[
https://issues.apache.org/jira/browse/HBASE-8000?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13609807#comment-13609807
]
Sergey Shelukhin commented on HBASE-8000:
-----------------------------------------
Preliminary results:
1) I found some efficiency of selection issues in stripe policy, updated the
respective patch... a few tiny bugs too, mostly in HStore using the thing.
2) Haven't tested after fixing (1), but even before I can see the observed
pattern that stripe compaction is slightly slower on average with large data
due to constant small compactions, but doesn't have huge latency spikes that
default policy creates. I'll see how it can be tweaked, and how much of the
overall slowdown will be fixed by fixing (1).
3) Now the sad part. This can only be evaluated on actual large data (unless
somebody knows some tricks to make Hadoop really slow or otherwise test in such
cases?). I have to run it for a few hours on EC2.
> create integration/perf tests for stripe compactions
> ----------------------------------------------------
>
> Key: HBASE-8000
> URL: https://issues.apache.org/jira/browse/HBASE-8000
> Project: HBase
> Issue Type: Sub-task
> Components: Compaction
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
>
> While writing tests I seem to be finding more errors with edge cases
> unrelated to what test actually tests compared to what is expected.
> Integration test will be needed... probably good for perf/compare too.
--
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
