[
https://issues.apache.org/jira/browse/TAJO-1049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14141816#comment-14141816
]
ASF GitHub Bot commented on TAJO-1049:
--------------------------------------
Github user jinossy commented on the pull request:
https://github.com/apache/tajo/pull/147#issuecomment-56259080
+1
Cluster limitation was temporary solution.
> Remove the parallel degree limit up to the maximum cluster capacity
> -------------------------------------------------------------------
>
> Key: TAJO-1049
> URL: https://issues.apache.org/jira/browse/TAJO-1049
> Project: Tajo
> Issue Type: Improvement
> Components: data shuffle, distributed query plan
> Reporter: Hyunsik Choi
> Assignee: Hyunsik Choi
> Priority: Trivial
> Fix For: 0.9.0
>
> Attachments: TAJO-1049.patch, TAJO-1049_2.patch, TAJO-1049_3.patch
>
>
> The maximum parallel degree had been limited by the number of cluster node
> capacity. This was because the old hash shuffle approach created too many
> small files depending on the number of parallel degree of a stage.
> But, TAJO-992 solved this problem. We don't need to use this limit anymore.
> This patch removes the limit.
> I think that this patch requires cluster test and it is hard to verify this
> patch in unit tests.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
