Posted On: Jan 6, 2023

Amazon EC2 Auto Scaling now produces predictive scaling policy forecasts 4x per day, up from 1x per day prior to now. Reducing our forecast interval from 24 hours to 6 hours produces more accurate predictive scaling policies that quickly adapt to changing demand trends. All customers will benefit from this new default for predictive scaling. Customers use predictive scaling policies to scale out the capacity of their Auto Scaling groups based on forecasted demand, improving application availability, and alleviating the need for costly capacity buffers to accommodate spikes in demand.

Predictive Scaling policies are best-suited for applications that experience repeatable patterns of demand changes - such as daily spikes in user traffic or service demand. Predictive scaling learns from historical demand patterns and scales out capacity in advance of the forecasted demand. Predictive scaling policies are tuned to account for time consuming initialization steps to prepare instances to serve traffic - such as loading gigabytes of data, provisioning services, or running custom scripts. 

Using the Auto Scaling console, customers can validate the forecast accuracy by visually comparing the generated predictive scaling policy against actual demand. Predictive scaling can be enabled by customers through the Amazon EC2 Auto Scaling console, SDK, CLI, AWS CloudFormation, or AWS Cloud Development Kit.

For more details on getting started with predictive scaling and supported region list, refer the documentation.