The official release of Apache Hadoop 2.6.0 does not include the required binaries (e.g., winutils.exe) necessary to run hadoop. In order to use Hadoop on Windows, it must be compiled from source. This takes a bit of effort, so I’ve provided a pre-compiled, unofficial distribution below:

• hadoop-2.6.0.tar.gz
  SHA1: 205b235d77213b958f126647241a5092845a0ff8
• hadoop-dist-2.6.0-javadoc.jar
  SHA1: cbccbc5d7eeb03261e533cbe0b1b367fce83b181
• hadoop-2.6.0-src.tar.gz
  SHA1: 4a1dfa9bd34d5efb7f2a0cd4dcf03db3eab46a5d

I compiled the source using:

• Apache Maven 3.2.5
• CMake 3.1.0
• MinGW: Minimalist GNU for Windows, specifically with sh, mkdir, rm, cp, tar, and gzip.
• Oracle Java 7u71. Note that the newer Java 8 will not work out of the box due to doclint errors.
• Google Protocol Buffers 2.5.0. The pom file is hard-coded to use 2.5.0.
• Microsoft Windows SDK for Windows 7 and .NET Framework. Alternatively, you may use Visual Studio 2010 Professional. Visual Studio 2012 and later will not work without modification of the build scripts.

Then, using the Windows SDK 7.1 Command Prompt or Visual Studio Command Prompt (2010):

set JAVA_HOME=C:\PROGRA~1\Java\jdk1.7.0_71
set Platform=x64 

The build system requires that you use the 8.3 short filename for JAVA_HOME (no spaces!). The environment variables (Platform) are also case sensitive. Finally:

mvn package -Pdist -DskipTests -Dtar 

The binaries will be available in hadoop-dist/target.