[GitHub] [druid] suneet-s edited a comment on issue #9321: Performance degradation in topN queries when SQL-compatible null handling is enabled

[GitBox]

suneet-s edited a comment on issue #9321: Performance degradation in topN 
queries when SQL-compatible null handling is enabled
URL: https://github.com/apache/druid/issues/9321#issuecomment-591551174
 
 
   Thanks for the flamegraphs! I read the code last night and based on the 
flamegraphs, the best thing I can think of that might be happening is maybe 
something in the hotspot is less optimal after the group by query is run. The 
buffers are shared across the TopNQueryEngine and the GroupByQueryEngine, so 
somehow running a group by query is impacting the group by performance.
   
   Since it's this low level, I'd also question the reality of the scenario. It 
might be less visible if the performance tests were running a mixture of 
queries in both runs.
   
   **My guess at this point is this an edge case, and I don't think we'd see a 
huge difference if this were a mixed load scenario**
   
   I'm also attaching screenshots from your flamegraphs that highlight what you 
pointed out, so others with more knowledge than me can either confirm these 
suspicions or suggest what might actually be happening.
   When only topN queries were run - the time spent in 
NullableNumericBuffer#aggregate was mainly spent in reading the bytes (LZ4...)
   <img width="1423" alt="topN-after-groupBy" 
src="https://user-images.githubusercontent.com/44787917/75371559-d2423f00-587b-11ea-819b-96502be7012f.png";>
   
   After a groupBy query was run, and topN queries were profiled - 
NullableNumericBuffer#aggregate has a bunch of time spent where it isn't 
reading the bytes, but doing something else
   <img width="1428" alt="topN-before-groupBy" 
src="https://user-images.githubusercontent.com/44787917/75371555-cfdfe500-587b-11ea-98ca-d0ef75d6d7ee.png";>
   
   

----------------------------------------------------------------
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.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org