[jira] [Commented] (PHOENIX-3591) Group by is slower for immutable encoded table

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

    [ https://issues.apache.org/jira/browse/PHOENIX-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15824490#comment-15824490 ] 

Mujtaba Chohan commented on PHOENIX-3591:
-----------------------------------------

Few more details on this. This is with 5K dense columns and using column names that would be the first and last columns for both immutable encoded and non-encoded scheme. Generated 20K rows so data fits in OS page cache when executing query multiple times.

Query 
{noformat}
select /*+ SERIAL NO_CACHE*/ ZZZLAST,A00FIRST,sum(k1) from $T group by ZZZLAST,A00FIRST
{noformat}

Query time for immutable encoded is 13.2sec vs 4.0 seconds for immutable non-encoded.

> Group by is slower for immutable encoded table
> ----------------------------------------------
>
>                 Key: PHOENIX-3591
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3591
>             Project: Phoenix
>          Issue Type: Sub-task
>            Reporter: Mujtaba Chohan
>            Assignee: Samarth Jain
>
> {noformat}
>  select /*+ SERIAL NO_CACHE*/ c0,sum(k1) from $T group by c0
> {noformat}
> This is >2X slower when compared to non-encoded table. There are 100 distinct values in c0 column.
> For data and schema see data generator script in https://issues.apache.org/jira/browse/PHOENIX-3560



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