Merge pull request apache#399 from pwendell/consolidate-off

Disable shuffle file consolidation by default

After running various performance tests for the 0.9 release, this still seems to have performance issues even on XFS. So let's keep this off-by-default for 0.9 and users can experiment with it depending on their disk configurations.

Author: pwendell

core/src/main/scala/org/apache/spark/storage/ShuffleBlockManager.scala

@@ -64,7 +64,7 @@ class ShuffleBlockManager(blockManager: BlockManager) {
   // Turning off shuffle file consolidation causes all shuffle Blocks to get their own file.
   // TODO: Remove this once the shuffle file consolidation feature is stable.
   val consolidateShuffleFiles =
-    conf.getBoolean("spark.shuffle.consolidateFiles", true)
+    conf.getBoolean("spark.shuffle.consolidateFiles", false)
 
   private val bufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 100) * 1024
 

docs/configuration.md

@@ -382,7 +382,7 @@ Apart from these, the following properties are also available, and may be useful
 
 <tr>
   <td>spark.shuffle.consolidateFiles</td>
-  <td>true</td>
+  <td>false</td>
   <td>
     If set to "true", consolidates intermediate files created during a shuffle. Creating fewer files can improve filesystem performance for shuffles with large numbers of reduce tasks. It is recommended to set this to "true" when using ext4 or xfs filesystems. On ext3, this option might degrade performance on machines with many (>8) cores due to filesystem limitations.
   </td>