Pārlūkot izejas kodu

MAPREDUCE-4887. Add RehashPartitioner, to smooth distributions with poor implementations of Object#hashCode(). Contributed by Radim Kolar.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1424158 13f79535-47bb-0310-9956-ffa450edef68
Doug Cutting 12 gadi atpakaļ
vecāks
revīzija
aa30056d09

+ 3 - 0
hadoop-mapreduce-project/CHANGES.txt

@@ -16,6 +16,9 @@ Trunk (Unreleased)
 
     MAPREDUCE-4807. Allow MapOutputBuffer to be pluggable. (masokan via tucu)
 
+    MAPREDUCE-4887. Add RehashPartitioner, to smooth distributions
+    with poor implementations of Object#hashCode().  (Radim Kolar via cutting)
+
   IMPROVEMENTS
 
     MAPREDUCE-3787. [Gridmix] Optimize job monitoring and STRESS mode for

+ 48 - 0
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/lib/partition/RehashPartitioner.java

@@ -0,0 +1,48 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.hadoop.mapreduce.lib.partition;
+
+import org.apache.hadoop.classification.InterfaceAudience;
+import org.apache.hadoop.classification.InterfaceStability;
+import org.apache.hadoop.mapreduce.Partitioner;
+
+/**
+  *  This partitioner rehashes values returned by {@link Object#hashCode()}
+  *  to get smoother distribution between partitions which may improve
+  *  reduce reduce time in some cases and should harm things in no cases.
+  *  This partitioner is suggested with Integer and Long keys with simple
+  *  patterns in their distributions.
+  *  @since 2.0.3
+ */
+@InterfaceAudience.Public
+@InterfaceStability.Stable
+public class RehashPartitioner<K, V> extends Partitioner<K, V> {
+
+  /** prime number seed for increasing hash quality */
+  private static final int SEED = 1591267453;
+
+  /** Rehash {@link Object#hashCode()} to partition. */
+  public int getPartition(K key, V value, int numReduceTasks) {
+    int h = SEED ^ key.hashCode();
+    h ^= (h >>> 20) ^ (h >>> 12);
+    h = h ^ (h >>> 7) ^ (h >>> 4);
+
+    return (h & Integer.MAX_VALUE) % numReduceTasks;
+  }
+}

+ 71 - 0
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/test/java/org/apache/hadoop/mapreduce/lib/partition/TestRehashPartitioner.java

@@ -0,0 +1,71 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.mapreduce.lib.partition;
+
+import static org.junit.Assert.*;
+
+import java.util.Arrays;
+import java.util.Collections;
+
+import org.apache.commons.lang.ArrayUtils;
+import org.apache.hadoop.io.IntWritable;
+import org.apache.hadoop.io.NullWritable;
+
+import org.junit.*;
+
+public class TestRehashPartitioner {
+
+  /** number of partitions */
+  private static final int PARTITIONS = 32;
+
+  /** step in sequence */
+  private static final int STEP = 3;
+
+  /** end of test sequence */
+  private static final int END = 100000;
+
+  /** maximum error for considering too big/small bucket */
+  private static final double MAX_ERROR = 0.20;
+
+  /** maximum number of oddly sized buckets */
+  private static final double MAX_BADBUCKETS = 0.10;
+
+  /** test partitioner for patterns */
+  @Test
+  public void testPatterns() {
+    int results[] = new int[PARTITIONS];
+    RehashPartitioner <IntWritable, NullWritable> p = new RehashPartitioner < IntWritable, NullWritable> ();
+    /* test sequence 4, 8, 12, ... 128 */
+    for(int i = 0; i < END; i+= STEP) {
+      results[p.getPartition(new IntWritable(i), null, PARTITIONS)]++;
+    }
+    int badbuckets = 0;
+    Integer min = Collections.min(Arrays.asList(ArrayUtils.toObject(results)));
+    Integer max = Collections.max(Arrays.asList(ArrayUtils.toObject(results)));
+    Integer avg = (int) Math.round((max+min)/2.0);
+    System.out.println("Dumping buckets distribution: min="+min+" avg="+avg+" max="+max);
+    for (int i = 0; i < PARTITIONS; i++) {
+      double var = (results[i]-avg)/(double)(avg);
+      System.out.println("bucket "+i+" "+results[i]+" items, variance "+var);
+      if (Math.abs(var) > MAX_ERROR)
+        badbuckets++;
+    }
+    System.out.println(badbuckets + " of "+PARTITIONS+" are too small or large buckets");
+    assertTrue("too many overflow buckets", badbuckets < PARTITIONS * MAX_BADBUCKETS);
+  }
+}