snleee commented on pull request #7492:
URL: https://github.com/apache/pinot/pull/7492#issuecomment-930443143


   @mqliang Our old behavior is to use `min(numOperator, 
MAX_NUM_THREADS_PER_QUERY)` for all incoming queries. So, even if the user 
tries to be greedy (I hope this is not the case in reality), they will get the 
same behavior as before. I think that we can consider enabling the `server` 
level config by default for our large MT cluster given that the latency 
requirement is not too strict. This can protect the cluster from a single super 
expensive query. In our current scheme, one super-expensive query can bring 
down the cluster. I think that it's already a good start.
   
   Yes, if we want to solve the problem of "fairness" among all clients for our 
MT cluster resource, a more sophisticated scheduling algorithm will be needed 
and this would be a complex research problem. 


-- 
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]

Reply via email to