[
https://issues.apache.org/jira/browse/GORA-56?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13141046#comment-13141046
]
Ferdy commented on GORA-56:
---------------------------
I ran some benchmarks for the patch in single thread mode and I could not
notice a performance impact compared to the current implementation. This is to
be expected:
-The ThreadLocal HTable code should not incur a performance penalty. It is a
recommended way to avoid synchronization altogether.
-The fields in HBaseStore are made volatile, however since there are just
written once and read only afterwards the penalty for this is truly minimal.
Removing the volatile keywords will not make a difference in practice, but
technically they are required since the initialization thread must publish its
objects safely to other threads somehow. (However I probably could have gotten
away with using just one volatile variable as a memory-barrier.) Anyway I
noticed that some of the other store implementations (including superclass
DataStoreBase) do not do any of this, so I guess this is something to keep in
mind.
> HBaseStore is not thread safe.
> ------------------------------
>
> Key: GORA-56
> URL: https://issues.apache.org/jira/browse/GORA-56
> Project: Gora
> Issue Type: Bug
> Components: storage-hbase
> Reporter: Ferdy
> Priority: Critical
> Attachments: GORA-56-v2.patch, GORA-56.patch
>
>
> Will be working on a patch.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
