pltbkd commented on PR #28305:
URL: https://github.com/apache/flink/pull/28305#issuecomment-5103652879

   I've updated the PR with the new semantics. Please take another look.
   
   I also ran the benchmark with the updated version for another 10 rounds * 5 
iterations:
   Group | Scenarios | Avg Diff | Range (±SE)
   -- | -- | -- | --
   With network exchange | ORDERED | +0.25% | ±1.9%
   With network exchange | UNORDERED | -0.18% | ±2.1%
   Without network exchange | ORDERED | +1.09% | ±2.2%
   Without network exchange | UNORDERED | +1.53% | ±4.2%
   
   Compared with the previous run I posted earlier, the diffs didn't reproduce 
and everything is now within the noise range, in the opposite direction in some 
cases. So the deviations in the previous run were most likely environment 
fluctuations rather than the change itself. Combining both runs, I think it's 
fair to say there is no observable regression, with or without a downstream 
network exchange. @1996fanrui @pnowojski 


-- 
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]

Reply via email to