I've been working on a new company that's doing cloud hosted postgres foreign data wrappers as a service (plus a bunch of interface / workflow stuff on top) and

I've been working on a new company that's doing cloud hosted postgres foreign data wrappers as a service (plus a bunch of interface / workflow stuff on top) and while I'm biased I totally agree. The postgres FDW system has gotten very sophisticated in later releases. It does performance optimizations like pushing down filters / aggregates, and it even goes further by supporting sending remote explain calls out to the underlying datasources to use in query planning. In one of my local tests vs. Denodo it smartly decided _not_ to push a group by down when the cardinality of the column being grouped on is very high (since the grouping doesn't reduce the number of rows sent over the wire very much), which blew me away. If anyone is interested I'd love to chat: steve (at) syndetic.co