Apache > Hadoop > Core
 

Native Libraries Guide

はじめに

Hadoop has native implementations of certain components for reasons of both performance and non-availability of Java implementations. These components are available in a single, dynamically-linked, native library. On the *nix platform it is libhadoop.so. This document describes the usage and details on how to build the native libraries.

Components

Hadoop currently has the following compression codecs as the native components:

Of the above, the availability of native hadoop libraries is imperative for the gzip and bzip2 compression codecs to work.

使い方

It is fairly simple to use the native hadoop libraries:

  • Take a look at the supported platforms.
  • Either download the pre-built 32-bit i386-Linux native hadoop libraries (available as part of hadoop distribution in lib/native directory) or build them yourself.
  • Make sure you have any of or all of >zlib-1.2, >gzip-1.2, and >bzip2-1.0 packages for your platform installed; depending on your needs.

The bin/hadoop script ensures that the native hadoop library is on the library path via the system property -Djava.library.path=<path>.

To check everything went alright check the hadoop log files for:

DEBUG util.NativeCodeLoader - Trying to load the custom-built native-hadoop library...
INFO util.NativeCodeLoader - Loaded the native-hadoop library

If something goes wrong, then:

INFO util.NativeCodeLoader - Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

Supported Platforms

Hadoop native library is supported only on *nix platforms only. Unfortunately it is known not to work on Cygwin and Mac OS X and has mainly been used on the GNU/Linux platform.

It has been tested on the following GNU/Linux distributions:

On all the above platforms a 32/64 bit Hadoop native library will work with a respective 32/64 bit jvm.

Building Native Hadoop Libraries

Hadoop native library is written in ANSI C and built using the GNU autotools-chain (autoconf, autoheader, automake, autoscan, libtool). This means it should be straight-forward to build them on any platform with a standards compliant C compiler and the GNU autotools-chain. See supported platforms.

In particular the various packages you would need on the target platform are:

Once you have the pre-requisites use the standard build.xml and pass along the compile.native flag (set to true) to build the native hadoop library:

$ ant -Dcompile.native=true <target>

The native hadoop library is not built by default since not everyone is interested in building them.

You should see the newly-built native hadoop library in:

$ build/native/<platform>/lib

where <platform> is combination of the system-properties: ${os.name}-${os.arch}-${sun.arch.data.model}; for e.g. Linux-i386-32.

Notes

  • It is mandatory to have the zlib, gzip, and bzip2 development packages on the target platform for building the native hadoop library; however for deployment it is sufficient to install one of them if you wish to use only one of them.
  • It is necessary to have the correct 32/64 libraries of both zlib depending on the 32/64 bit jvm for the target platform for building/deployment of the native hadoop library.

Loading native libraries through DistributedCache

User can load native shared libraries through DistributedCache for distributing and symlinking the library files

Here is an example, describing how to distribute the library and load it from map/reduce task.

  1. First copy the library to the HDFS.
    bin/hadoop fs -copyFromLocal mylib.so.1 /libraries/mylib.so.1
  2. The job launching program should contain the following:
    DistributedCache.createSymlink(conf);
    DistributedCache.addCacheFile("hdfs://host:port/libraries/mylib.so.1#mylib.so", conf);
  3. The map/reduce task can contain:
    System.loadLibrary("mylib.so");