It's not about scale, it's about manageability. Having a set of uniform nodes where you schedule containers is nicer, in my opinion, than managing an infrastru

It's not about scale, it's about manageability. Having a set of uniform nodes where you schedule containers is nicer, in my opinion, than managing an infrastructure where you've scripted applications to go places. Sure, you could make puppet manage containers on uniform nodes but then we're having a massive convention vs configuration argument, which we shouldn't bother with because the distributed schedules are giving us a lot of important stuff for mostly free.