[
https://issues.apache.org/jira/browse/PIG-895?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daniel Dai updated PIG-895:
---------------------------
Attachment: PIG-895-1.patch
Preliminary patch for this. Need to add unit test.
> Default parallel for Pig
> ------------------------
>
> Key: PIG-895
> URL: https://issues.apache.org/jira/browse/PIG-895
> Project: Pig
> Issue Type: New Feature
> Components: impl
> Affects Versions: 0.3.0
> Reporter: Daniel Dai
> Fix For: 0.4.0
>
> Attachments: PIG-895-1.patch
>
>
> For hadoop 20, if user don't specify the number of reducers, hadoop will use
> 1 reducer as the default value. It is different from previous of hadoop, in
> which default reducer number is usually good. 1 reducer is not what user want
> for sure. Although user can use "parallel" keyword to specify number of
> reducers for each statement, it is wordy. We need a convenient way for users
> to express a desired number of reducers. Here is my propose:
> 1. Add one property "default_parallel" to Pig. User can set default_parallel
> in script. Eg:
> set default_parallel '10';
> 2. default_parallel is a hint to Pig. Pig is free to optimize the number of
> reducers (unlike parallel keyword). Currently, since we do not have a
> mechanism to determine the optimal number of reducers, default_parallel will
> be always granted, unless it is override by "parallel" keyword.
> 3. If user put multiple default_parallel inside script, the last entry will
> be taken.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
