[
https://issues.apache.org/jira/browse/NUTCH-1033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13060640#comment-13060640
]
Julien Nioche commented on NUTCH-1033:
--------------------------------------
Please ignore my previous comment about committing - was about another issue.
Am a bit perplexed by this one - it's only in 2.0 that the fetching is done as
part of the reduce step. in 1.x it is still done in the map step; setting the
number of tasks to use for the reducer could be done thanks to a parameter as
per your patch but I can't think of a reason why we should not simply rely on
the generic mapred.map.tasks for doing this, especially as the value specified
would probably be the same across the various jobs involved in a crawl
(generate, fetch, parse etc...)
> Backport FetcherJob should run more reduce tasks than default
> -------------------------------------------------------------
>
> Key: NUTCH-1033
> URL: https://issues.apache.org/jira/browse/NUTCH-1033
> Project: Nutch
> Issue Type: Improvement
> Components: fetcher
> Affects Versions: 1.3, 1.4
> Reporter: Markus Jelsma
> Fix For: 1.4
>
> Attachments: NUTCH-1033-1.4-1.patch, NUTCH-1033-1.4-2.patch
>
>
> Andrzej wrote:"FetcherJob now performs fetching in the reduce phase. This
> means that in a typical Hadoop setup there will be many fewer reduce tasks
> than map tasks, and consequently the max. total throughput of Fetcher will be
> proportionally reduced. I propose that FetcherJob should set the number of
> reduce tasks to the number of map tasks. This way the fetching will be more
> granular."
> This issue covers the backport of NUTCH-884 to Nutch 1.4-dev.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
