This message was deleted.

I have seen larger segments using Range when the data skew throws off the algorithm a bit. I would also check segment scan times when you run queries to target 250ms or less ... if the scan times are larger then maybe try to drop your segment size down a bit.

I would say either is OK as a starting point; they aren't that different The main guidance is you don't want ones that are super small or super big

@Srikanth Kantamaneni We designed sql based ingestion so that the user does not have to think about segments sizes rite off the bat. The defaults should be good enough for most use cases. I would say please try out : https://druid.apache.org/docs/latest/multi-stage-query/reference.html#replace that would help you range partition your datasource quite easily.