Since the last major release in the Hadoop 2.x series, hadoop-2.1.0-beta represents a significant set of enhancements:
- API & Protocol Stabilization. The community is now very confident that we can, henceforth, support full compatibility (both API & wire-protocol) for applications built on both HDFS & YARN. See HADOOP-8990 & YARN-386 for details.
- Binary Compatibility for existing MapReduce applications built for hadoop-1.x. The community decided, at the beginning of the year, to support full binary compatibility for existing MapReduce applications built for Apache Hadoop 1.x i.e. current stable release. This meant one could take an existing MapReduce application (jars, scripts etc.) and run them unchanged on both hadoop-1.x and hadoop-2.x. This release represents a culmination of that effort and removes last of the barriers for adoption by easing migration from hadoop-1.x to hadoop-2.x. See MAPREDUCE-5108 for more details.
- Support for Microsoft Windows. As most people are aware, engineers from Microsoft & Hortonworks have been collaborating with the community to support Hadoop on Windows. As the first official ASF release of Apache Hadoop to support Hadoop 2.x on Microsoft Windows, this represents a major milestone. See HADOOP-8562 for details.
- HDFS Snapshots. This is the first Apache Hadoop release from the ASF which has full-support HDFS snapshots. See HDFS-2802 for more details.
- NFS-v3 Access for HDFS. This is the first Apache Hadoop release from the ASF which has full-support for NFSv3 access to HDFS. See HDFS-4750 for more details.
- Client APIs for YARN Application Developers. The YARN developer community has completely revamped and simplified client libraries for people developing new YARN applications. See YARN-418 for more details.
- Integration Testing. Substantial amount of integration testing with the the entire Apache Hadoop ecosystem including Apache HBase, Apache Pig, Apache Hive etc.
See also: http://bit.ly/15dCSNN