Author: cutting Date: Wed Feb 28 11:59:40 2007 New Revision: 512936 URL: http://svn.apache.org/viewvc?view=rev&rev=512936 Log: HADOOP-1043. Optimize shuffle, increasing parallelism. Contributed by Devaraj.
Modified: lucene/hadoop/trunk/CHANGES.txt lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/ReduceTaskRunner.java Modified: lucene/hadoop/trunk/CHANGES.txt URL: http://svn.apache.org/viewvc/lucene/hadoop/trunk/CHANGES.txt?view=diff&rev=512936&r1=512935&r2=512936 ============================================================================== --- lucene/hadoop/trunk/CHANGES.txt (original) +++ lucene/hadoop/trunk/CHANGES.txt Wed Feb 28 11:59:40 2007 @@ -159,6 +159,9 @@ 47. HADOOP-972. Optimize HDFS's rack-aware block placement algorithm. (Hairong Kuang via cutting) +48. HADOOP-1043. Optimize shuffle, increasing parallelism. + (Devaraj Das via cutting) + Release 0.11.2 - 2007-02-16 Modified: lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/ReduceTaskRunner.java URL: http://svn.apache.org/viewvc/lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/ReduceTaskRunner.java?view=diff&rev=512936&r1=512935&r2=512936 ============================================================================== --- lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/ReduceTaskRunner.java (original) +++ lucene/hadoop/trunk/src/java/org/apache/hadoop/mapred/ReduceTaskRunner.java Wed Feb 28 11:59:40 2007 @@ -608,9 +608,17 @@ numInFlight--; } + boolean busy = true; // ensure we have enough to keep us busy if (numInFlight < lowThreshold && (numOutputs-numCopied) > probe_sample_size) { - break; + busy = false; + } + //Check whether we have more CopyResult to check. If there is none, and + //we are not busy enough, break + synchronized (copyResults) { + if (copyResults.size() == 0 && !busy) { + break; + } } }