[
https://issues.apache.org/jira/browse/HBASE-15082?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15111703#comment-15111703
]
stack commented on HBASE-15082:
-------------------------------
A rough measure of this patch on small cluster of with patch and without doing
increments with the bundled IncrementPerformanceTest tool shows about the same
speedup as with and without for the branch-1 patch.
Without patch:
{code}
2016-01-21 16:20:08,602 INFO [main] hbase.IncrementPerformanceTest:
75th=1.3486201E7, 95th=1.7464251E7, 99th=2.0276807E7
real 1m48.443s
user 1m56.227s
sys 0m58.596s
{code{
With the patch:
{code}
2016-01-21 16:40:26,667 INFO [main] hbase.IncrementPerformanceTest:
75th=9020464.0, 95th=1.2099562E7, 99th=1.9464278E7
real 1m15.537s
user 1m57.670s
sys 0m55.586s
{code}
.What is interesting is that this patch DOES NOT do what the branch-1 patch
does to get speed, by-passing mvcc, where we READ_UNCOMMITTED, and then on the
end of the increment, we just do a complete of the mvcc rather than a
completeAndWait.
So, it is interesting that it is faster. Let me try and see why and see if I
can go faster still by taking on the branch-1 trick.... which I should be able
to do AND keep row consistency since all mvcc messing and memstore changes are
under exclusive row lock.
> Fix merge of MVCC and SequenceID performance regression
> -------------------------------------------------------
>
> Key: HBASE-15082
> URL: https://issues.apache.org/jira/browse/HBASE-15082
> Project: HBase
> Issue Type: Sub-task
> Components: Performance
> Reporter: stack
> Assignee: stack
> Priority: Critical
> Fix For: 2.0.0, 1.3.0
>
> Attachments: 15082.patch, 15082v10.patch, 15082v12.patch,
> 15082v13.patch, 15082v14.patch, 15082v14.patch, 15082v15.patch,
> 15082v16.patch, 15082v2.patch, 15082v2.txt, 15082v3.txt, 15082v4.patch,
> 15082v5.patch, 15082v6.patch, 15082v7.patch, 15082v8.patch
>
>
> This is general fix for increments (appends, checkAnd*) perf-regression
> identified in the parent issue. HBASE-15031 has a narrow fix for branch-1.1
> and branch-1.0.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
