Apache Hadoop 2.7.0 發布

Apache Hadoop 2.7.0發布。一共修復了來自社區的535個JIRAs，其中：Hadoop Common有160個；HDFS有192個；YARN有148個；MapReduce有35個。Hadoop 2.7.0是2015年第一個Hadoop release版本，不過需要注意的是
（1）、不要將Hadoop 2.7.0用于生產環境，因為一些關鍵Bug還在測試中，如果需要在生產環境使用，需要等Hadoop 2.7.1/2.7.2，這些版本很快會發布。
（2）、這個版本的Hadoop已經不支持JDK 1.6運行環境，只支持JDK 7+。
此版本主要做的改進如下：

Hadoop Common

1、支持Windows Azure Storage，BLOB作為Hadoop中的文件系統。

Hadoop HDFS

1、支持文件截斷（file truncate）；
2、支持每個存儲類型配額(Support for quotas per storage type);
3、支持可變長度的塊文件

Hadoop YARN

1、YARN安全模塊可插拔
2、YARN的本地化資源可以自動共享，全局緩存（測試版）

Hadoop MapReduce

1、能夠限制運行的Map/Reduce作業的任務
2、為非常的大Job（有許多輸出文件）加快了FileOutputCommitter。

英文原文

Hadoop Common
Windows Azure Storage Blob support (HADOOP-9629), available in trunk for a while, is now integrated into branch-2 and released as part of 2.7.0.
Hadoop HDFS
Enable new read/write scenarios in HDFS by adding support for truncate (HDFS-3107) and support for files with variable-length blocks (HDFS-3689)
Enforce quotas at the Heterogeneous Storage Type granularity in HDFS (HDFS-7584)
Enhance management (HDFS-7424) and monitoring (HDFS-7449) for the NFS Gateway Server
Hadoop YARN
YARN-3100 – Makes YARN authorization pluggable so that tools like Apache Ranger can provide authorization for YARN job submission operations
YARN-1492 – Automatic shared, global caching of YARN localized resources (beta)
Hadoop MapReduce
MAPREDUCE-5583 – Allows the ability to limit the size of a running MapReduce job by restricting the maximum number of Map or Reduce tasks running at any point in time
MAPREDUCE-4815 – Speed up Hive, Pig and MapReduce jobs that deal with many output files by making enhancements to the FileOutputCommitter

Hadoop2.7.0下載

http://hadoop.apache.org/releases.html#Download