rkhachatryan commented on pull request #18989:
URL: https://github.com/apache/flink/pull/18989#issuecomment-1064008740


   Thanks for the review @Myasuka.
   I've updated the PR, PTAL.
   
   > What do you mean here?
   > I've found experimentally that the distribution isn't that much different 
(true random being slightly better than Murmur3 hashing).
   > Is that to say, new Random().nextInt((int) periodicMaterializeDelay) 
behaves better than current implementation of this PR?
   
   I generated offset numbers using the two approaches and then analyzed the 
distributions using K-MEANS.
   `new Random().nextInt((int) periodicMaterializeDelay)` creates slightly less 
clusters. 
   So the answer to your question is yes, persumably it would work silghtly 
better. On the downside, it would be more susceptive to bad PRNG and not 
reproducible to some degree.


-- 
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