Hadoop Map-Reduce Tutorial

This document comprehensively describes all user-facing facets of the Hadoop Map-Reduce framework and serves as a tutorial.

Ensure that Hadoop is installed, configured and is running. More details:

• Hadoop Quickstart for first-time users.
• Hadoop Cluster Setup for large, distributed clusters.

Hadoop Map-Reduce is a software framework for easily writing applications which process vast amounts of data (multi-terabyte data-sets) in-parallel on large clusters (thousands of nodes) of commodity hardware in a reliable, fault-tolerant manner.

A Map-Reduce job usually splits the input data-set into independent chunks which are processed by the map tasks in a completely parallel manner. The framework sorts the outputs of the maps, which are then input to the reduce tasks. Typically both the input and the output of the job are stored in a file-system. The framework takes care of scheduling tasks, monitoring them and re-executes the failed tasks.

Typically the compute nodes and the storage nodes are the same, that is, the Map-Reduce framework and the Distributed FileSystem are running on the same set of nodes. This configuration allows the framework to effectively schedule tasks on the nodes where data is already present, resulting in very high aggregate bandwidth across the cluster.

The Map-Reduce framework consists of a single master JobTracker and one slave TaskTracker per cluster-node. The master is responsible for scheduling the jobs' component tasks on the slaves, monitoring them and re-executing the failed tasks. The slaves execute the tasks as directed by the master.

Minimally, applications specify the input/output locations and supply map and reduce functions via implementations of appropriate interfaces and/or abstract-classes. These, and other job parameters, comprise the job configuration. The Hadoop job client then submits the job (jar/executable etc.) and configuration to the JobTracker which then assumes the responsibility of distributing the software/configuration to the slaves, scheduling tasks and monitoring them, providing status and diagnostic information to the job-client.

Although the Hadoop framework is implemented in Java^TM, Map-Reduce applications need not be written in Java.

• Hadoop Streaming is a utility which allows users to create and run jobs with any executables (e.g. shell utilities) as the mapper and/or the reducer.
• Hadoop Pipes is a SWIG- compatible C++ API to implement Map-Reduce applications (non JNI^TM based).

The Map-Reduce framework operates exclusively on <key, value> pairs, that is, the framework views the input to the job as a set of <key, value> pairs and produces a set of <key, value> pairs as the output of the job, conceivably of different types.

The key and value classes have to be serializable by the framework and hence need to implement the Writable interface. Additionally, the key classes have to implement the WritableComparable interface to facilitate sorting by the framework.

Input and Output types of a Map-Reduce job:

(input) <k1, v1> -> map -> <k2, v2> -> combine -> <k2, v2> -> reduce -> <k3, v3> (output)

Before we jump into the details, lets walk through an example Map-Reduce application to get a flavour for how they work.

WordCount is a simple application that counts the number of occurences of each word in a given input set.

This works with a local-standalone, pseudo-distributed or fully-distributed Hadoop installation.

Source Code

	WordCount.java
1.	package org.myorg;
2.
3.	import java.io.IOException;
4.	import java.util.*;
5.
6.	import org.apache.hadoop.fs.Path;
7.	import org.apache.hadoop.conf.*;
8.	import org.apache.hadoop.io.*;
9.	import org.apache.hadoop.mapred.*;
10.	import org.apache.hadoop.util.*;
11.
12.	public class WordCount {
13.
14.	public static class Map extends MapReduceBase implements Mapper<LongWritable, Text, Text, IntWritable> {
15.	private final static IntWritable one = new IntWritable(1);
16.	private Text word = new Text();
17.
18.	public void map(LongWritable key, Text value, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException {
19.	String line = value.toString();
20.	StringTokenizer tokenizer = new StringTokenizer(line);
21.	while (tokenizer.hasMoreTokens()) {
22.	word.set(tokenizer.nextToken());
23.	output.collect(word, one);
24.	}
25.	}
26.	}
27.
28.	public static class Reduce extends MapReduceBase implements Reducer<Text, IntWritable, Text, IntWritable> {
29.	public void reduce(Text key, Iterator<IntWritable> values, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException {
30.	int sum = 0;
31.	while (values.hasNext()) {
32.	sum += values.next().get();
33.	}
34.	output.collect(key, new IntWritable(sum));
35.	}
36.	}
37.
38.	public static void main(String[] args) throws Exception {
39.	JobConf conf = new JobConf(WordCount.class);
40.	conf.setJobName("wordcount");
41.
42.	conf.setOutputKeyClass(Text.class);
43.	conf.setOutputValueClass(IntWritable.class);
44.
45.	conf.setMapperClass(Map.class);
46.	conf.setCombinerClass(Reduce.class);
47.	conf.setReducerClass(Reduce.class);
48.
49.	conf.setInputFormat(TextInputFormat.class);
50.	conf.setOutputFormat(TextOutputFormat.class);
51.
52.	conf.setInputPath(new Path(args[0]));
53.	conf.setOutputPath(new Path(args[1]));
54.
55.	JobClient.runJob(conf);
57.	}
58.	}
59.

