[GitHub] [flink] zhijiangW commented on issue #7911: [FLINK-11082][network] Fix the calculation of backlog in PipelinedSubpartition

[GitBox <gi...@apache.org>]

zhijiangW commented on issue #7911: [FLINK-11082][network] Fix the calculation of backlog in PipelinedSubpartition
URL: https://github.com/apache/flink/pull/7911#issuecomment-508718353
 
 
   Thanks for reviews and merging @pnowojski !
   
   As we discussed offline, this PR changes the backlog semantic from the number of producer buffers to the number of consumable buffers. In common sense I think this new semantic is more fit for the initial thought of introducing backlog for credit-based.
   
   Your concern is reasonable.  As we discussed offline, it is hard to determine the specific effects for reporting high backlog size in different cases. If the extra requested floating buffers for high backlog are actually used soon on consumer side, it might have no effects. Otherwise it might bring negative effects because of floating buffers misdistribution. I agree with that there should have some follow-up works for further verifying the conclusion. 

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services