that will work for running sql queries, but it gets trickier when you start to think about stuff like efficiently batching uploads from spark to redshift