Usage

Assuming HADOOP_HOME is the root of the installation and HADOOP_VERSION is the Hadoop version installed, compile WordCount.java and create a jar:

$ mkdir wordcount_classes
$ javac -classpath ${HADOOP_HOME}/hadoop-${HADOOP_VERSION}-core.jar -d wordcount_classes WordCount.java
$ jar -cvf /usr/joe/wordcount.jar -C wordcount_classes/ .

Assuming that:

• /usr/joe/wordcount/input - input directory in HDFS
• /usr/joe/wordcount/output - output directory in HDFS

Sample text-files as input:

$ bin/hadoop dfs -ls /usr/joe/wordcount/input/
/usr/joe/wordcount/input/file01
/usr/joe/wordcount/input/file02

$ bin/hadoop dfs -cat /usr/joe/wordcount/input/file01
Hello World Bye World

$ bin/hadoop dfs -cat /usr/joe/wordcount/input/file02
Hello Hadoop Goodbye Hadoop

Run the application:

$ bin/hadoop jar /usr/joe/wordcount.jar org.myorg.WordCount /usr/joe/wordcount/input /usr/joe/wordcount/output

Output:

$ bin/hadoop dfs -cat /usr/joe/wordcount/output/part-00000
Bye 1
Goodbye 1
Hadoop 2
Hello 2
World 2

Walk-through

The WordCount application is quite straight-forward.

The Mapper implementation (lines 14-26), via the map method (lines 18-25), processes one line at a time, as provided by the specified TextInputFormat (line 49). It then splits the line into tokens separated by whitespaces, via the StringTokenizer, and emits a key-value pair of < <word>, 1>.

For the given sample input the first map emits:
< Hello, 1>
< World, 1>
< Bye, 1>
< World, 1>

The second map emits:
< Hello, 1>
< Hadoop, 1>
< Goodbye, 1>
< Hadoop, 1>

We'll learn more about the number of maps spawned for a given job, and how to control them in a fine-grained manner, a bit later in the tutorial.

WordCount also specifies a combiner (line 46). Hence, the output of each map is passed through the local combiner (which is same as the Reducer as per the job configuration) for local aggregation, after being sorted on the keys.

The output of the first map:
< Bye, 1>
< Hello, 1>
< World, 2>

The output of the second map:
< Goodbye, 1>
< Hadoop, 2>
< Hello, 1>

The Reducer implementation (lines 28-36), via the reduce method (lines 29-35) just sums up the values, which are the occurence counts for each key (i.e. words in this example).

Thus the output of the job is:
< Bye, 1>
< Goodbye, 1>
< Hadoop, 2>
< Hello, 2>
< World, 2>

The run method specifies various facets of the job, such as the input/output paths (passed via the command line), key/value types, input/output formats etc., in the JobConf. It then calls the JobClient.runJob (line 55) to submit the and monitor its progress.

We'll learn more about JobConf, JobClient, Tool and other interfaces and classes a bit later in the tutorial.

This section provides a reasonable amount of detail on every user-facing aspect of the Map-Reduce framwork. This should help users implement, configure and tune their jobs in a fine-grained manner. However, please note that the javadoc for each class/interface remains the most comprehensive documentation available; this is only meant to be a tutorial.

Let us first take the Mapper and Reducer interfaces. Applications typically implement them to provide the map and reduce methods.

We will then discuss other core interfaces including JobConf, JobClient, Partitioner, OutputCollector, Reporter, InputFormat, OutputFormat and others.

Finally, we will wrap up by discussing some useful features of the framework such as the DistributedCache, IsolationRunner etc.

