I'm working on a survey project running on RDS Postgres in which we're trying to get our queries down from seconds to milliseconds, and I'm investigating swapping our database from Traditional Relational, to a service which will be able to run and scale queries and accurate aggregations.

The obvious choice seemed to be Redshift, but currently Postgres is struggling most on a query along a many to many relation. Being: of these people, select those in the following demographic groups (is male, in Chicago, in accounting, etc) for which each person many have hundreds of demographic ids describing them in the through table between people and demographics. Im worried that redshift may also have the same issues querying so much data without denormalization, which is something I'd like to avoid for code complexity sake.

I know much less about graph databases. They seem to really champion queries like these on related data, however, I'm more hesitant to dive into using Neptune first for a PoC due to the friction of tying it into our python webapp.

I was wondering what other's experiences are with either implementing Neptune's graph database in a production setting, or optimizing their redshift database for querying through tables