[
https://issues.apache.org/jira/browse/HIVE-11107?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15036686#comment-15036686
]
Ashutosh Chauhan commented on HIVE-11107:
-----------------------------------------
* It will be more useful to have tez plans for this instead of MR, since MR is
deprecated.
* Move function setupMetaStoreTableColumnStatsFor30TBTPCDSWorkload() to
QTestUtil that way it will be easier to maintain, since its in java file and
not in template.
* Derby is now integrated into jdk7 so instead of accessing over jdbc
connection, you may want to use apis directly, that will make it much easier to
maintain and debug this function.
http://www.oracle.com/technetwork/java/javadb/overview/javadb-156712.html
> Support for Performance regression test suite with TPCDS
> --------------------------------------------------------
>
> Key: HIVE-11107
> URL: https://issues.apache.org/jira/browse/HIVE-11107
> Project: Hive
> Issue Type: Task
> Reporter: Hari Sankar Sivarama Subramaniyan
> Assignee: Hari Sankar Sivarama Subramaniyan
> Attachments: HIVE-11107.1.patch, HIVE-11107.2.patch,
> HIVE-11107.3.patch, HIVE-11107.4.patch
>
>
> Support to add TPCDS queries to the performance regression test suite with
> Hive CBO turned on.
> This benchmark is intended to make sure that subsequent changes to the
> optimizer or any hive code do not yield any unexpected plan changes. i.e.
> the intention is to not run the entire TPCDS query set, but just "explain
> plan" for the TPCDS queries.
> As part of this jira, we will manually verify that expected hive
> optimizations kick in for the queries (for given stats/dataset). If there is
> a difference in plan within this test suite due to a future commit, it needs
> to be analyzed and we need to make sure that it is not a regression.
> The test suite can be run in master branch from itests by
> {code}
> mvn test -Dtest=TestPerfCliDriver -Phadoop-2
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
