Amazon ECS Service Auto Scaling Now Scales Faster

What's new

Amazon Elastic Container Service (ECS) service auto scaling has been updated to react to changes in load much more quickly. By leveraging 20-second metric resolutions for CPU and memory utilization (down from the previous 60-second default) and optimizing metric publishing, ECS can now trigger scale-out events significantly faster. AWS's internal tests show scale-out triggers improving from over 5 minutes to under 1.5 minutes, and the full scaling and provisioning process dropping from over 6 minutes to under 2 minutes.

Why it matters

This is a substantial improvement for applications with spiky or unpredictable traffic patterns. Previously, a sudden surge in demand could lead to a noticeable degradation in performance before auto scaling caught up. The reduced latency means your services can maintain better availability and responsiveness under stress, potentially allowing you to run with leaner baseline capacity and save on costs. For stateful applications or those sensitive to latency during scaling events, this reduction in time-to-scale is critical.

How to use it

Configure 20-second resolution for CPU or memory utilization metrics when creating or updating your ECS service via the AWS Console, CLI, CloudFormation, or SDKs.