Posted On: Sep 16, 2022
Amazon Elastic Container Service (Amazon ECS) has improved Amazon ECS Capacity Providers to deliver a faster Cluster Auto Scaling experience for scale-in events. Amazon ECS now scales-in excess capacity at a much faster rate, which helps you improve utilization of your infrastructure and saves compute costs.
Amazon ECS is a fully managed container orchestration service that makes it easier for you to deploy, manage, and scale containerized applications. Cluster auto scaling (CAS) is an Amazon ECS capability that manages the Amazon EC2 instances in an Auto Scaling group (ASG) on your behalf, so that you can focus on running your tasks without having to worry about managing infrastructure. With CAS, Amazon ECS manages both scaling-out and scaling-in of Amazon EC2 instances so that your cluster has the optimal capacity for your workloads. During scale-in, Amazon ECS reduces ASG capacity iteratively over several steps to maximize availability while stabilizing capacity to meet demand. Previously, Amazon ECS limited each scale-in step to reduce 5% ASG capacity at a time. With today’s launch, Amazon ECS has increased the scale-in step limit to reduce up to 50% ASG capacity, which translates to fewer scale-in steps and a more responsive scaling-in experience, while still maintaining availability of capacity for spiky traffic patterns. We tested this enhancement for an Amazon ECS service that had scaled-out to 334 c5.xlarge instances for 1000 tasks and required scaling-in to 34 instances for 100 tasks at steady state . The scale-in time period saw an 8x improvement from 80 minutes down to 10 minutes post the scale-in step enhancement. For more details on our performance benchmarking, refer to the blog post.