[
https://issues.apache.org/jira/browse/OPENJPA-160?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12476039
]
Patrick Linskey commented on OPENJPA-160:
-----------------------------------------
The profiling data seems to point to reflection as the culprit; the init() call
is very very fast.
Could someone with access to the benchmark try out just calling 'new
BrokerImpl() from JDBCConfigurationImpl or something? That should tell us a lot
about where the cost is coming from.
WRT per-thread pooling -- it sounds like you're proposing that we would
actually share brokers between multiple EMs in the same thread. I think that
that would have a lot of undesired consequences, would certainly violate a
bunch of the intent of the JPA spec, and would probably fail in the CTS. In
OpenJPA, each logical EM definitely needs to have access to a unique Broker.
Pooling could help us reduce the cost of obtaining such a Broker, but sharing
would be a pretty significant semantic change.
> Reuse BrokerImpl objects
> ------------------------
>
> Key: OPENJPA-160
> URL: https://issues.apache.org/jira/browse/OPENJPA-160
> Project: OpenJPA
> Issue Type: Sub-task
> Reporter: Michael Dick
> Attachments: perf2.jpg, perf3.jpg
>
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
