Hi Vincent, Following your tests, Matthieu made some changes/optimizations in the OpenJPA persistence to avoid some table scans. So you should see performance improvements in your test when using the latest 1.1-branch code.
Depending on your requirements, you may also consider tweaking event generation ( http://ode.apache.org/user-guide.html#UserGuide-ODEExecutionEvents) to reduce the amount of data written to the database. alex On 12/3/07, Vincent <[EMAIL PROTECTED]> wrote: > > Re Alex, Matthieu, > > 1) Thanks for having made me remind I forgot to specify the hardware I > used ;o) > I am working on a PC workstation - Pentium IV 2.8 Ghz - 2 GB RAM - > Seagate Barracuda 7200rpm / 8.5ms - Win XP SP2 Pro. > While running the tests, only WLS and MySQL were running, plus 2 or 3 > other softwares such as MySQL monitor or Windows Explorer. > So I do think that neither the hardware not the software can be blamed > for the results. > > 2) Except the changes I listed in my previous email, I did not modify > anything to the ODE persistency mechanism. So I do think that OpenJPA > was used. > I attached to this email a zip file (mysql_queries.zip) containing the > db traces collected while deploying and running the BPEL process, > either in "in-memory" mode or "persisting" one. > You'll notice that the request as well as the answer are persisted, > but also a lot of events (I think that's what Matthieu was "looking > for") / serialized java objects. > NB: I can not publicly disclose the BPEL process that was used, sorry. > > =============== > > As requested, I opened a JIRA issue to post the WLSFactory, see > http://issues.apache.org/jira/browse/ODE-216. > > Regards, > > Vincent > >
