[
https://issues.apache.org/jira/browse/HADOOP-1986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12534099
]
Doug Cutting commented on HADOOP-1986:
--------------------------------------
> I think that 2 levels of indirection are overkill, but I'm open to discussing
> it.
In most scenarios we've discussed, this isn't actually removing a level of
indirection but moving it out of the inner loop, where it might be a problem.
JVMs can do wonderful things, but removing a method call from an inner loop
still frequently makes things faster. It could become significant in, e.g., a
<String,String> mapreduce job that's, e.g., inverting links. If we can't make
that significant, it *might* even argue that we've not optimized things enough
elsewhere!
This is an area where we want both transparent simplicity, supporting things
like introspection, and also want to permit the highest performance, since it
is the innermost loop for many applications.
> Add support for a general serialization mechanism for Map Reduce
> ----------------------------------------------------------------
>
> Key: HADOOP-1986
> URL: https://issues.apache.org/jira/browse/HADOOP-1986
> Project: Hadoop
> Issue Type: New Feature
> Components: mapred
> Reporter: Tom White
> Assignee: Tom White
> Fix For: 0.16.0
>
> Attachments: SerializableWritable.java
>
>
> Currently Map Reduce programs have to use WritableComparable-Writable
> key-value pairs. While it's possible to write Writable wrappers for other
> serialization frameworks (such as Thrift), this is not very convenient: it
> would be nicer to be able to use arbitrary types directly, without explicit
> wrapping and unwrapping.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
