Allen Wittenauer commented on YARN-3170:

I'll be honest:  I greatly dislike that entire first paragraph. MRv2 needs to 
be struck from the vocabulary.  It was a marketing ploy to get YARN's 
acceptance into Hadoop as a subproject.   It also helps underscore the problems 
of what to call the "new" MR API in the actual MR subproject.  I'm inclined to 
think the entire paragraph should just get deleted.

The second paragraph should be rewritten.  There's little value in comparing 
YARN to earlier versions of Hadoop at this point.  Don't describe YARN in terms 
of the JobTracker.  If I'm new to Hadoop, I have no idea what the heck a JT 
even is.

{code} An application is either a single job in the classical sense of 
Map-Reduce jobs or a DAG of jobs.{code}

* Drop "in the classical sense of Map-Reduce jobs". 

{code} The ResourceManager and per-node slave, the NodeManager (*NM*), form the 
data-computation framework. The ResourceManager is the ultimate authority that 
arbitrates resources among all the applications in the system.

* Drop "per-node slave,"
* Drop "(*NM*)"
* Add a description of the node manager after the description of the resource 
manager in this paragraph.

> YARN architecture document needs updating
> -----------------------------------------
>                 Key: YARN-3170
>                 URL: https://issues.apache.org/jira/browse/YARN-3170
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: documentation
>            Reporter: Allen Wittenauer
>            Assignee: Brahma Reddy Battula
>              Labels: BB2015-05-TBR
>         Attachments: YARN-3170-002.patch, YARN-3170-003.patch, 
> YARN-3170-004.patch, YARN-3170.patch
> The marketing paragraph at the top, "NextGen MapReduce", etc are all 
> marketing rather than actual descriptions. It also needs some general 
> updates, esp given it reads as though 0.23 was just released yesterday.

This message was sent by Atlassian JIRA

Reply via email to