[
https://issues.apache.org/jira/browse/HIVE-5634?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Shelukhin updated HIVE-5634:
-----------------------------------
Description:
See HIVE-4002.
We can replace OutputCollector in ReduceSink to output to say a sequence file.
Then instead of fetching file output written by file sinks in map tasks, the
client-side reducer can fetch reducer output via some simple operator and work
the same way as normal reducer.
It can also take advantage of additional ReduceSink functionality.
was:
See HIVE-4002.
We can replace OutputCollector in ReduceSink to output to say a sequence file.
Then instead of fetching file output written by file sinks in map tasks, the
client-side reducer can fetch reducer output via some simple task and work the
same way as normal reducer.
It can also take advantage of additional ReduceSink functionality.
> use ReduceSinkOperator when outputting for client-side reducer
> --------------------------------------------------------------
>
> Key: HIVE-5634
> URL: https://issues.apache.org/jira/browse/HIVE-5634
> Project: Hive
> Issue Type: Improvement
> Reporter: Sergey Shelukhin
>
> See HIVE-4002.
> We can replace OutputCollector in ReduceSink to output to say a sequence
> file. Then instead of fetching file output written by file sinks in map
> tasks, the client-side reducer can fetch reducer output via some simple
> operator and work the same way as normal reducer.
> It can also take advantage of additional ReduceSink functionality.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