Payload

Applications typically implement the Mapper and Reducer interfaces to provide the map and reduce methods. These form the core of the job.

Mapper

Mapper maps input key/value pairs to a set of intermediate key/value pairs.

Maps are the individual tasks that transform input records into intermediate records. The transformed intermediate records do not need to be of the same type as the input records. A given input pair may map to zero or many output pairs.

The Hadoop Map-Reduce framework spawns one map task for each InputSplit generated by the InputFormat for the job.

Overall, Mapper implementations are passed the JobConf for the job via the JobConfigurable.configure(JobConf) method and override it to initialize themselves. The framework then calls map(WritableComparable, Writable, OutputCollector, Reporter) for each key/value pair in the InputSplit for that task. Applications can then override the Closeable.close() method to perform any required cleanup.

Output pairs do not need to be of the same types as input pairs. A given input pair may map to zero or many output pairs. Output pairs are collected with calls to OutputCollector.collect(WritableComparable,Writable).

Applications can use the Reporter to report progress, set application-level status messages and update Counters, or just indicate that they are alive.

All intermediate values associated with a given output key are subsequently grouped by the framework, and passed to the Reducer(s) to determine the final output. Users can control the grouping by specifying a Comparator via JobConf.setOutputKeyComparatorClass(Class).

The Mapper outputs are sorted and then partitioned per Reducer. The total number of partitions is the same as the number of reduce tasks for the job. Users can control which keys (and hence records) go to which Reducer by implementing a custom Partitioner.

Users can optionally specify a combiner, via JobConf.setCombinerClass(Class), to perform local aggregation of the intermediate outputs, which helps to cut down the amount of data transferred from the Mapper to the Reducer.

The intermediate, sorted outputs are always stored in files of SequenceFile format. Applications can control if, and how, the intermediate outputs are to be compressed and the CompressionCodec to be used via the JobConf.

How Many Maps?

The number of maps is usually driven by the total size of the inputs, that is, the total number of blocks of the input files.

The right level of parallelism for maps seems to be around 10-100 maps per-node, although it has been set up to 300 maps for very cpu-light map tasks. Task setup takes awhile, so it is best if the maps take at least a minute to execute.

Thus, if you expect 10TB of input data and have a blocksize of 128MB, you'll end up with 82,000 maps, unless setNumMapTasks(int) (which only provides a hint to the framework) is used to set it even higher.

Reducer

Reducer reduces a set of intermediate values which share a key to a smaller set of values.

The number of reduces for the job is set by the user via JobConf.setNumReduceTasks(int).

Overall, Reducer implementations are passed the JobConf for the job via the JobConfigurable.configure(JobConf) method and can override it to initialize themselves. The framework then calls reduce(WritableComparable, Iterator, OutputCollector, Reporter) method for each <key, (list of values)> pair in the grouped inputs. Applications can then override the Closeable.close() method to perform any required cleanup.

Reducer has 3 primary phases: shuffle, sort and reduce.

Shuffle

Input to the Reducer is the sorted output of the mappers. In this phase the framework fetches the relevant partition of the output of all the mappers, via HTTP.

Sort

The framework groups Reducer inputs by keys (since different mappers may have output the same key) in this stage.

The shuffle and sort phases occur simultaneously; while map-outputs are being fetched they are merged.

Secondary Sort

If equivalence rules for grouping the intermediate keys are required to be different from those for grouping keys before reduction, then one may specify a Comparator via JobConf.setOutputValueGroupingComparator(Class). Since JobConf.setOutputKeyComparatorClass(Class) can be used to control how intermediate keys are grouped, these can be used in conjunction to simulate secondary sort on values.

Reduce

In this phase the reduce(WritableComparable, Iterator, OutputCollector, Reporter) method is called for each <key, (list of values)> pair in the grouped inputs.

The output of the reduce task is typically written to the FileSystem via OutputCollector.collect(WritableComparable, Writable).

Applications can use the Reporter to report progress, set application-level status messages and update Counters, or just indicate that they are alive.

The output of the Reducer is not sorted.

How Many Reduces?

The right number of reduces seems to be 0.95 or 1.75 multiplied by (<no. of nodes> * mapred.tasktracker.reduce.tasks.maximum).

