Predictive Scaling and Capacity Planning With Machine Learning at Amazon.com

Explore how Amazon.com handles massive traffic spikes during high-velocity events like Prime Day, Black Friday, and Diwali through advanced predictive scaling and capacity planning techniques. Learn from Principal Engineer Artur Souza and Senior Applied Scientist Chunpeng Wang as they reveal how machine learning algorithms and strategic record-keeping enable proactive scaling of thousands of services running on AWS infrastructure. Discover the limitations of traditional reactive scaling approaches that respond to CPU utilization and traffic metrics, and understand how predictive methods prevent customer-facing errors such as high latency and checkout failures during peak demand periods. Gain insights into the sophisticated systems that automatically scale services up before anticipated traffic surges and then intelligently de-scale back to forecasted loads for normal business operations, ensuring optimal customer experience while maintaining cost efficiency across Amazon's massive cloud native architecture.