You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kylin.apache.org by "Zhong Yanghong (JIRA)" <ji...@apache.org> on 2017/09/08 02:11:00 UTC
[jira] [Updated] (KYLIN-2728) Introduce a new cuboid scheduler
based on cuboid tree rather than static rules
[ https://issues.apache.org/jira/browse/KYLIN-2728?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhong Yanghong updated KYLIN-2728:
----------------------------------
Attachment: APACHE-KYLIN-2783+APACHE-KYLIN-2728.patch
> Introduce a new cuboid scheduler based on cuboid tree rather than static rules
> ------------------------------------------------------------------------------
>
> Key: KYLIN-2728
> URL: https://issues.apache.org/jira/browse/KYLIN-2728
> Project: Kylin
> Issue Type: Sub-task
> Affects Versions: v2.2.0
> Reporter: Zhong Yanghong
> Assignee: Zhong Yanghong
> Attachments: APACHE-KYLIN-2783+APACHE-KYLIN-2728.patch
>
>
> Rather than depending on static rules to define a parent-child relationship between cuboids, a cuboid tree can be constructed for this relationship. In this tree, the root is the base cuboid. For a cuboid, it may have many legal parent cuboids. We should choose the {color:#f79232}best parent {color}cuboid as its parent in this tree. Then how to define the best? It's related two rules as follows:
> * If we know the row count, choose the one with the least row count
> * Otherwise, choose the smallest cuboid
> Based on the cuboid tree, we introduced a new cuboid scheduler called {{TreeCuboidScheduler}}. And both cube building & query can be based on {{TreeCuboidScheduler}}.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)