Hey, Which option is better from performance point of view? 1) Flattening nested json objects (with nested array of jsons) in one single table leading to significant data replication. 2) Splitting it to multiple tables and querying via (inner) joins We are using multi stage engine.

It depends on your requirements and the tradeoffs you are willing to take (space vs time)

Does joins scale well in Pinot as the table size increase? We like the clean approach with joins. They are quite fast with small tables. I don’t know/haven’t checked how will they scale with big tables.

Joins is relatively new, but our goal is to ensure to make it fast and efficient (as long as allowed by laws of physics - a 10Mx10M cross join won’t magically be fast).