[
https://issues.apache.org/jira/browse/KYLIN-4888?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17280429#comment-17280429
]
ASF subversion and git services commented on KYLIN-4888:
--------------------------------------------------------
Commit a4a8480a05449b7ab19c7eace096dc3e39697dcb in kylin's branch
refs/heads/kylin-on-parquet-v2 from feng.zhu
[ https://gitbox.apache.org/repos/asf?p=kylin.git;h=a4a8480 ]
KYLIN-4888, Performance optimization of union query with spark engine
> Performance optimization of union query with spark engine
> ----------------------------------------------------------
>
> Key: KYLIN-4888
> URL: https://issues.apache.org/jira/browse/KYLIN-4888
> Project: Kylin
> Issue Type: Improvement
> Components: Spark Engine
> Affects Versions: v4.0.0-alpha
> Reporter: Feng Zhu
> Assignee: Feng Zhu
> Priority: Major
> Fix For: v4.0.0-GA
>
> Attachments: spark_union_plan_comparison, stages before.png,
> stages_after.png
>
>
> when using union query with spark engine, UnionPlan transforms OLAPUnionRel
> to spark
> DataFrame, when OLAPUnionRel.all = false, distinct transformation of spark
> will be used, but
> it's used in a loop which traversing the DataFrame collection so that we
> don't have an excepted optimized flattenUnion plan(the CombineUnions rule of
> spark optimize the distinct, but the nested union plan does not be
> flattened),there are so many stages in spark dag. Actuall, distinct
> transformation should be used only once at last.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
