Github user pwendell commented on the pull request:
https://github.com/apache/spark/pull/2438#issuecomment-62332004
Hey @nfergu - I was looking for any older PR's that have fallen through the
cracks and came across this. This is a very well written patch - kudos! When I
suggested this registry concept initially, I was actually envisioning this
happening in user-space rather than in Spark itself. I think automatically
broadcasting all named accumulators is not going to work because some
applications create thousands of accumulators (e.g. streaming applications),
and it could end up with an unexpected performance regression.
For some applications this might be acceptable though. How hard would it be
for a user-space library to implement this rather than having it be inside of
Spark proper?
---
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]
