“初探Hadoop MapReduce NextGen|MRv2|YARN(工作上下文紧急切换， 该篇待续)”

fujohnwang

2011-12-16

人家年初提出来的想法，我年末才刚开始看， 惭愧…

Architecture

  1. further separation of concerns on JobTracker
    • resource management concern
    • job scheduling/monitoring concern
  2. not limited to MapReduce pattern, multiple different parallel programming patterns can be used too, like master-workers, MPI, iterative monads, etc. This is a more generic distributed parallel programming framework.

Components

  1. ResourceManager
    • Scheduler
    • ApplicationManager
  2. NodeManager
  3. ApplicationMaster

Interesting projects built on hadoop

  1. http://www.lilyproject.org/lily/index.html - LILY - SMART DATA, AT SCALE, MADE EASY

Hadoop in multiple com

##Hadoop @ 百度 1. 透明压缩 - linux idle class调度 - 异步block级别压缩

2. 资源隔离 
                - cgroup

References

  1. http://developer.yahoo.com/blogs/hadoop/posts/2011/02/mapreduce-nextgen/
  2. http://developer.yahoo.com/blogs/hadoop/posts/2011/03/mapreduce-nextgen-scheduler/
  3. http://www.mesosproject.org/
  4. http://developer.yahoo.com/blogs/hadoop/posts/2011/02/capacity-scheduler/
  5. http://wiki.apache.org/hadoop/SupportingProjects
>>>>>> 更多阅读 <<<<<<

「为AI疯狂」星球上，扶墙老师正在和朋友们讨论有趣的AI话题，你要不要⼀起来呀？^-^
这里

  1. 不但有及时新鲜的AI资讯和深度探讨
  2. 还分享AI工具、产品方法和商业机会
  3. 更有体系化精品付费内容等着你，加入星球(https://t.zsxq.com/0dI3ZA0sL) 即可免费领取。(加入之后一定记得看置顶消息呀！)

知识星球二维码

存量的时代，省钱就是赚钱。
在增量的时代，省钱其实是亏钱。
避坑儿是省钱的一种形式，更是真正聪明人的选择！
弯路虽然也是路，但还是能少走就少走，背后都是高昂的试错成本。
订阅「福报」，少踩坑，少走弯路，多走一步，就是不一样的胜率！

订阅「福报Premium订阅」