Hello there,
does anyone have experience in profiling SCA applications?
I have tried profiling my application using YourKit, a commercial
profiler that integrates nicely with Eclipse. My first attempt has failed,
however, in that I get profiling data only for the Tuscany classes, but
not for my application components.
The way I run my application, I start several Java processes (in profiling
mode) using the node API, somewhat like this:
SCANodeFactory factory = SCANodeFactory.newInstance();
SCANode node = factory.createSCANode( compositeURI, scaContribs );
node.start();
Services expose Tuscany RMI bindings. I use hand-coded RMI clients
to exercise the services, either from a web frontend or from test classes.
This code would not be running under the profiler:
MyService service = (MyService) Naming.lookup( “myServiceRmiName” );
service.doSomething();
Do you think this approach is basically alright? Should I rather talk about
setting
filetering/sampling options with the YourKit people, or is there something about
the Tuscany runtime that makes profiling SCA applications inherently difficult?
n Sebastian
