shatejas commented on PR #14076:
URL: https://github.com/apache/lucene/pull/14076#issuecomment-2558748026

   > e.g. when search and merges are running at the same time
   
   @jimczi, https://github.com/apache/lucene/pull/13985 was put out after a 
discussion since there was the this tradeoff being made. You can find the 
discussion here 
https://github.com/apache/lucene/issues/13920#issuecomment-2419196663. 
   
   > using multiple read advices on the same file can lead to unpredictable 
behavior—it might seem fine until you test it in a constrained setup.
   
   The 
[experiments](https://github.com/opensearch-project/k-NN/issues/2134#issuecomment-2420793541)
 seemed to suggest it picked the last advice. I understand the setup wasn't 
constrained, would you be able to share the details of constrained setup where 
this caused unpredictable behavior.
   
   > That said, we could still improve merge performance with a RANDOM read 
advice in the future, for instance, by adding eager prefetching.
   
   Are we sure that introducing prefetch during merge won't impact search?
   
   I understand the main point of contention, before reverting the change would 
want to see if there is a path forward for optimizing merge or if there is a 
middle ground.
   
   A thought here would be to see the behavior if the read advice is updated 
(and revert after merge is over) to normal since sequential read advice can be 
aggressive as [per 
documentation](https://man7.org/linux/man-pages/man2/madvise.2.html). Reading 
ahead of pages will then be decided by kernel heuristics for normal read advice 
as per my understanding.
   
   @uschindler @jpountz any thoughts or suggestions?


-- 
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: issues-unsubscr...@lucene.apache.org

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


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

Reply via email to