This is a great post demonstrating the wide variety of things you can do with postgres foreign data wrappers. But I've worked on analytics systems built on thi

This is a great post demonstrating the wide variety of things you can do with postgres foreign data wrappers. But I've worked on analytics systems built on this and it should just be clear that when you use this kind of system you're accepting ease of operation/ease of integration for actual interactive-level performance. At some point if you want interactive-level performance on this historic data (which is very much possible) you'll have to start putting this data in clickhouse or elasticsearch and some system on top of postgres with a better sharing scheme. If you wanted better performance without giving up the postgres frontend it would (to me) make more sense for folks to invest in building FDWs for elastic or clickhouse rather than over parquet or csv. But that's just a theory.