MAPREDUCE-3827. Changed Counters to use ConcurrentSkipListMap for performance. Contributed by Vinod K V.

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1241711 13f79535-47bb-0310-9956-ffa450edef68

hadoop-mapreduce-project/CHANGES.txt

@@ -764,6 +764,9 @@ Release 0.23.1 - Unreleased
     MAPREDUCE-3823. Ensure counters are calculated only once after a job
     finishes. (Vinod Kumar Vavilapalli via sseth)
 
+    MAPREDUCE-3827. Changed Counters to use ConcurrentSkipListMap for
+    performance. (vinodkv via acmurthy)  
+
 Release 0.23.0 - 2011-11-01 
 
   INCOMPATIBLE CHANGES

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/counters/AbstractCounterGroup.java

@@ -22,11 +22,8 @@ import java.io.DataInput;
 import java.io.DataOutput;
 import java.io.IOException;
 import java.util.Iterator;
-import java.util.Map;
+import java.util.concurrent.ConcurrentMap;
-
+import java.util.concurrent.ConcurrentSkipListMap;
-import com.google.common.collect.ImmutableSet;
-import com.google.common.collect.Iterators;
-import com.google.common.collect.Maps;
 
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.io.Text;
@@ -34,6 +31,8 @@ import org.apache.hadoop.io.WritableUtils;
 import org.apache.hadoop.mapreduce.Counter;
 import org.apache.hadoop.mapreduce.util.ResourceBundles;
 
+import com.google.common.collect.Iterators;
+
 /**
  * An abstract class to provide common implementation of the
  * generic counter group in both mapred and mapreduce package.
@@ -46,7 +45,8 @@ public abstract class AbstractCounterGroup<T extends Counter>
 
   private final String name;
   private String displayName;
-  private final Map<String, T> counters = Maps.newTreeMap();
+  private final ConcurrentMap<String, T> counters =
+      new ConcurrentSkipListMap<String, T>();
   private final Limits limits;
 
   public AbstractCounterGroup(String name, String displayName,
@@ -80,7 +80,7 @@ public abstract class AbstractCounterGroup<T extends Counter>
   @Override
   public synchronized T addCounter(String counterName, String displayName,
                                    long value) {
-    String saveName = limits.filterCounterName(counterName);
+    String saveName = Limits.filterCounterName(counterName);
     T counter = findCounterImpl(saveName, false);
     if (counter == null) {
       return addCounterImpl(saveName, displayName, value);
@@ -97,7 +97,9 @@ public abstract class AbstractCounterGroup<T extends Counter>
 
   @Override
   public synchronized T findCounter(String counterName, String displayName) {
-    String saveName = limits.filterCounterName(counterName);
+    // Take lock to avoid two threads not finding a counter and trying to add
+    // the same counter.
+    String saveName = Limits.filterCounterName(counterName);
     T counter = findCounterImpl(saveName, false);
     if (counter == null) {
       return addCounterImpl(saveName, displayName, 0);
@@ -106,10 +108,12 @@ public abstract class AbstractCounterGroup<T extends Counter>
   }
 
   @Override
-  public synchronized T findCounter(String counterName, boolean create) {
+  public T findCounter(String counterName, boolean create) {
-    return findCounterImpl(limits.filterCounterName(counterName), create);
+    return findCounterImpl(Limits.filterCounterName(counterName), create);
   }
 
+  // Lock the object. Cannot simply use concurrent constructs on the counters
+  // data-structure (like putIfAbsent) because of localization, limits etc.
   private synchronized T findCounterImpl(String counterName, boolean create) {
     T counter = counters.get(counterName);
     if (counter == null && create) {
@@ -142,8 +146,8 @@ public abstract class AbstractCounterGroup<T extends Counter>
   protected abstract T newCounter();
 
   @Override
-  public synchronized Iterator<T> iterator() {
+  public Iterator<T> iterator() {
-    return ImmutableSet.copyOf(counters.values()).iterator();
+    return counters.values().iterator();
   }
 
   /**

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/counters/AbstractCounters.java

