[
https://issues.apache.org/jira/browse/HIVE-16337?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chao Sun updated HIVE-16337:
----------------------------
Comment: was deleted
(was: cc [~lirui].)
> HoS: use separate config for mapjoin hash table size limit rather than
> hive.auto.convert.join.noconditionaltask.size
> --------------------------------------------------------------------------------------------------------------------
>
> Key: HIVE-16337
> URL: https://issues.apache.org/jira/browse/HIVE-16337
> Project: Hive
> Issue Type: Improvement
> Components: Spark
> Reporter: Chao Sun
>
> HoS currently share the same config
> {{hive.auto.convert.join.noconditionaltask.size}} with MR, however it
> interprets it differently: in MR this is the file size of the input, while in
> Spark it is the data size. This could cause Spark to be over conservative in
> many mapjoin optimizations. I suggest we use a separate config for HoS for
> this purpose.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
