[
https://issues.apache.org/jira/browse/HIVE-28489?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17877925#comment-17877925
]
Sungwoo Park commented on HIVE-28489:
-------------------------------------
On 10TB TPC-DS benchmark (tested with Hive 4 on MR3),
query 18, before: 30.6s, after: 28.2s
query 22, before: 53.1s, after: 18.0s
query 67, before: 842.3s, after: 429.1s
> Partitioning the input data of Grouping Set GroupBy operator
> ------------------------------------------------------------
>
> Key: HIVE-28489
> URL: https://issues.apache.org/jira/browse/HIVE-28489
> Project: Hive
> Issue Type: Improvement
> Reporter: Seonggon Namgung
> Assignee: Seonggon Namgung
> Priority: Major
> Attachments: 2.PartitionDataBeforeGroupingSet.pptx
>
>
> GroupBy operator with grouping sets often emits too many rows, which becomes
> the bottleneck of query execution. To reduce the number output rows, this
> JIRA proposes partitioning the input data of such GroupBy operator.
> Please check out the attached slides for detailed explanation.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
