Apache Spark @Scale: A 60 TB+ production use case Through a series of performance and reliability improvements, we were able to scale Spark to handle a TB-scale entity ranking system in production. Click to view the original at code.facebook.com
