The basic analysis and behaviors will remain same in terms of applying your logic to optimize the performance. But Spark 3 version is highly optimized and way faster compare to Spark 2.x in few areas. 46% of the deployment were on Spark SQL area to make it more efficient and fast. Few concepts like Dynamic Partition Pruning, Adaptive Query Execution, Optimizer Hints have been implemented which make the overall execution faster. I've prepared a small overview covering these features which you can refer @ https://medium.com/@ch.nabarun/whats-new-in-spark-3-8250a65b3144