[jira] [Commented] (KYLIN-867) Hybrid model for multiple realizations/cubes

["fengYu (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KYLIN-867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14738356#comment-14738356 ] 

fengYu commented on KYLIN-867:
------------------------------

I has a questing about this : if I build a cube with 3 dimensions A/B/C, segment partition date is [M, N), some days later, I want to add 2 more dimensions D/E to this cube model, and build a new segment partition date is [N, S), now, I want to query this sql : select A, D, count(1) from fact_table group by A, D, what is the result ?I think the result can only generate by the new segment which means the result do not contain datas in [M,N). 
Is that true? 

> Hybrid model for multiple realizations/cubes
> --------------------------------------------
>
>                 Key: KYLIN-867
>                 URL: https://issues.apache.org/jira/browse/KYLIN-867
>             Project: Kylin
>          Issue Type: Improvement
>          Components: Metadata, Query Engine
>            Reporter: Shaofeng SHI
>            Assignee: Shaofeng SHI
>             Fix For: v1.0
>
>
> Scenario: users have a cube already built with the history data; Now they create a new cube which adding more dimensions; Due to some reason, user expects both the history cube and current cube can be scanned to get a full result set;
> A hybrid realization will be introduced, which is composed by multiple cubes, and these cubes have no overlap/gap on the partition date; Then the hybrid will delegate the query to all cubes, and then merge the results;  



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)