Auto Scaling - ESS - Supports Predictive Scaling Rules
Target customers: enterprise customers with periodically fluctuating business loads, such as customers in gaming, e-commerce, and online video industries. Feature released: Predictive scaling rules can obtain the historical monitoring data of a scaling group over the last 24 hours or more, and leverage machine learning to predict the monitoring values of the scaling group over the next 48 hours, with a prediction granularity of 60 minutes. Algorithms are used to predict the number of ECS instances needed in the scaling group per hour based on the predicted metric values. Prediction results are updated once a day and 48 prediction tasks are generated, with one prediction task being executed per hour. After a prediction task is executed, the maximum and minimum values of the scaling group are adjusted.