Amazon ECS introduces new high-resolution metrics for faster service auto scaling | Amazon Web Services (opens in new tab)
Amazon ECS now supports 20-second high-resolution metrics for faster service auto scaling. AWS reports that scale-out triggers improved from 363 to 86 seconds, while total scaling and task provisioning dropped from 386 to 109 seconds. The update helps applications handle traffic spikes more reliably, reduce excess baseline capacity, and achieve aggressive scaling with simpler target-tracking policies. ## Faster Scaling with High-Resolution Metrics - ECS service auto scaling can use predictive, scheduled, or reactive target-tracking policies. - New 20-second metrics allow ECS to detect workload changes faster than standard 60-second metrics. - AWS benchmarking showed: - 76% faster scale-out triggering, or 4.2× improvement. - 72% faster total scaling and task provisioning, or 3.5× improvement. - Faster scaling can reduce latency and failures during sudden demand increases. ## Lower Costs and Simpler Configuration - Applications may reduce baseline task counts because capacity can be added quickly during spikes. - This can lower compute costs without sacrificing availability. - High-resolution target tracking can provide aggressive scaling behavior that previously required custom step-scaling policies. ## Configuration and Supported Metrics - Enable high-resolution metrics for the ECS service, then configure a target-tracking scaling policy. - New options include: - `ECSServiceAverageCPUUtilizationHighResolution` - `ECSServiceAverageMemoryUtilizationHighResolution` - The feature works with AWS Fargate, ECS Managed Instances, and Amazon EC2. - It can be configured through the ECS console, AWS SDKs, AWS CLI, Application Auto Scaling, or CloudFormation. - Existing services must first enable high-resolution metrics and complete deployment before their scaling policy can be updated. ## Cost and Availability - The ECS feature itself has no additional charge. - High-resolution CloudWatch metrics incur additional CloudWatch costs, unlike standard 60-second metrics. - The capability is available now. For workloads with unpredictable or rapid traffic changes, enabling high-resolution metrics with target tracking can improve responsiveness while reducing the need for overprovisioned ECS tasks.