Hi Thomas, I think the throughput variations are because of the overall number of events (events processed:), the lower the number the lower the throughput, this being visible for relatively small numbers (the upper limit for throughput is several thousand events/sec, depending on how fast is your system). The filter set, collection mode (also JVMTI vs JVMPI), and application behavior affects this metric.
You could see more consistent results (from an event load perspective at least) if you profile to file (and get several thousand or millions events), then import the files with a filter set that should give you similar results. If you profile to file, you should also get better run time for the profiled application, closer to regular run times (when not profiled), this is good if your application is sensitive to execution time. If you have a filter that filters out most of the events when profiling one application and loads most of the events when profiling another application, you could still see big differences. By using the right filters you could reduce considerable the profiling overhead. Thank you, Marius Slavescu IBM Tivoli Autonomic Computing - Toronto Lab Phone: 905-413-3610 Thomas Eichstädt-Engelen <[EMAIL PROTECTED]> Sent by: [EMAIL PROTECTED] 02/10/2008 07:13 AM Please respond to [EMAIL PROTECTED]; Please respond to TPTP Tracing and Profiling Tools Project developer discussions <tptp-tracing-profiling-tools-dev@eclipse.org> To Marius Slavescu/Toronto/[EMAIL PROTECTED] cc TPTP Tracing and Profiling Tools Project developer discussions <tptp-tracing-profiling-tools-dev@eclipse.org> Subject Re: [tptp-tracing-profiling-tools-dev] DataThrougput while profiling JUnit-Tests Hi Marius, Am Mo, 4.02.2008, 22:55, schrieb Marius Slavescu: > Right, it measures only the time sent in processing the events from first > event to last event (or dataServerExit call) in one session (for each > XMLTraceDataProcessor instance). thanks for your quick response! Regarding my data throughput question. Could you think of a reason why it is so volatile? Cheers, Thomas E.-E. _______________________________________________ tptp-tracing-profiling-tools-dev mailing list tptp-tracing-profiling-tools-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/tptp-tracing-profiling-tools-dev
_______________________________________________ tptp-tracing-profiling-tools-dev mailing list tptp-tracing-profiling-tools-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/tptp-tracing-profiling-tools-dev
