GitHub user vgankidi opened a pull request:
https://github.com/apache/spark/pull/19425
[SPARK-22196][Core] Combine multiple input splits into a HadoopPartition
## What changes were proposed in this pull request?
Spark native read path allows tuning the partition size based on
spark.sql.files.maxPartitionBytes and spark.sql.files.openCostInBytes. It would
be useful to add a similar functionality/behavior to HadoopRDD, i.e, pack
multiple input splits into a single partition based on maxPartitionBytes and
openCostInBytes. We have had several use-cases to merge small files by
coalescing them by size to reduce the number of tasks launched.
## How was this patch tested?
Added a unit test. It was also tested manually in a few production jobs.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/vgankidi/spark SPARK-22196
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/19425.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 #19425
----
commit 2f4e32681e50d4b42ed5b3d05d91e45483679bee
Author: Vinitha Gankidi <vgank...@netflix.com>
Date: 2017-10-04T06:36:56Z
[SPARK-22196][Core] Combine multiple input splits into a HadoopPartition
----
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
