I'm actually dealing with the same thing. A request on my local machine against a tiny test dataset is taking 9 seconds to return ~30 rows (which is the entire table) in sorted order. The reason its so slow is because of queries that are overfetching and ORM code scattered throughout backend code, so stuff like "get list of things from DB, then loop over things and do more queries for each one" happens a lot, instead of having a single query join the data it needs. By cutting the columns down to only what was needed and moving some of these extra queries into joins or subqueries, and removing the ones that were unnecessary, I got the 9 second request to complete in about 400ms. And that's without even looking into indexing or other optimisations.