Github user JoshRosen commented on the pull request:
https://github.com/apache/spark/pull/1648#issuecomment-51253115
I've spent a bit of time looking into some of the performance issues that
we've seen in this patch.
After this patch, it looks like some of the `mapPartitions` stages in the
`correlationoptimizer.*` tests are taking ~7 seconds instead of a few tens of
milliseconds (these were called from the SparkSQL Exchange operator).
@marmbrus, maybe we should chat about this, since you're more familiar with
that code.
Operating under the theory that deserializing Hadoop Configuration /
JobConfs was expensive, I tried a few alternative serialization approaches,
including using WritableUtils to manually serialize the configuration and
writing my own code to read that back into a configuration; this didn't seem to
make a huge difference.
I'm going to put this fix on hold for now until I have more time to figure
out why we're seeing this slowdown.
@ash211 Do you have a way to reliably reproduce the thread-safety issues
that you reported in
[SPARK-2546](https://issues.apache.org/jira/browse/SPARK-2546)? That would be
helpful in order to know whether I've actually fixed the problem with `clone()`.
---
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]
