hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "eric baldeschwieler (JIRA)" <j...@apache.org>
Subject [jira] Commented: (MAPREDUCE-326) The lowest level map-reduce APIs should be byte oriented
Date Mon, 08 Feb 2010 07:41:31 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12830867#action_12830867
] 

eric baldeschwieler commented on MAPREDUCE-326:
-----------------------------------------------

I'm very interested in proposals for how this can be done.  If it can be done in a performance
and API neutral way, than yes, interested parties should pursue it.

The issue here is just that we need to hold a reasonably high bar on performance and API changes
at this point.  Incremental checkins that change user facing APIs in a non-backwards compatible
way are a challenge.  Ditto performance regressions.  I owe the community a blog post on the
challenges we are facing with our user base on backwards compatibility and why this has lead
us to choose not to deploy the 21 release at Yahoo.

When I suggest it might be better to defer this work, it is looking at it through that lens.
 And that is the lens I suggest reviewers use considering patches in this area.  Hope that
clarifies things!  Go forth and propose improvements in this area, I'm a fan.  Just be aware
that since hadoop's users are demanding we behave as if hadoop is 1.0, think about supporting
them while thinking about how to manage this.

> 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