hadoop-pig-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pradeep Kamath (JIRA)" <j...@apache.org>
Subject [jira] Updated: (PIG-545) PERFORMANCE: Sampler for order bys does not produce a good distribution
Date Wed, 18 Feb 2009 00:20:59 GMT

     [ https://issues.apache.org/jira/browse/PIG-545?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Pradeep Kamath updated PIG-545:

    Attachment: PIG-545-v4.patch

Attached patch which addresses observation (1) in the previous comment.

Regarding (2), the differences from the current SortPartitioner are:
 - In the configure of the WeightedRangePartitioner the entire sample file is loaded and a
hash is built for those sample elements which repeat across partition boundaries - this is
more expensive than previously but one time set up
- IN the getPartition() call, if the input element is present in the hash, then probabilistically
a partition is assigned based on the information in the hash else (as previously), a binary
search of the quantiles array is done to figure out the partition - overall this should be
nearly the same as previously performance wise.

Regarding (3), getPartition() is a hadoop interface method and hence we cannot throw any other
exception besides RuntimeException

> PERFORMANCE: Sampler for order bys does not produce a good distribution
> -----------------------------------------------------------------------
>                 Key: PIG-545
>                 URL: https://issues.apache.org/jira/browse/PIG-545
>             Project: Pig
>          Issue Type: Bug
>          Components: impl
>    Affects Versions: types_branch
>            Reporter: Alan Gates
>            Assignee: Pradeep Kamath
>             Fix For: types_branch
>         Attachments: PIG-545-v3.patch, PIG-545-v4.patch, WRP.patch, WRP1.patch
> In running tests on actual data, I've noticed that the final reduce of an order by has
skewed partitions.  Some reduces finish in a few seconds while some run for 20 minutes.  Getting
a better distribution should lead to much better performance for order by.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message