[jira] [Commented] (PHOENIX-3131) improve "order by " performance with aggregated query

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

    [ https://issues.apache.org/jira/browse/PHOENIX-3131?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15402939#comment-15402939 ] 

Samarth Jain commented on PHOENIX-3131:
---------------------------------------

[~tuyuri] - can you post explain plans for the two queries you mentioned. FWIW, doing an order by on a non-row key column will take longer. Also, does it make a difference if you don't include offset in your queries? 

> improve "order by " performance with aggregated query 
> ------------------------------------------------------
>
>                 Key: PHOENIX-3131
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3131
>             Project: Phoenix
>          Issue Type: Improvement
>    Affects Versions: 4.8.0
>            Reporter: tu nguyen khac
>            Priority: Critical
>
> I created a table in phoenix with query : ( 4 node , ram 8gb, 4 cores / node ) 
> CREATE TABLE pageview_site (
>     url varchar(255) not null,
>     pageview bigint,
>     dt date not null,
>     CONSTRAINT PK PRIMARY KEY (url, dt ROW_TIMESTAMP)
> ) SALT_BUCKETS = 4;
> After that : 
> 1. I tried to upsert about : 13 milions rows to this table . 
> 2. Run 2 queries : 
>     a. select url,sum(pageview) as pv FROM pageview_site where dt > to_date ('2016-06-01') group by url limit 100 offset 2;
> the duration this query  in about : 0.5 second
>     b. select url,sum(pageview) as pv FROM pageview_site where dt > to_date ('2016-06-01') group by ur order by pv descl limit 100 offset 2;
> the duration this query  in about : 9.5 seconds
> what happens with 2nd query ?? I think we should improve performance for "order by " command 



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