Today, AWS announces support for predictive scaling for Amazon Elastic Container Service (Amazon ECS). Predictive scaling leverages advanced machine learning algorithms to proactively scale your Amazon ECS services ahead of demand surges, reducing overprovisioning costs while improving application responsiveness and availability.
Amazon ECS offers a rich set of service auto scaling options, including target tracking and step scaling policies, that automatically adjust task counts in response to observed load, as well as scheduled scaling to manually define rules to adjust capacity for routine demand patterns. Many applications observe recurring patterns of steep demand changes, such as early morning spikes when business resumes, wherein a reactive scaling policy can be slow to respond. Predictive scaling is a new capability that harnesses advanced machine learning algorithms, pre-trained on millions of data points, to proactively scale out ECS services ahead of anticipated demand surges. You can use predictive scaling alongside your existing auto scaling policies, such as target tracking or step scaling, so that your applications scale based on both real-time and historic patterns. You can also choose a “forecast only” mode to evaluate its accuracy and suitability, before enabling it to “forecast and scale“. Predictive scaling enhances responsiveness and availability for applications with recurring demand patterns, while also reducing the operational effort of manually configuring scaling policies and the costs from overprovisioning.
You can use AWS management console, SDK, CLI, CloudFormation, and CDK to configure predictive auto scaling for your ECS services. For a list of supported AWS Regions, see documentation. To learn more, visit this blog post and documentation.
Source:: Amazon AWS