|> I don't think that JUnit can handle speed measurements the |> way we want, as it only does pass/fail. |> But it could be possible to write such a system. |> We should probably have a baseline speed test not depending |> on JBoss that never changes and use that to correct for |> differences in speed of the machine used for testing.
we don't need something very complex, even just the simple test from the tomcat installation are enough to measure theoretical throughput on the bare container. I would plug in fake persistence engines so that we are really looking at **container** numbers and not **serialization/db** number. Then there needs to be a series of tests for the persistence engines. While the time on the container has been fairly stable the time on the cmp engines can vary greatly. Since 80% of the time is spent in these stacks (queries with large sets etc) it really impacts the performance a lot. These tests should substract the measured time on container stacks and just measure the CMP time. marcf _______________________________________________ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development
