hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Lars Hofhansl (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9428) Regex filters are at least an order of magnitude slower since 0.94.3
Date Wed, 04 Sep 2013 17:42:52 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13758016#comment-13758016
] 

Lars Hofhansl commented on HBASE-9428:
--------------------------------------

The difference after HBASE-7279 is that we pass the initial byte[] + offset around. Previous
we'd make a copy of the row key. So now the JDK make a copy of the initial byte[] during decoding
(i.e. the entire HFile block).
                
> Regex filters are at least an order of magnitude slower since 0.94.3
> --------------------------------------------------------------------
>
>                 Key: HBASE-9428
>                 URL: https://issues.apache.org/jira/browse/HBASE-9428
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Jean-Daniel Cryans
>             Fix For: 0.98.0, 0.94.12, 0.96.1
>
>         Attachments: 9428-0.94.txt
>
>
> I found this issue after debugging a performance problem on an OpenTSDB cluster, it was
basically unusable after an upgrade from 0.94.2 to 0.94.6. It was caused by HBASE-7279 (ping
[~lhofhansl]).
> The easiest way to see it is to run a simple 1 client PE:
> {noformat}
> $ ./bin/hbase org.apache.hadoop.hbase.PerformanceEvaluation sequentialWrite 1
> {noformat}
> Then in the shell do a filter scan (flush the table first and make sure if fits in your
blockcache if you want stable numbers).
> Pre HBASE-7279:
> {noformat}
> hbase(main):028:0> scan 'TestTable', {FILTER => "(RowFilter (=, 'regexstring:0000055872')
)"}
> ROW                                                 COLUMN+CELL                     
                                                                                         
                         
>  0000055872                                         column=info:data, timestamp=1378248850191,
value=(blanked)                                                                          
                                                         
> 1 row(s) in 1.2780 seconds
> {noformat}
> Post HBASE-7279
> {noformat}
> hbase(main):037:0* scan 'TestTable', {FILTER => "(RowFilter (=, 'regexstring:0000055872')
)"}
> ROW                                                 COLUMN+CELL                     
                                                                                         
                         
>  0000055872                                         column=info:data, timestamp=1378248850191,
value=(blanked)                                                                          
                                                           
> 1 row(s) in 24.2940 seconds
> {noformat}
> I tried a bunch of 0.94, up to 0.94.11, and the tip of 0.96. They are all slow like this.
> It seems that since that jira went in we do a lot more row matching, and running the
regex gets super expensive.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message