Agreed. I'm also a bit frustrated that he never tuned vertica in his test of it - he just loaded the data into the default superprojection and queried it like that. Nearly all of vertica's power comes from its concept of projections - just like normal DBs benefit from indexes. I'd really like to see how it performs to these other systems once it's been tuned properly because in my experience it's always been the fastest DB I've ever worked with.