[
https://issues.apache.org/jira/browse/HIVE-18079?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16275835#comment-16275835
]
Gopal V commented on HIVE-18079:
--------------------------------
The patch decreases the memory used in the metastore by 16x, while staying
under the hive.stats.ndv.error defaults (which is ~20%).
> Statistics: Allow HyperLogLog to be merged to the lowest-common-denominator
> bit-size
> ------------------------------------------------------------------------------------
>
> Key: HIVE-18079
> URL: https://issues.apache.org/jira/browse/HIVE-18079
> Project: Hive
> Issue Type: Improvement
> Components: Standalone Metastore, Statistics
> Affects Versions: 3.0.0
> Reporter: Gopal V
> Assignee: Gopal V
> Attachments: HIVE-18079.1.patch
>
>
> HyperLogLog can merge a 14 bit HLL into a 10 bit HLL bitset, because of its
> mathematical hash distribution & construction.
> Allow the squashing of a 14 bit HLL -> 10 bit HLL without needing a second
> scan over the data-set.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
