[
https://issues.apache.org/jira/browse/KYLIN-4035?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jack updated KYLIN-4035:
------------------------
Description:
Kylin will calculate column cardinality when loading hive table. This stage is
only supported by MR engine without spark. I think spark engine should be used
in this stage because of the following:
1) Kylin users can choose which engine they apply when calculating column
cardinality;
2) Some good spark features(e.g. dynamic resource allocation) can be used;
3) The code written in spark is simple.
I finish this work and test ok. But "kylin.engine.spark-cardinality=true"
should be added in kylin.properties(default is false). Look forwards to
suggestions.
Best regards.
was:
Kylin will calculate column cardinality when loading hive table. This stage is
only supported by MR engine without spark. I think spark engine should be used
in this stage because of the following:
1) Kylin users can choose which engine they apply when calculating column
cardinality;
2) Some good spark features(e.g. dynamic resource allocation) can be used;
3) The code written in spark is simple.
I finish this work and test ok. But "kylin.engin.spark-cardinality=true" should
be added in kylin.properties(default is false). Look forwards to suggestions.
Best regards.
> Calculate column cardinality by using spark engine
> --------------------------------------------------
>
> Key: KYLIN-4035
> URL: https://issues.apache.org/jira/browse/KYLIN-4035
> Project: Kylin
> Issue Type: Improvement
> Components: Spark Engine
> Environment: kylin: master/3.0.0-alpha
> spark: 2.4.3
> hadoop: 2.6.5
> Reporter: Jack
> Assignee: Jack
> Priority: Minor
> Fix For: v3.0.0-beta
>
>
> Kylin will calculate column cardinality when loading hive table. This stage
> is only supported by MR engine without spark. I think spark engine should be
> used in this stage because of the following:
> 1) Kylin users can choose which engine they apply when calculating column
> cardinality;
> 2) Some good spark features(e.g. dynamic resource allocation) can be used;
> 3) The code written in spark is simple.
> I finish this work and test ok. But "kylin.engine.spark-cardinality=true"
> should be added in kylin.properties(default is false). Look forwards to
> suggestions.
> Best regards.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
