gianm commented on issue #19344: URL: https://github.com/apache/druid/issues/19344#issuecomment-4266616139
I guess "TopN Ranking (24h)" is the one you're worried about? The other cases seem to be the same or faster in Druid 37 (on JDK 21) vs Druid 27. The TopN case is likely suffering from not being vectorized. (None of the `topN` query algorithms are currently vectorized.) With numeric vector selectors (`VectorValueSelector`) there is a `getNullVector` method that is called once per row batch, and returns `null` itself to indicate that no nulls exist in the batch. The vectorized query engines take advantage of that to skip null handling for the entire batch. So, vectorizing topN would be very helpful for this one. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
