Github user andrewor14 commented on the pull request:
https://github.com/apache/spark/pull/1106#issuecomment-54670726
@WangTaoTheTonic I looked at this more and I think it will actually be
slower with the new changes. Before this patch we shuffle all the workers only
once, but here we shuffle it for each waiting driver. If we have thousands of
workers, shuffling can become fairly expensive.
As for the original problem you set out to solve, it seems to me that the
existing code is already doing the randomization correctly. This is because we
draw the first worker from a list of *shuffled* workers, so it will already be
random by the time we draw from 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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
