atris commented on issue #854: Shared PQ Based Early Termination for Concurrent 
Search
URL: https://github.com/apache/lucene-solr/pull/854#issuecomment-530675406
 
 
   > There is also the option of something in between. To avoid thread 
contention, could we just make uncoordinated read/write of a minimum minimum 
score. If a thread fails to write (its write is overwritten by some other 
thread with a worse minimum, still an improvement), all it costs you (or some 
other thread) is some extra work, perhaps cheaper than the coordination and 
priority queue or array scanning.
   
   Like a CAS? That can be a good idea.
   
   I wonder if we are over stressing the time complexity of calculating the 
global minimum/maximum. Ideally, the number of slices should not be **too** 
high, so the worst case complexity of the calculation (O(m)) should be still 
ok. Another advantage of this approach is that there is no scope of a thread 
performing rework in case the CAS fails.
   
   
   >You could also choose to write only occasionally, to further reduce 
contention.
   
   That could lead to potentially stale hits being collected?
   
   
   

----------------------------------------------------------------
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: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to