@@ -18,19 +18,18 @@
 
 package org.apache.hadoop.mapreduce.counters;
 
+import static org.apache.hadoop.mapreduce.counters.CounterGroupFactory.getFrameworkGroupId;
+import static org.apache.hadoop.mapreduce.counters.CounterGroupFactory.isFrameworkGroup;
+
 import java.io.DataInput;
 import java.io.DataOutput;
 import java.io.IOException;
 import java.util.Iterator;
 import java.util.Map;
+import java.util.concurrent.ConcurrentSkipListMap;
 
-import com.google.common.collect.ImmutableSet;
-import com.google.common.collect.Iterables;
-import com.google.common.collect.Iterators;
-import com.google.common.collect.Maps;
-
-import org.apache.commons.logging.LogFactory;
 import org.apache.commons.logging.Log;
+import org.apache.commons.logging.LogFactory;
 import org.apache.hadoop.classification.InterfaceAudience;
 import org.apache.hadoop.classification.InterfaceStability;
 import org.apache.hadoop.io.Text;
@@ -40,7 +39,10 @@ import org.apache.hadoop.mapreduce.Counter;
 import org.apache.hadoop.mapreduce.FileSystemCounter;
 import org.apache.hadoop.mapreduce.JobCounter;
 import org.apache.hadoop.mapreduce.TaskCounter;
-import static org.apache.hadoop.mapreduce.counters.CounterGroupFactory.*;
+
+import com.google.common.collect.Iterables;
+import com.google.common.collect.Iterators;
+import com.google.common.collect.Maps;
 
 /**
  * An abstract class to provide common implementation for the Counters
@@ -61,8 +63,10 @@ public abstract class AbstractCounters<C extends Counter,
    * A cache from enum values to the associated counter.
    */
   private Map<Enum<?>, C> cache = Maps.newIdentityHashMap();
-  private Map<String, G> fgroups = Maps.newTreeMap(); // framework & fs groups
+  //framework & fs groups
-  private Map<String, G> groups = Maps.newTreeMap();  // other groups
+  private Map<String, G> fgroups = new ConcurrentSkipListMap<String, G>();
+  // other groups
+  private Map<String, G> groups = new ConcurrentSkipListMap<String, G>();
   private final CounterGroupFactory<C, G> groupFactory;
 
   // For framework counter serialization without strings
@@ -181,14 +185,13 @@ public abstract class AbstractCounters<C extends Counter,
    * @return Set of counter names.
    */
   public synchronized Iterable<String> getGroupNames() {
-    return Iterables.concat(ImmutableSet.copyOf(fgroups.keySet()),
+    return Iterables.concat(fgroups.keySet(), groups.keySet());
-                            ImmutableSet.copyOf(groups.keySet()));
   }
 
   @Override
-  public synchronized Iterator<G> iterator() {
+  public Iterator<G> iterator() {
-    return Iterators.concat(ImmutableSet.copyOf(fgroups.values()).iterator(),
+    return Iterators.concat(fgroups.values().iterator(),
-                            ImmutableSet.copyOf(groups.values()).iterator());
+                            groups.values().iterator());
   }
 
   /**
@@ -216,7 +219,7 @@ public abstract class AbstractCounters<C extends Counter,
   private String filterGroupName(String oldName) {
     String newName = legacyMap.get(oldName);
     if (newName == null) {
-      return limits.filterGroupName(oldName);
+      return Limits.filterGroupName(oldName);
     }
     LOG.warn("Group "+ oldName +" is deprecated. Use "+ newName +" instead");
     return newName;

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/counters/Limits.java

@@ -42,11 +42,11 @@ public class Limits {
     return name.length() > maxLen ? name.substring(0, maxLen - 1) : name;
   }
 
-  public String filterCounterName(String name) {
+  public static String filterCounterName(String name) {
     return filterName(name, COUNTER_NAME_MAX);
   }
 
-  public String filterGroupName(String name) {
+  public static String filterGroupName(String name) {
     return filterName(name, GROUP_NAME_MAX);
   }