GitHub user viirya opened a pull request:
https://github.com/apache/spark/pull/5572
[SPARK-6307][Core] Speed up RDD.cartesian by caching remotely received
blocks
JIRA: https://issues.apache.org/jira/browse/SPARK-6307 and
https://issues.apache.org/jira/browse/SPARK-6922
The problematic cases are stated clearly in the related JIRAs.
Currently, `CacheManager.getOrCompute` only caches locally computed values.
If a required block is retrieved remotely, it will not be cached. In most use
cases, this is working well because the same block will not be required in
multiple times.
However, `CartesianRDD` hits this problem and it will require same blocks
in multiple times to compute the cross product split. Thus, the same blocks are
fetched again and again and the performance is very slow.
This pr tried to cache the remotely received blocks if enabled by
configuration.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/viirya/spark-1 cache_remote_block
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/5572.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5572
----
commit 7e1bb1331d20881ab72a4ee0b55ebc8ab73cdf86
Author: Liang-Chi Hsieh <[email protected]>
Date: 2015-04-18T14:12:44Z
Try to cache remotely received block if enabled by configuration.
----
---
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]
