Github user WeichenXu123 commented on the issue:
https://github.com/apache/spark/pull/16574
@mridulm
Year, I know you are worried about the shuffling cost here. Currently when
`spark.shuffle.reduceLocality.enabled` is true(by default), each shuffling
reducer will be launched on the node with the largest outputs. So in this PR
implementation it will generate good data-locality so that its network transfer
cost is similar to current `NarrowDependency` implementation, IMO.
BUT, you mention that Cartesian has more efficient way to implement using
shuffling... I would like to research about it and consider better solution.
Thanks!
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
