Github user ChengXiangLi commented on the pull request:
https://github.com/apache/flink/pull/1255#issuecomment-166297616
Sorry, @fhueske , i misunderstood your test data, the keys should be skewed
on some value, while in my previous test, the keys are now skewed. it's
complicate to calculate how many samples should be taken from a dataset to meet
an a priori specified accuracy guarantee, one of the algorithm is described at
http://research.microsoft.com/pubs/159275/MSR-TR-2012-18.pdf which i used
before, but it should not totally fit into the case which keys are skewed.
Would you continue to test how much it required to make partition roughly
balanced? Raise the sample number should not add much overhead, i'm totally
support of it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
