I mean data size. It's anecdotally the threshold where "normal" ETL methods start having problems. Obviously performance is hard and the only correct answer is

I mean data size. It's anecdotally the threshold where "normal" ETL methods start having problems. Obviously performance is hard and the only correct answer is "it depends", but I have never seen Spark improve things on datasets that small.