[
https://issues.apache.org/jira/browse/TAJO-2007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15055590#comment-15055590
]
ASF GitHub Bot commented on TAJO-2007:
--------------------------------------
Github user jihoonson commented on the pull request:
https://github.com/apache/tajo/pull/900#issuecomment-164371417
Hi @hyunsik, I found some query plans are changed, especially related to
broadcast join. I suppose this change is due to the table volumes stored in
catalog, but it is difficult to find where the code is. Would you let me know
some points where need to be reviewed?
> By default, Optimizer should use the table volume in TableStat.
> ---------------------------------------------------------------
>
> Key: TAJO-2007
> URL: https://issues.apache.org/jira/browse/TAJO-2007
> Project: Tajo
> Issue Type: Improvement
> Components: Planner/Optimizer
> Reporter: Hyunsik Choi
> Assignee: Hyunsik Choi
> Fix For: 0.12.0, 0.11.1
>
>
> Currently, the optimizer by default gets table volumes through storage
> manager and employ them for join optimization. But, in some cases, it causes
> performance degradation because aggregating all file volumes is not cheap in
> large partitioned tables on S3 or HDFS.
> So, this patch improves TableStatUpdateRewriter to use table volumes of
> TableStat by default, and it also adds a session variable 'USE_TABLE_VOLUME'
> to allow the optimizer to use the table volume through storage handler.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
