Is Apache Spark Too Costly? An Amazon Engineer Tells His Story

The New Stack Podcast

Is Apache Spark too costly? Amazon Principal Engineer Patrick Ames tackled this question during an interview with The New Stack Makers, sharing insights into transitioning from Spark to Ray for managing large-scale data. Ames, described as a "go-to" engineer for exabyte-scale projects, emphasized a goal-driven approach to solving complex engineering problems, from simplifying daily chores to optimizing software solutions.

Initially, Spark was chosen at Amazon for its simplicity and open-source flexibility, allowing efficient merging of data with minimal SQL code. The team leveraged Spark in a decoupled architecture over S3 storage, scaling it to handle thousands of jobs daily. However, as data volumes grew to hundreds of terabytes and beyond, Spark’s limitations became apparent. Long processing times and high costs prompted a search for alternatives.

Enter Ray—a unified framework designed for scaling AI and Python applications. After experimentation, Ames and his team noted significant efficiency improvements, driving the shift from Spark to Ray to meet scalability and cost-efficiency needs.

Learn more from The New Stack about Apache Spark and Ray:

Amazon to Save Millions Moving From Apache Spark to Ray

How Ray, a Distributed AI Framework, Helps Power ChatGPT

Join our community of newsletter subscribers to stay on top of the news and at the top of your game.