https://blog.youkuaiyun.com/npucloudzh-cnhttps://blog.youkuaiyun.com/https://blog.youkuaiyun.com/npucloud/article/details/6597867https://blog.youkuaiyun.com/npucloud/article/details/6597867npucloudMon, 11 Jul 2011 15:55:26 +0800https://blog.youkuaiyun.com/npucloud/article/details/6544834https://blog.youkuaiyun.com/npucloud/article/details/6544834npucloudTue, 14 Jun 2011 22:08:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6541577https://blog.youkuaiyun.com/npucloud/article/details/6541577npucloudMon, 13 Jun 2011 16:07:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6450921https://blog.youkuaiyun.com/npucloud/article/details/6450921npucloudFri, 27 May 2011 20:50:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6432145https://blog.youkuaiyun.com/npucloud/article/details/6432145npucloudThu, 19 May 2011 14:53:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6429692https://blog.youkuaiyun.com/npucloud/article/details/6429692npucloudWed, 18 May 2011 15:31:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6429688https://blog.youkuaiyun.com/npucloud/article/details/6429688npucloudWed, 18 May 2011 15:30:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6429684https://blog.youkuaiyun.com/npucloud/article/details/6429684npucloudWed, 18 May 2011 15:29:00 +0800
1、概述
1TB排序通常用于衡量分布式数据处理框架的数据处理能力。Terasort是Hadoop中的的一个排序作业，在2008年，Hadoop在1TB排序基准评估中赢得第一名 ，耗时209秒。那么Terasort在Hadoop中是怎样实现的呢？本文主要从算法设计角度分析Terasort作业。
2、算法思想
实际上，当我们要把传统的串行排序算法设计成并行的排序算法时，通常会想到分而治之的策略，即：把要排序的数据划成M个数据块（可以用Hash的方 法做到），]]>https://blog.youkuaiyun.com/npucloud/article/details/6414112https://blog.youkuaiyun.com/npucloud/article/details/6414112npucloudThu, 12 May 2011 11:26:00 +0800https://blog.youkuaiyun.com/npucloud/article/details/6414072https://blog.youkuaiyun.com/npucloud/article/details/6414072npucloudThu, 12 May 2011 11:16:00 +0800