Do you see yourself getting better performance with the Dataframes+ Pipelines infra...or is it simply a better mental model.