A big thank you goes to Daniel Haviv for his suggestion to use ORC with Snappy compression over Tez (with Vectorised reads) as well as the advice he provided to easily set this up.