hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (JIRA)" <j...@apache.org>
Subject [jira] Commented: (MAPREDUCE-326) The lowest level map-reduce APIs should be byte oriented
Date Thu, 04 Feb 2010 19:02:28 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12829699#action_12829699
] 

Todd Lipcon commented on MAPREDUCE-326:
---------------------------------------

I'm worried about performance if you demote Java to a streaming-like interface. The splice
system call, afaik, allows you to move data between two pipes without going through userland.
If you want to actually look at the data, it doesn't really help much. In order to do the
map side sort or the reduce side shuffle, it will have to look at all the data coming out
of the mapper (or into the reducer), no?

Is the assumption also that the InputFormat would run entirely in the child process, and we'd
reimplement the common inputformats in every language? This raises issues for splits other
than FileSplit. How do we turn a dbinputformat split into a bytebuffer if the inputformat
lives on the child side?

> The lowest level map-reduce APIs should be byte oriented
> --------------------------------------------------------
>
>                 Key: MAPREDUCE-326
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-326
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>            Reporter: eric baldeschwieler
>
> As discussed here:
> https://issues.apache.org/jira/browse/HADOOP-1986#action_12551237
> The templates, serializers and other complexities that allow map-reduce to use arbitrary
types complicate the design and lead to lots of object creates and other overhead that a byte
oriented design would not suffer.  I believe the lowest level implementation of hadoop map-reduce
should have byte string oriented APIs (for keys and values).  This API would be more performant,
simpler and more easily cross language.
> The existing API could be maintained as a thin layer on top of the leaner API.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message