[jira] [Commented] (KYLIN-2064) add non-runtime-aggregation measure support

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

    [ https://issues.apache.org/jira/browse/KYLIN-2064?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15635022#comment-15635022 ] 

Shaofeng SHI commented on KYLIN-2064:
-------------------------------------

Hi Yu, I understand your point; it is almost the same as the "Holistic Count Distinct" measure in Kylin's early version (1.3 may still have); It can fulfill some specific scenario but it is inflexible because need the query must not has post-aggregation, so later Kylin removed that piece of code, and now all measures in Kylin need be mergable. 

> add non-runtime-aggregation measure support
> -------------------------------------------
>
>                 Key: KYLIN-2064
>                 URL: https://issues.apache.org/jira/browse/KYLIN-2064
>             Project: Kylin
>          Issue Type: Improvement
>    Affects Versions: v1.5.2
>            Reporter: fengYu
>            Assignee: fengYu
>         Attachments: 0001-KYLIN-2064-add-non-runtime-aggregation-measure-for-d.patch
>
>
> Kylin is based on pre-computation and store result in hbase. however, It support runtime aggregation to satisfy the query which can not match computed data.
> But the runtime aggregation slow down the query and need more storage space in hbase(which will slow down scan speed),  such as distinct count/ topn measures .
> So we can use more pre-compution and less runtime aggregation, in some scenario we do not need results cross different partition(date),  we add a kind of measure which only support result for count distinct, it will speed up query and need less storage.
> what's more, If a query on this measure which is not computed it will return exception.
> I will arrange our solution in my patch later.



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