Posted On: May 19, 2022
Starting today, when you create a new predictive scaling policy, Amazon EC2 Auto Scaling goes back 14 days to generate capacity forecasts for the past dates, enabling you to see how predictive scaling would have scaled your Auto Scaling group. This allows you to quickly decide if the predictive scaling policy is accurate for your applications by comparing the demand and capacity forecasts against the actuals immediately after you create a predictive scaling policy. Previously, you would have had to wait at least a few days after creating the policy to build up sufficient forecast history for the same comparison to determine the forecast accuracy.
Amazon EC2 Auto Scaling is a service that helps you meet application demand by automatically adding or removing EC2 instances to an Auto Scaling group according to the conditions you define. In addition to offering dynamic scaling policies that react to demand changes, Auto Scaling offers you predictive scaling policies that can proactively change the size of your Auto Scaling group in advance of predictable demand spikes. This helps you maintain high application availability and responsiveness without the need to overprovision, resulting in lower EC2 costs. For example, predictive scaling might observe recurring early morning spikes when business resumes after a weekend, and predictively scale out your application every weekday in advance of that traffic so that you don’t have to overprovision your Auto Scaling group overnight. You can start by creating a predictive scaling policy in a “ForecastOnly” mode, in which predictive scaling does not take any scaling actions and only shows you the forecasts. With this release, immediately after creating a policy, you can compare the historical forecast against the actual usage to determine the accuracy of forecasts and then decide to whether to actually enabling predictive scaling in the “ForecastAndScale” mode to start scaling according to the forecasted capacity.