hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jay Booth (JIRA)" <j...@apache.org>
Subject [jira] Commented: (MAPREDUCE-326) The lowest level map-reduce APIs should be byte oriented
Date Thu, 04 Feb 2010 21:35:29 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12829798#action_12829798
] 

Jay Booth commented on MAPREDUCE-326:
-------------------------------------

Well consider me an enthusiastic +whatever :)

In the "crazy ideas to spur discussion" vein, maybe we could consider delegating reduce output
to the framework as well?  Reducers don't have to write anything to stdout except for maybe
progress reports?  That would open up:
* reduce outputs that go to multiple locations on the filesystem (could be a big win for jobs
producing hive partitions)
* reduce outputs which send results directly to the launching process somehow, avoiding an
intermediate file write/read

That's sort of in crazy land but figured I'd throw the idea out there -- might not be worth
the added burden on frameworks to support super edgey cases when you can always read a file
back in after the job or move it or whatever.

Anyways, awesome idea, I think if we got something like this proposal in place on top of avro's
promised cross-language easyness, a bunch of mapreduce frameworks in different languages could
be developed really easily.

> 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