[
https://issues.apache.org/jira/browse/HIVE-23032?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17065807#comment-17065807
]
Peter Vary commented on HIVE-23032:
-----------------------------------
[~dkuzmenko]:
Thanks for the patch, few ideas:
* Could you please add the {{rewriteBatchedStatements}},
{{reWriteBatchedInserts}} configurations to other DataSources as well (BoneCP,
DBCP)?
* Maybe cache not the date, but the whole getDbEpochFn result in case of
Derby?
* Maybe query the DIRECT_SQL_MAX_ELEMENTS_VALUES_CLAUSE only once before the
loop?
* What happens if the DIRECT_SQL_MAX_ELEMENTS_VALUES_CLAUSE is 10, and we have
exactly 10 lock components?
* Would be good to have some perf results for differnt batch sizes. Do not
have to be exhaustive, just to have a clear understanding how batch size
changes the equation.
Thanks for the patch!
Peter
> Add batching in Lock generation
> -------------------------------
>
> Key: HIVE-23032
> URL: https://issues.apache.org/jira/browse/HIVE-23032
> Project: Hive
> Issue Type: Improvement
> Components: Locking
> Reporter: Denys Kuzmenko
> Assignee: Denys Kuzmenko
> Priority: Major
> Attachments: HIVE-23032.1.patch, HIVE-23032.2.patch,
> HIVE-23032.3.patch, HIVE-23032.4.patch
>
>
> Replace multi-row insert in Oracle with batching. Performance tests showed
> significant performance improvement after turning batching on.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
