hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dan Washusen (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-2249) The PerformanceEvaluation read tests don't take the MemStore into account.
Date Wed, 24 Feb 2010 01:17:27 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12837560#action_12837560
] 

Dan Washusen commented on HBASE-2249:
-------------------------------------

Posted on HBASE-2248 based by Dave:
{quote}
@Dan: Took a read over the patch, though it seemed to be based in a different dir and didn't
want to apply nicely. From what I can see the ScanTest still does getScanner in testSetup
before the timer is begun. This may be fine, if the point of this test is to measure scan
performance per-row and not setup/teardown time. It just explains why the ScanTest doesn't
exhibit this issue. It does look like other tests, such as the RandomSeekScanTest and the
new RandomScanWithRangeTest do test setup/teardown time as part of each "testRow" and should
exhibit this issue, if run.
{quote}

Dave has a very good point, the ScanTest is a strange one.  It appears to be timing each call
to ResetScanner#next and not including the cost of setting up or closing the scanner.  IMO,
this should be changed but doing so would change any previous results...

> The PerformanceEvaluation read tests don't take the MemStore into account.
> --------------------------------------------------------------------------
>
>                 Key: HBASE-2249
>                 URL: https://issues.apache.org/jira/browse/HBASE-2249
>             Project: Hadoop HBase
>          Issue Type: Improvement
>    Affects Versions: 0.20.3
>            Reporter: Dan Washusen
>             Fix For: 0.20.4
>
>         Attachments: HBASE-2249.patch
>
>
> The write tests in the PerformanceEvaluation flush the MemStore after they complete,
as a result the read tests don't take the MemStore into account (because it's always empty).
 An optional flag could be provided when the performance evaluation starts that dictates if
the table should be flushed after a test completes.  This optional flag would allow changes
to the MemStore to be performance tested...

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message