Watch this video on YouTube
Where in the Spark UI can one diagnose a performance problem induced by not leveraging predicate push-down?