[
https://issues.apache.org/jira/browse/FLINK-11425?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16753944#comment-16753944
]
Stephan Ewen commented on FLINK-11425:
--------------------------------------
I would suggest to reach out to [~ykt836] to see if that makes sense at this
point and, if yes, to coordinate with him.
> Support of “Hash Teams” in Hybrid Hash Join
> -------------------------------------------
>
> Key: FLINK-11425
> URL: https://issues.apache.org/jira/browse/FLINK-11425
> Project: Flink
> Issue Type: New Feature
> Components: Core, Optimizer
> Reporter: LiuJi
> Priority: Major
>
> Hybrid Hash Join is already supported in current version. The join starts
> operating in memory and gradually starts spilling contents to disk, when the
> memory is not sufficient.
>
> Current hash join only support two inputs, so when a job contains multiple
> hash joins which have the same join keys, it will consume some unnecessary
> resources (I/O, memory, etc) because some upstream output data may useless
> for downstream hash join.
>
> According to the above observations, we want to provide a HashTeamManager to
> implement multiway inputs hash join by combining several two way hash join
> which have same join keys. HashTeamManager manage the relations of multiple
> HashTables and improve efficiency in memory use and lower I/O operations by
> joining multiple relations at one time.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
