[
https://issues.apache.org/jira/browse/PIG-1264?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12841984#action_12841984
]
Richard Ding commented on PIG-1264:
-----------------------------------
Somehow during the many merges between the trunk and LSR branch, the random
sampler is used in place of Poisson sampler for skewed join. This was corrected
with PIG-1273.
> Skewed join sampler misses out the key with the highest frequency
> -----------------------------------------------------------------
>
> Key: PIG-1264
> URL: https://issues.apache.org/jira/browse/PIG-1264
> Project: Pig
> Issue Type: Bug
> Affects Versions: 0.7.0
> Reporter: Sriranjan Manjunath
> Assignee: Richard Ding
> Fix For: 0.7.0
>
>
> I am noticing two issues with the sampler used in skewed join:
> 1. It does not allocate multiple reducers to the key with the highest
> frequency.
> 2. It seems to be allocating the same number of reducers to every key (8 in
> this case).
> Query:
> a = load 'studenttab10k' using PigStorage() as (name, age, gpa);
> b = load 'votertab10k' as (name, age, registration, contributions);
> e = join a by name right, b by name using "skewed" parallel 8;
> store e into 'SkewedJoin_9.out';
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
