Something related that I found out about from HN a few months back is another engine called quokka. It's particularly interesting and applicable how quokka schedules distributed queries to outperform Spark https://github.com/marsupialtail/quokka/blob/master/blog/why...