Show HN: Faster, Cheaper Databricks Spark with Real-Time ML Config Management

The most counterintuitive thing we found doing this is how often jobs (especially old jobs) are undersized. You can run faster and cheaper if you're running out of resources (especially memory).