I am currenty using Postgres/PostGIS (one single instance) to run some geo queries to catch points that fall within specific shapes along with some extra data from the relative columns.

However even with indexes (and scaling up the db) some of the queries are too slow.

I have tried using GPU databases like BrytLyt which run smoothly on top of Postgres and it solves the problem but unfortunately is way far too expensive (AWS per hour cost).

The total amount of records is ~100M and the queries at the moment are taking 6/7 secs each one (I need ~ 0.1 secs or so).

Can you suggest any database that could suit for the problem? I am trying to avoid some “manual” sharding and pre-calculating values.

Example of structure of the data:

longitude | latitude | price | bedrooms 

So for example the average price for house with 1 bedroom in a specific area.