Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
Pig >> mail # dev >> Review Request 15261: PIG-3555 Initial implementation of Tez combiner optimization

Cheolsoo Park 2013-11-06, 11:04
Cheolsoo Park 2013-11-06, 17:00
Copy link to this message
Re: Review Request 15261: PIG-3555 Initial implementation of Tez combiner optimization

This is an automatically generated e-mail. To reply, visit:

(Updated Nov. 6, 2013, 11:55 p.m.)
Review request for pig, Alex Bain, Daniel Dai, Mark Wagner, and Rohini Palaniswamy.

Upload a new patch that includes the following changes-
* Adds two Map<OperatorKey, TezEdgeDescriptor>'s to TezOperator.
* Adds combine plans to outbound (map/onfilesortedoutput) instead of inbound (reduce/shufflemergeinput). This is the same as MR-Pig.
* Adds a few Pig-specific properties to the edge payload to make PigCombiner work.

I still have to go through Mark's comments, but with this patch, combiners seem to work now. I can see counters in task logs as follows-

Combine input records=3, Combine output records=8
Bugs: PIG-3555
Repository: pig-git

Initial implementation of Tez combiner optimizer. The patch includes the following changes-
* Factored out CombinerOptimizer code into a utility class called CombinerOptimizerUtil. So both MR and Tez CombinerOptimizer use this utility class instead of duplicating code.
* Introduced a new class called TezEdgeDescriptor that holds combine plans as well as various edge properties.
* Added TezEdgeDescriptors to TezOperator. Note that I added multiple descriptors for inbound edges but a single descriptor for all the outbound edges. This is because TezDagBuilder always creates an edge by connecting predecessors to the current vertex. Please let me know if you think we should allow multiple descriptors for outbound edges too.
* Refactored some code in TezDagBuilder while touching it.
Diffs (updated)

  src/org/apache/pig/backend/hadoop/executionengine/mapReduceLayer/CombinerOptimizer.java 18a382b
  src/org/apache/pig/backend/hadoop/executionengine/tez/CombinerOptimizer.java e69de29
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezCompiler.java 0b1f3c9
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezDagBuilder.java 45e47b0
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezEdgeDescriptor.java e69de29
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezLauncher.java 3f14644
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezOperator.java e612d88
  src/org/apache/pig/backend/hadoop/executionengine/tez/TezPrinter.java 5a42ded
  src/org/apache/pig/backend/hadoop/executionengine/util/CombinerOptimizerUtil.java e69de29
  test/org/apache/pig/test/data/GoldenFiles/TEZC1.gld 925f07e
  test/org/apache/pig/test/data/GoldenFiles/TEZC2.gld a3974fe
  test/org/apache/pig/test/data/GoldenFiles/TEZC3.gld a8c942b
  test/org/apache/pig/test/data/GoldenFiles/TEZC4.gld fb7c903
  test/org/apache/pig/test/data/GoldenFiles/TEZC5.gld e6cd25e

Diff: https://reviews.apache.org/r/15261/diff/

ant test-tez passes.
ant test-e2e-tez passes.

I didn't add new test cases, but an e2e test case (Checkin_3) includes an algebraic udf (count) following group-by. I also manually tested it on a live cluster.

Cheolsoo Park

Alex Bain 2013-11-08, 00:55
Cheolsoo Park 2013-11-08, 07:37
Cheolsoo Park 2013-11-08, 07:57
Rohini Palaniswamy 2013-11-08, 23:12
Cheolsoo Park 2013-11-09, 00:33
Cheolsoo Park 2013-11-10, 04:31
Mark Wagner 2013-11-11, 21:16
Cheolsoo Park 2013-11-11, 21:41
Mark Wagner 2013-11-06, 22:54
Cheolsoo Park 2013-11-08, 07:37
Rohini Palaniswamy 2013-11-06, 19:46
Rohini Palaniswamy 2013-11-06, 22:15
Cheolsoo Park 2013-11-06, 20:55