With 0.95 all of the reduces can launch immediately and start transfering map outputs as the maps finish. With 1.75 the faster nodes will finish their first round of reduces and launch a second wave of reduces doing a much better job of load balancing.

Increasing the number of reduces increases the framework overhead, but increases load balancing and lowers the cost of failures.

The scaling factors above are slightly less than whole numbers to reserve a few reduce slots in the framework for speculative-tasks and failed tasks.

Reducer NONE

It is legal to set the number of reduce-tasks to zero if no reduction is desired.

In this case the outputs of the map-tasks go directly to the FileSystem, into the output path set by setOutputPath(Path). The framework does not sort the map-outputs before writing them out to the FileSystem.

Partitioner

Partitioner partitions the key space.

Partitioner controls the partitioning of the keys of the intermediate map-outputs. The key (or a subset of the key) is used to derive the partition, typically by a hash function. The total number of partitions is the same as the number of reduce tasks for the job. Hence this controls which of the m reduce tasks the intermediate key (and hence the record) is sent to for reduction.

HashPartitioner is the default Partitioner.

Reporter

Reporter is a facility for Map-Reduce applications to report progress, set application-level status messages and update Counters.

Mapper and Reducer implementations can use the Reporter to report progress or just indicate that they are alive. In scenarios where the application takes a significant amount of time to process individual key/value pairs, this is crucial since the framework might assume that the task has timed-out and kill that task. Another way to avoid this is to set the configuration parameter mapred.task.timeout to a high-enough value (or even set it to zero for no time-outs).

Applications can also update Counters using the Reporter.

OutputCollector

OutputCollector is a generalization of the facility provided by the Map-Reduce framework to collect data output by the Mapper or the Reducer (either the intermediate outputs or the output of the job).

Hadoop Map-Reduce comes bundled with a library of generally useful mappers, reducers, and partitioners.

Job Configuration

JobConf represents a Map-Reduce job configuration.

JobConf is the primary interface for a user to describe a map-reduce job to the Hadoop framework for execution. The framework tries to faithfully execute the job as described by JobConf, however:

• f Some configuration parameters may have been marked as final by administrators and hence cannot be altered.
• While some job parameters are straight-forward to set (e.g. setNumReduceTasks(int)), other parameters interact subtly with the rest of the framework and/or job configuration and are more complex to set (e.g. setNumMapTasks(int)).

JobConf is typically used to specify the Mapper, combiner (if any), Partitioner, Reducer, InputFormat and OutputFormat implementations. JobConf also indicates the set of input files (setInputPath(Path)/addInputPath(Path)) and where the output files should be written (setOutputPath(Path)).

Optionally, JobConf is used to specify other advanced facets of the job such as the Comparator to be used, files to be put in the DistributedCache, whether intermediate and/or job outputs are to be compressed (and how), debugging via user-provided scripts (setMapDebugScript(String)/setReduceDebugScript(String)) , whether job tasks can be executed in a speculative manner (setMapSpeculativeExecution(boolean))/(setReduceSpeculativeExecution(boolean)) , maximum number of attempts per task (setMaxMapAttempts(int)/setMaxReduceAttempts(int)) , percentage of tasks failure which can be tolerated by the job (setMaxMapTaskFailuresPercent(int)/setMaxReduceTaskFailuresPercent(int)) etc.

Of course, users can use set(String, String)/get(String, String) to set/get arbitrary parameters needed by applications. However, use the DistributedCache for large amounts of (read-only) data.

Task Execution & Environment

The TaskTracker executes the Mapper/ Reducer task as a child process in a separate jvm.

The child-task inherits the environment of the parent TaskTracker. The user can specify additional options to the child-jvm via the mapred.child.java.opts configuration parameter in the JobConf such as non-standard paths for the run-time linker to search shared libraries via -Djava.library.path=<> etc. If the mapred.child.java.opts contains the symbol @taskid@ it is interpolated with value of taskid of the map/reduce task.

Here is an example with multiple arguments and substitutions, showing jvm GC logging, and start of a passwordless JVM JMX agent so that it can connect with jconsole and the likes to watch child memory, threads and get thread dumps. It also sets the maximum heap-size of the child jvm to 512MB and adds an additional path to the java.library.path of the child-jvm.

