[
https://issues.apache.org/jira/browse/HBASE-7805?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13575501#comment-13575501
]
Ted Yu commented on HBASE-7805:
-------------------------------
If 0.94 tests output both serial and parallel runtime statistics upon failure,
that would help us gauge performance characteristics in various scenarios.
By raising amplifier to 7/4 I could get the test to fail:
{code}
Failed tests:
testParallelScanAgainstSameRegionWithCoProc(org.apache.hadoop.hbase.client.TestParallelScan):
serial 305, parallel 186
testParallelScanAgainstSameRegion(org.apache.hadoop.hbase.client.TestParallelScan):
serial 223, parallel 133
{code}
Good news is that parallel scanning is faster than serial.
But as James found out, the presence of region observer does slow down the
overall performance.
> Scan of same region processed serially if table has region observer
> coprocessor
> -------------------------------------------------------------------------------
>
> Key: HBASE-7805
> URL: https://issues.apache.org/jira/browse/HBASE-7805
> Project: HBase
> Issue Type: Bug
> Components: Coprocessors, Scanners
> Affects Versions: 0.94.4
> Reporter: James Taylor
> Fix For: 0.96.0, 0.94.6
>
> Attachments: 7805-0.94.patch, 7805-trunk.txt, 7805-trunk-v2.txt,
> ParallelScanTestCoProc.java, ParallelScanTest.java
>
>
> If multiple scans to different parts of the same region are executed in
> parallel, they are processed serially if the table has a region observer
> coprocessor.
--
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
