of running ECS tasks based on load or metrics.

• Powered by AWS Application Auto Scaling.

Key Metrics for Scaling

• ECS Service Average CPU Utilization

• ECS Service Average Memory Utilization

• ALB Request Count per Target (from Application Load Balancer)

Scaling Methods

• Target Tracking Scaling – Keeps a CloudWatch metric at a specific target value.

• Step Scaling – Responds to CloudWatch alarms with scaling actions based on thresholds.

• Scheduled Scaling – Runs scaling actions at fixed times/dates (good for predictable traffic patterns).

Scope of Scaling

• ECS Service Auto Scaling → Adjusts task count.

• EC2 Auto Scaling → Adjusts container instance count (EC2 launch type).

• Fargate Auto Scaling → Only adjusts task count (no instance scaling needed).

EC2 Launch Type – Auto Scaling EC2 Instances

Auto Scaling Group (ASG)

• Manages the number of EC2 instances in the cluster.

• Scales based on CloudWatch metrics such as CPU utilization.

• Ensures additional EC2 instances are launched when needed to run more tasks.

ECS Cluster Capacity Provider

• Integrates ECS with an ASG for automated infrastructure scaling.

• Monitors ECS task capacity needs (CPU, RAM).

• Automatically provisions EC2 instances when there’s insufficient cluster capacity.

ECS Scaling – CPU Usage Example

1. CloudWatch Metric
• ECS service reports Average CPU Utilization to CloudWatch.

2. CloudWatch Alarm
• Alarm triggers when CPU exceeds a set threshold (e.g., 70%).

3. Service Auto Scaling Action
• Policy adds new ECS tasks (e.g., from 2 to 3 tasks) to handle increased demand.

4. Infrastructure Scaling (Optional)
• If EC2 launch type is used and the cluster lacks resources,
the Capacity Provider + ASG launches more EC2 instances.