<property>
  <name>mapred.child.java.opts</name>
  <value>
     -Xmx512M -Djava.library.path=/home/mycompany/lib -verbose:gc -Xloggc:/tmp/@[email protected]
     -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false
  </value>
</property>

The DistributedCache can also be used as a rudimentary software distribution mechanism for use in the map and/or reduce tasks. It can be used to distribute both jars and native libraries. The DistributedCache.addArchiveToClassPath(Path, Configuration) or DistributedCache.addFileToClassPath(Path, Configuration) api can be used to cache files/jars and also add them to the classpath of child-jvm. Similarly the facility provided by the DistributedCache where-in it symlinks the cached files into the working directory of the task can be used to distribute native libraries and load them. The underlying detail is that child-jvm always has its current working directory added to the java.library.path and hence the cached libraries can be loaded via System.loadLibrary or System.load.

Job Submission and Monitoring

JobClient is the primary interface by which user-job interacts with the JobTracker.

JobClient provides facilities to submit jobs, track their progress, access component-tasks' reports/logs, get the Map-Reduce cluster's status information and so on.

The job submission process involves:

1. Checking the input and output specifications of the job.
2. Computing the InputSplit values for the job.
3. Setting up the requisite accounting information for the DistributedCache of the job, if necessary.
4. Copying the job's jar and configuration to the map-reduce system directory on the FileSystem.
5. Submitting the job to the JobTracker and optionally monitoring it's status.

Normally the user creates the application, describes various facets of the job via JobConf, and then uses the JobClient to submit the job and monitor its progress.

Job Control

Users may need to chain map-reduce jobs to accomplish complex tasks which cannot be done via a single map-reduce job. This is fairly easy since the output of the job typically goes to distributed file-system, and the output, in turn, can be used as the input for the next job.

However, this also means that the onus on ensuring jobs are complete (success/failure) lies squarely on the clients. In such cases, the various job-control options are:

• runJob(JobConf) : Submits the job and returns only after the job has completed.
• submitJob(JobConf) : Only submits the job, then poll the returned handle to the RunningJob to query status and make scheduling decisions.
• JobConf.setJobEndNotificationURI(String) : Sets up a notification upon job-completion, thus avoiding polling.

Job Input

InputFormat describes the input-specification for a Map-Reduce job.

The Map-Reduce framework relies on the InputFormat of the job to:

1. Validate the input-specification of the job.
2. Split-up the input file(s) into logical InputSplit instances, each of which is then assigned to an individual Mapper.
3. Provide the RecordReader implementation used to glean input records from the logical InputSplit for processing by the Mapper.

The default behavior of file-based InputFormat implementations, typically sub-classes of FileInputFormat, is to split the input into logical InputSplit instances based on the total size, in bytes, of the input files. However, the FileSystem blocksize of the input files is treated as an upper bound for input splits. A lower bound on the split size can be set via mapred.min.split.size.

Clearly, logical splits based on input-size is insufficient for many applications since record boundaries must be respected. In such cases, the application should implement a RecordReader, who is responsible for respecting record-boundaries and presents a record-oriented view of the logical InputSplit to the individual task.

TextInputFormat is the default InputFormat.

If TextInputFormat is the InputFormat for a given job, the framework detects input-files with the .gz and .lzo extensions and automatically decompresses them using the appropriate CompressionCodec. However, it must be noted that compressed files with the above extensions cannot be split and each compressed file is processed in its entirety by a single mapper.

InputSplit

InputSplit represents the data to be processed by an individual Mapper.

Typically InputSplit presents a byte-oriented view of the input, and it is the responsibility of RecordReader to process and present a record-oriented view.

FileSplit is the default InputSplit. It sets map.input.file to the path of the input file for the logical split.

RecordReader

RecordReader reads <key, value> pairs from an InputSplit.

Typically the RecordReader converts the byte-oriented view of the input, provided by the InputSplit, and presents a record-oriented to the Mapper implementations for processing. RecordReader thus assumes the responsibility of processing record boundaries and presents the tasks with keys and values.

Job Output

OutputFormat describes the output-specification for a Map-Reduce job.

The Map-Reduce framework relies on the OutputFormat of the job to:

1. Validate the output-specification of the job; for example, check that the output directory doesn't already exist.
2. Provide the RecordWriter implementation used to write the output files of the job. Output files are stored in a FileSystem.

TextOutputFormat is the default OutputFormat.

