Comment by dogman123

This could be incredibly useful for me. Currently struggling to complete jobs with massive amounts of shuffle with Spark on EMR (large joins yielding 150+ billion rows). We use Glue currently, but it has become cost prohibitive.