Optimizing Apache Spark on Databricks