Task Side-Effect Files

In some applications, component tasks need to create and/or write to side-files, which differ from the actual job-output files.

In such cases there could be issues with two instances of the same Mapper or Reducer running simultaneously (for example, speculative tasks) trying to open and/or write to the same file (path) on the FileSystem. Hence the application-writer will have to pick unique names per task-attempt (using the taskid, say task_200709221812_0001_m_000000_0), not just per task.

To avoid these issues the Map-Reduce framework maintains a special ${mapred.output.dir}/_${taskid} sub-directory for each task-attempt on the FileSystem where the output of the task-attempt is stored. On successful completion of the task-attempt, the files in the ${mapred.output.dir}/_${taskid} (only) are promoted to ${mapred.output.dir}. Of course, the framework discards the sub-directory of unsuccessful task-attempts. This process is completely transparent to the application.

The application-writer can take advantage of this feature by creating any side-files required in ${mapred.output.dir} during execution of a task via JobConf.getOutputPath(), and the framework will promote them similarly for succesful task-attempts, thus eliminating the need to pick unique paths per task-attempt.

RecordWriter

RecordWriter writes the output <key, value> pairs to an output file.

RecordWriter implementations write the job outputs to the FileSystem.

Other Useful Features

Counters

Counters represent global counters, defined either by the Map-Reduce framework or applications. Each Counter can be of any Enum type. Counters of a particular Enum are bunched into groups of type Counters.Group.

Applications can define arbitrary Counters (of type Enum) and update them via Reporter.incrCounter(Enum, long) in the map and/or reduce methods. These counters are then globally aggregated by the framework.

DistributedCache

DistributedCache distributes application-specific, large, read-only files efficiently.

DistributedCache is a facility provided by the Map-Reduce framework to cache files (text, archives, jars and so on) needed by applications.

Applications specify the files to be cached via urls (hdfs:// or http://) in the JobConf. The DistributedCache assumes that the files specified via hdfs:// urls are already present on the FileSystem.

The framework will copy the necessary files to the slave node before any tasks for the job are executed on that node. Its efficiency stems from the fact that the files are only copied once per job and the ability to cache archives which are un-archived on the slaves.

DistributedCache tracks the modification timestamps of the cached files. Clearly the cache files should not be modified by the application or externally while the job is executing.

DistributedCache can be used to distribute simple, read-only data/text files and more complex types such as archives and jars. Archives (zip files) are un-archived at the slave nodes. Optionally users can also direct the DistributedCache to symlink the cached file(s) into the current working directory of the task via the DistributedCache.createSymlink(Path, Configuration) api. Files have execution permissions set.

Tool

The Tool interface supports the handling of generic Hadoop command-line options.

Tool is the standard for any Map-Reduce tool or application. The application should delegate the handling of standard command-line options to GenericOptionsParser via ToolRunner.run(Tool, String[]) and only handle its custom arguments.

The generic Hadoop command-line options are:
-conf <configuration file>
-D <property=value>
-fs <local|namenode:port>
-jt <local|jobtracker:port>

IsolationRunner

IsolationRunner is a utility to help debug Map-Reduce programs.

To use the IsolationRunner, first set keep.failed.tasks.files to true (also see keep.tasks.files.pattern).

Next, go to the node on which the failed task ran and go to the TaskTracker's local directory and run the IsolationRunner:
$ cd <local path>/taskTracker/${taskid}/work
$ bin/hadoop org.apache.hadoop.mapred.IsolationRunner ../job.xml

IsolationRunner will run the failed task in a single jvm, which can be in the debugger, over precisely the same input.

JobControl

JobControl is a utility which encapsulates a set of Map-Reduce jobs and their dependencies.

Data Compression

Hadoop Map-Reduce provides facilities for the application-writer to specify compression for both intermediate map-outputs and the job-outputs i.e. output of the reduces. It also comes bundled with CompressionCodec implementations for the zlib and lzo compression algorithms. The gzip file format is also supported.

Hadoop also provides native implementations of the above compression codecs for reasons of both performance (zlib) and non-availability of Java libraries (lzo). More details on their usage and availability are available here.

Intermediate Outputs

Applications can control compression of intermediate map-outputs via the JobConf.setCompressMapOutput(boolean) api and the CompressionCodec to be used via the JobConf.setMapOutputCompressorClass(Class) api. Since the intermediate map-outputs are always stored in the SequenceFile format, the SequenceFile.CompressionType (i.e. RECORD / BLOCK - defaults to RECORD) can be specified via the JobConf.setMapOutputCompressionType(SequenceFile.CompressionType) api.

