My main point is that there is work behind the scenes happening that you don't have to worry about, on an undefined number of machines that you don't have to care about. Yes, there is still a cluster which does some kind of planning and post-processing now, plus there is obviously a lot of coordination, but it's not like Hadoop or classic Redshift, where you are constrained only to the hardware that you paid for and set up. It's the provisioning in the central phase that resembles BQ the most.