This week, Amazon Web Services is using machine learning to refine its Auto Scaling tool, providing it with predictive abilities. The latest predictive scaling attribute forecasts user’s likely EC2 usage and traffic, containing weekly and daily patterns. It can assist a user to develop a scaling strategy by looking forward weekly and daily peaks.
Predictive scaling’s machine learning (ML) model is trained using the data gathered from the user’s own EC2 usage, along with billions of more data points Amazon sees. The ML model requires at least one-day historical data to begin forecasting. After every 24 hours, the model is re-examined to make the prediction for the coming 48 hours.
In 2006, Amazon EC2 was released and in 2009, Auto Scaling got launched, together with Elastic Load Balancing and CloudWatch Metrics. Chief Evangelist for Amazon Web Services, Jeff Barr, mentioned in a blog post that these two releases “truly signify the fundamentally dynamic, on-demand nature of the cloud.”
Auto Scaling is being updated one week earlier than the annual AWS re:Invent conference. In the previous year, re:Invent conference was utilized to release a swing of latest facilities, with many charged by machine learning. At the beginning of this week, AWS introduced advancements to Amazon Polly, Transcribe and Translate, devices which refine personal products of users by utilizing machine learning.