Job Outputs

Applications can control compression of job-outputs via the OutputFormatBase.setCompressOutput(JobConf, boolean) api and the CompressionCodec to be used can be specified via the OutputFormatBase.setOutputCompressorClass(JobConf, Class) api.

If the job outputs are to be stored in the SequenceFileOutputFormat, the required SequenceFile.CompressionType (i.e. RECORD / BLOCK - defaults to RECORD)can be specified via the SequenceFileOutputFormat.setOutputCompressionType(JobConf, SequenceFile.CompressionType) api.

Here is a more complete WordCount which uses many of the features provided by the Map-Reduce framework we discussed so far.

This needs the HDFS to be up and running, especially for the DistributedCache-related features. Hence it only works with a pseudo-distributed or fully-distributed Hadoop installation.

Source Code

	WordCount.java
1.	package org.myorg;
2.
3.	import java.io.*;
4.	import java.util.*;
5.
6.	import org.apache.hadoop.fs.Path;
7.	import org.apache.hadoop.filecache.DistributedCache;
8.	import org.apache.hadoop.conf.*;
9.	import org.apache.hadoop.io.*;
10.	import org.apache.hadoop.mapred.*;
11.	import org.apache.hadoop.util.*;
12.
13.	public class WordCount extends Configured implements Tool {
14.
15.	public static class Map extends MapReduceBase implements Mapper<LongWritable, Text, Text, IntWritable> {
16.
17.	static enum Counters { INPUT_WORDS }
18.
19.	private final static IntWritable one = new IntWritable(1);
20.	private Text word = new Text();
21.
22.	private boolean caseSensitive = true;
23.	private Set<String> patternsToSkip = new HashSet<String>();
24.
25.	private long numRecords = 0;
26.	private String inputFile;
27.
28.	public void configure(JobConf job) {
29.	caseSensitive = job.getBoolean("wordcount.case.sensitive", true);
30.	inputFile = job.get("map.input.file");
31.
32.	if (job.getBoolean("wordcount.skip.patterns", false)) {
33.	Path[] patternsFiles = new Path[0];
34.	try {
35.	patternsFiles = DistributedCache.getLocalCacheFiles(job);
36.	} catch (IOException ioe) {
37.	System.err.println("Caught exception while getting cached files: " + StringUtils.stringifyException(ioe));
38.	}
39.	for (Path patternsFile : patternsFiles) {
40.	parseSkipFile(patternsFile);
41.	}
42.	}
43.	}
44.
45.	private void parseSkipFile(Path patternsFile) {
46.	try {
47.	BufferedReader fis = new BufferedReader(new FileReader(patternsFile.toString()));
48.	String pattern = null;
49.	while ((pattern = fis.readLine()) != null) {
50.	patternsToSkip.add(pattern);
51.	}
52.	} catch (IOException ioe) {
53.	System.err.println("Caught exception while parsing the cached file '" + patternsFile + "' : " + StringUtils.stringifyException(ioe));
54.	}
55.	}
56.
57.	public void map(LongWritable key, Text value, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException {
58.	String line = (caseSensitive) ? value.toString() : value.toString().toLowerCase();
59.
60.	for (String pattern : patternsToSkip) {
61.	line = line.replaceAll(pattern, "");
62.	}
63.
64.	StringTokenizer tokenizer = new StringTokenizer(line);
65.	while (tokenizer.hasMoreTokens()) {
66.	word.set(tokenizer.nextToken());
67.	output.collect(word, one);
68.	reporter.incrCounter(Counters.INPUT_WORDS, 1);
69.	}
70.
71.	if ((++numRecords % 100) == 0) {
72.	reporter.setStatus("Finished processing " + numRecords + " records " + "from the input file: " + inputFile);
73.	}
74.	}
75.	}
76.
77.	public static class Reduce extends MapReduceBase implements Reducer<Text, IntWritable, Text, IntWritable> {
78.	public void reduce(Text key, Iterator<IntWritable> values, OutputCollector<Text, IntWritable> output, Reporter reporter) throws IOException {
79.	int sum = 0;
80.	while (values.hasNext()) {
81.	sum += values.next().get();
82.	}
83.	output.collect(key, new IntWritable(sum));
84.	}
85.	}
86.
87.	public int run(String[] args) throws Exception {
88.	JobConf conf = new JobConf(getConf(), WordCount.class);
89.	conf.setJobName("wordcount");
90.
91.	conf.setOutputKeyClass(Text.class);
92.	conf.setOutputValueClass(IntWritable.class);
93.
94.	conf.setMapperClass(Map.class);
95.	conf.setCombinerClass(Reduce.class);
96.	conf.setReducerClass(Reduce.class);
97.
98.	conf.setInputFormat(TextInputFormat.class);
99.	conf.setOutputFormat(TextOutputFormat.class);
100.
101.	List<String> other_args = new ArrayList<String>();
102.	for (int i=0; i < args.length; ++i) {
103.	if ("-skip".equals(args[i])) {
104.	DistributedCache.addCacheFile(new Path(args[++i]).toUri(), conf);
105.	conf.setBoolean("wordcount.skip.patterns", true);
106.	} else {
107.	other_args.add(args[i]);
108.	}
109.	}
110.
111.	conf.setInputPath(new Path(other_args.get(0)));
112.	conf.setOutputPath(new Path(other_args.get(1)));
113.
114.	JobClient.runJob(conf);
115.	return 0;
116.	}
117.
118.	public static void main(String[] args) throws Exception {
119.	int res = ToolRunner.run(new Configuration(), new WordCount(), args);
120.	System.exit(res);
121.	}
122.	}
123.

