Finally the most awaited feature i.e AWS EC2 Predictive Scaling has been announced and launched by AWS. Now you can leverage the benefit of both types of scaling i.e AWS EC2 Predictive Scaling and AWS EC2 Dynamic Scaling. It was a very much needed feature of today as trends are shifting towards AI and ML.
The traditional autoscaling supported by AWS used to be either reactive or proactive. Incase of Reactive Scaling, customers setup manual scaling policies on cpu or network consumption etc. Once that policies are breached EC2 instances are scaled up or down by the control plane. Incase of Proactive Scaling, customers can setup schedules scale in and out policies where on certain scheduled times EC2 instances can be scaled in our out.
With the new advancements in the field of AI and ML, AWS has decided to bring Machine learning on its platform. So what exactly is the term predictive scaling. Predictive scaling is powered by Machine Learning which automatically shrinks or expand fleet of AWS EC2 instances depends on its algorithm without any manual intervention or devopsing. It does this based on customers AWS EC2 usage and data from multiple data points from tens of thousands of servers to prepare its own learning model. These data points are used to build an internal neural network to know average consumption of the fleet and decide when to scale in or out. This model works on historical data from at least one day to start making predictions. In every 1 Day (24 hours), this model is re-evaluated to create a forecasting for the next 48 hours.