[
https://issues.apache.org/jira/browse/HIVE-29824?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thomas Rebele updated HIVE-29824:
---------------------------------
Description:
When certain column level statistics (e.g., histograms) are stored at the
partition level, they require a lot of disk space. Hive should provide a way so
that the user can decide to store the statistics at the table level instead.
It needs to be decided whether to configure the property globally, for a table,
or a particular column, and which types of statistics (histograms, NDV/HLL).
This would provide a workaround for the KLL instability caused by the non-fixed
seed when merging KLL datasketches.
was:
When certain column level statistics (e.g., histograms) are stored at the
partition level, they require a lot of disk space. Hive should provide a way so
that the user can decide to store the statistics at the table level instead.
It needs to be decided whether to configure the property globally, for a table,
or a particular column.
This would provide a workaround for the KLL instability caused by the non-fixed
seed when merging KLL datasketches.
> Add possibility to store column level statistics at the table level
> -------------------------------------------------------------------
>
> Key: HIVE-29824
> URL: https://issues.apache.org/jira/browse/HIVE-29824
> Project: Hive
> Issue Type: New Feature
> Reporter: Thomas Rebele
> Priority: Major
>
> When certain column level statistics (e.g., histograms) are stored at the
> partition level, they require a lot of disk space. Hive should provide a way
> so that the user can decide to store the statistics at the table level
> instead.
> It needs to be decided whether to configure the property globally, for a
> table, or a particular column, and which types of statistics (histograms,
> NDV/HLL).
> This would provide a workaround for the KLL instability caused by the
> non-fixed seed when merging KLL datasketches.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
