[
https://issues.apache.org/jira/browse/MAPREDUCE-326?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12829993#action_12829993
]
eric baldeschwieler commented on MAPREDUCE-326:
-----------------------------------------------
heh! I still like this idea, but don't see it as the first thing we should
think about.
We'll want to be sure we do this in a performance and API neutral or positive
way if we do it.
I think a long term goal should be to refactor the whole MR framework into a
resource manager that is map-reduce unaware and "user" or library code that
executes MR.
If we got there, then we could have competing implementations of MR that did
this either way and let folks compare and contrast. I'd suggest that it might
make more sense to invest in moving the MR framework into "user space" rather
than replumbing the current implementation to be binary first.
We have a V2 MR framework bug kicking about somewhere. Anyone know it?
> 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.
