[SPARK-2503] Lower shuffle output buffer (spark.shuffle.file.buffer.kb) to 32KB.

This can substantially reduce memory usage during shuffle.

Author: rxin

core/src/main/scala/org/apache/spark/shuffle/sort/SortShuffleWriter.scala

@@ -40,7 +40,7 @@ private[spark] class SortShuffleWriter[K, V, C](
   private val ser = Serializer.getSerializer(dep.serializer.orNull)
 
   private val conf = SparkEnv.get.conf
-  private val fileBufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 100) * 1024
+  private val fileBufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 32) * 1024
 
   private var sorter: ExternalSorter[K, V, _] = null
   private var outputFile: File = null

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

@@ -73,7 +73,7 @@ class ShuffleBlockManager(blockManager: BlockManager) extends Logging {
   val sortBasedShuffle =
     conf.get("spark.shuffle.manager", "") == classOf[SortShuffleManager].getName
 
-  private val bufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 100) * 1024
+  private val bufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 32) * 1024
 
   /**
    * Contains all the state related to a particular shuffle. This includes a pool of unused

core/src/main/scala/org/apache/spark/util/collection/ExternalAppendOnlyMap.scala

@@ -107,7 +107,7 @@ class ExternalAppendOnlyMap[K, V, C](
   private var _memoryBytesSpilled = 0L
   private var _diskBytesSpilled = 0L
 
-  private val fileBufferSize = sparkConf.getInt("spark.shuffle.file.buffer.kb", 100) * 1024
+  private val fileBufferSize = sparkConf.getInt("spark.shuffle.file.buffer.kb", 32) * 1024
   private val keyComparator = new HashComparator[K]
   private val ser = serializer.newInstance()
 

core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala

@@ -83,7 +83,7 @@ private[spark] class ExternalSorter[K, V, C](
 
   private val conf = SparkEnv.get.conf
   private val spillingEnabled = conf.getBoolean("spark.shuffle.spill", true)
-  private val fileBufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 100) * 1024
+  private val fileBufferSize = conf.getInt("spark.shuffle.file.buffer.kb", 32) * 1024
 
   // Size of object batches when reading/writing from serializers.
   //

docs/configuration.md

@@ -258,7 +258,7 @@ Apart from these, the following properties are also available, and may be useful
 </tr>
 <tr>
   <td><code>spark.shuffle.file.buffer.kb</code></td>
-  <td>100</td>
+  <td>32</td>
   <td>
     Size of the in-memory buffer for each shuffle file output stream, in kilobytes. These buffers
     reduce the number of disk seeks and system calls made in creating intermediate shuffle files.