HADOOP 0.20.2以上版本MapReduce中采用压缩技术

Hadoop 可以在作业的Configuration对象中通过设定一系列参数来改变作业的行为，比如，我们需要进行一个map-reduce作业，并且吧最终作业reduce过程的结果输出为压缩的格式，我们可以在一般的map-reduce上进行一些定制。

实现

还是以以前做的删选最高气温的例子为参照：

以前的例子可以见这个博文：/article/4268165.html

我们现在要求让结果输出为压缩格式，所以保持Map类(MaxTemperatureMapper）和Reduce类（MaxTemperatureReducer）不变，只要在Job类的Configuration作一些压缩的配置即可，见第45-49行所示：

package com.charles.parseweather.compression;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.io.compress.CompressionCodec;
import org.apache.hadoop.io.compress.GzipCodec;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

/**
*
*
* Description: 这个类定义并且运行作业,压缩版
*
* @author charles.wang
* @created May 24, 2012 5:29:12 PM
*
*/

public class MaxTemperatureWithCompression {

/**
* @param args
*/
public static void main(String[] args) throws Exception{
// TODO Auto-generated method stub

if (args.length !=2){
System.err.println("Usage: MaxTemperature <input path> <output path>");
System.exit(-1);
}

//创建一个Map-Reduce的作业
Configuration conf = new Configuration();
conf.set("hadoop.job.ugi", "hadoop-user,hadoop-user");

//在这里我们配置一些和压缩有关的参数

//我们设定reduce输出结果使用gzip压缩的形式
conf.setBoolean("mapred.output.compress", true);
conf.setClass("mapred.output.compression.codec", GzipCodec.class, CompressionCodec.class);

Job job = new Job(conf,"Get Maximum Weather Information with Compression! ^_^");

//设定作业的启动类/
job.setJarByClass(MaxTemperatureWithCompression.class);

//解析输入和输出参数，分别作为作业的输入和输出，都是文件
FileInputFormat.addInputPath(job, new Path(args[0]));
FileOutputFormat.setOutputPath(job, new Path(args[1]));

//配置作业，设定Mapper类，Reducer类
job.setMapperClass(MaxTemperatureMapper.class);
job.setReducerClass(MaxTemperatureReducer.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);

System.exit(job.waitForCompletion(true)?0:1);

}

}

要运行这个例子，我们需要给出输入文件，因为Hadoop系统可以根据输入文件的扩展名自动识别基本文件，所以我们创建目录结构，并且上传一个gzip格式的文件作为map-reduce过程的输入：





然后我们运行的main中传入HDFS的输入文件和输出目录：





当执行完成之后，我们就可以在HDFS文件系统中看到最终的输出结果了，正如我们所预期的，这个结果是个gzip格式的文件：





通过日志观察压缩输出文件过程

我们可以观察日志来更细粒度的观察整个过程：

namenode:

2012-05-31 13:11:08,621 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=open    src=/user/hadoop-user/compress-input/1901.gz    dst=null    perm=null
2012-05-31 13:11:08,754 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=open    src=/user/hadoop-user/compress-input/1901.gz    dst=null    perm=null
2012-05-31 13:11:08,758 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=mkdirs  src=/user/hadoop-user/compress-output/_temporary    dst=null    perm=hadoop-user:supergroup:rwxr-xr-x
2012-05-31 13:11:08,853 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=open    src=/user/hadoop-user/compress-input/1901.gz    dst=null    perm=null
2012-05-31 13:11:09,203 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=create  src=/user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0/part-r-00000.gz dst=null    perm=hadoop-user:supergroup:rw-r--r--
2012-05-31 13:11:09,238 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* NameSystem.allocateBlock: /user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0/part-r-00000.gz. blk_-3869950436265612646_1016
2012-05-31 13:11:09,292 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* NameSystem.addStoredBlock: blockMap updated: 192.168.129.35:50010 is added to blk_-3869950436265612646_1016 size 29
2012-05-31 13:11:09,686 INFO org.apache.hadoop.hdfs.StateChange: DIR* NameSystem.completeFile: file /user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0/part-r-00000.gz is closed by DFSClient_-356100022
2012-05-31 13:11:09,692 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=listStatus  src=/user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0 dst=null    perm=null
2012-05-31 13:11:09,695 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=mkdirs  src=/user/hadoop-user/compress-output   dst=null    perm=hadoop-user:supergroup:rwxr-xr-x
2012-05-31 13:11:09,698 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=rename  src=/user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0/part-r-00000.gz dst=/user/hadoop-user/compress-output/part-r-00000.gz   perm=hadoop-user:supergroup:rw-r--r--
2012-05-31 13:11:09,699 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=delete  src=/user/hadoop-user/compress-output/_temporary/_attempt_local_0001_r_000000_0 dst=null    perm=null
2012-05-31 13:11:09,703 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.40.16   cmd=delete  src=/user/hadoop-user/compress-output/_temporary    dst=null    perm=null
2012-05-31 13:11:51,010 INFO org.apache.hadoop.hdfs.server.namenode.FSNamesystem.audit: ugi=hadoop-user,hadoop-user ip=/192.168.129.35  cmd=listStatus  src=/user/hadoop-user/compress-output   dst=null    perm=null

datanode:

2012-05-31 13:11:08,864 INFO org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: /192.168.129.35:50010, dest: /192.168.40.16:6233, bytes: 74447, op: HDFS_READ, cliID: DFSClient_-356100022, srvID: DS-1002949858-192.168.129.35-50010-1337839176422, blockid: blk_-4455870079864415553_1015
2012-05-31 13:11:09,248 INFO org.apache.hadoop.hdfs.server.datanode.DataNode: Receiving block blk_-3869950436265612646_1016 src: /192.168.40.16:6234 dest: /192.168.129.35:50010
2012-05-31 13:11:09,283 INFO org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: /192.168.40.16:6234, dest: /192.168.129.35:50010, bytes: 29, op: HDFS_WRITE, cliID: DFSClient_-356100022, srvID: DS-1002949858-192.168.129.35-50010-1337839176422, blockid: blk_-3869950436265612646_1016
2012-05-31 13:11:09,283 INFO org.apache.hadoop.hdfs.server.datanode.DataNode: PacketResponder 0 for block blk_-3869950436265612646_1016 terminating

我们在这里清楚的看到在目标目录下生成gzip格式的输出文件的整个过程,假定namenode第i行日志设为N(i)，datanode第i行日志设为D(i),则执行顺序为：

N1->N2->N3->N4->D1->N5->N6->D2->D3->D4->N7->N8...->N14,

其中N1->N4是namenode做一些准备工作，包括打开输入文件和创建输出目录及其临时子目录。

D1是datanode读取输入文件

N5,N6按照命名规则和配置中压缩文件的设定，创建输出文件到临时目录下（此时这个文件为空），然后用NameSystem吧这个块分配给datanode

D2-D4是datanode写最终reduce结果到被分配的块中。

N7-N14则是namenode吧输出文件的位置复制到命令行第二个参数指定的位置中，作为最终输出结果

ps. 需要提及的一点是Conf中压缩的配置需要写在Job定义之前，否则压缩无法成功。

本文出自 “平行线的凝聚” 博客，请务必保留此出处/article/4268176.html