Github user viirya commented on the pull request:
https://github.com/apache/spark/pull/2217#issuecomment-54396114
Thanks for your comments. I agree with Reynold too that it is a rough idea
and I need much consideration with a clear design doc for that if I think we
really need that primitive. I will close this PR.
I know that there are many possible workarounds for that purpose such as
@shivaram's suggestion, global static variable, etc. But these workarounds will
need collect data from executors and re-send (re-broadcast, etc.) to executors
again in next iteration. That is the main obstacle I want to overcome in order
to save the communication cost, especially if the data to collect and re-send
totally is huge.
---
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]
