I played with Denodo (data virtualization software) a couple years ago and thought it was pretty legit.
In theory, it could be used to provide that industrial strength abstraction layer between your Tableau/Looker/etc. and your bajillion weird and not-so-weird (RDBMS) data sources.
That would seem to make sense to me from the point of view of -- I would want my data visualization/analytics-type company to be able to concentrate on data visualization/analytics, not building some insane and never-ending data abstraction layer.
The part that surprised me was that Denodo could allegedly do a lot of smart data caching, thus speeding things up (esp hadoop-oriented data sources) and keeping costs down.
I'm guessing the other data virtualization providers can do similar.