Sample Runs

Sample text-files as input:

$ bin/hadoop dfs -ls /usr/joe/wordcount/input/
/usr/joe/wordcount/input/file01
/usr/joe/wordcount/input/file02

$ bin/hadoop dfs -cat /usr/joe/wordcount/input/file01
Hello World, Bye World!

$ bin/hadoop dfs -cat /usr/joe/wordcount/input/file02
Hello Hadoop, Goodbye to hadoop.

Run the application:

$ bin/hadoop jar /usr/joe/wordcount.jar org.myorg.WordCount /usr/joe/wordcount/input /usr/joe/wordcount/output

Output:

$ bin/hadoop dfs -cat /usr/joe/wordcount/output/part-00000
Bye 1
Goodbye 1
Hadoop, 1
Hello 2
World! 1
World, 1
hadoop. 1
to 1

Notice that the inputs differ from the first version we looked at, and how they affect the outputs.

Now, lets plug-in a pattern-file which lists the word-patterns to be ignored, via the DistributedCache.

$ hadoop dfs -cat /user/joe/wordcount/patterns.txt
\.
\,
\!
to

Run it again, this time with more options:

$ bin/hadoop jar /usr/joe/wordcount.jar org.myorg.WordCount -Dwordcount.case.sensitive=true /usr/joe/wordcount/input /usr/joe/wordcount/output -skip /user/joe/wordcount/patterns.txt

As expected, the output:

$ bin/hadoop dfs -cat /usr/joe/wordcount/output/part-00000
Bye 1
Goodbye 1
Hadoop 1
Hello 2
World 2
hadoop 1

Run it once more, this time switch-off case-sensitivity:

$ bin/hadoop jar /usr/joe/wordcount.jar org.myorg.WordCount -Dwordcount.case.sensitive=false /usr/joe/wordcount/input /usr/joe/wordcount/output -skip /user/joe/wordcount/patterns.txt

Sure enough, the output:

$ bin/hadoop dfs -cat /usr/joe/wordcount/output/part-00000
bye 1
goodbye 1
hadoop 2
hello 2
world 2

Highlights

The second version of WordCount improves upon the previous one by using some features offered by the Map-Reduce framework:

• Demonstrates how applications can access configuration parameters in the configure method of the Mapper (and Reducer) implementations (lines 28-43).
• Demonstrates how the DistributedCache can be used to distribute read-only data needed by the jobs. Here it allows the user to specify word-patterns to skip while counting (line 104).
• Demonstrates the utility of the Tool interface and the GenericOptionsParser to handle generic Hadoop command-line options (lines 87-116, 119).
• Demonstrates how applications can use Counters (line 68) and how they can set application-specific status information via the Reporter instance passed to the map (and reduce) method (line 72).