Hello Walter, Boundary blame itself takes out (from the Investigator view) intermediate components in a blame stack.
Basically all you get is the frontend and backend -- the "boundaries" of the JVM components.
This has resulted in a sizeable decrease in EM metric load as we used to have in previous Introscope 5.x,6.x releases, where all blame stack was collected affecting EM and Agent performance.
Please note that this can be switched off, but you lose some nicer naming of components if you switch this feature off.
Boundary blame works like this:There is a new tracer called a FrontendTracer that can be used to explicitly mark a frontend. If no FrontendTracer is configured, the first component in the blame stack will be the default frontend (this may not be desireable -- for example, in a BRTA environment you may want to make servlets show up as frontends instead of the browser component. If so, the FrontendTracer is for you!). There is also a BackendTracer to explicitly mark backends (databases, for instance). A default backend can be inferred in the absence of a backend tracer -- basically, anything that opens a client socket will be a default backend if none is explicitly marked.Finally, there is a BlamePointTracer (of which the FrontendTracer and BackendTracer are instances) that gives canonicalized metrics for any blamed component (avg. response time, per interval counts, concurrency, stalls, errors). A BlamePointTracer would go on a middle component. Using it instead of the old individual tracers in previous v5,6 simplifies configuring the common and more required metrics quickly. For more information, please refer to:- JavaAgent8.2.0.0.pdf, page 94, 134 - see attachment.- KB: 138 - Creating Custom Directives (Tracers) :
https://support.wilytech.com/cgi-bin/wilytech.cfg/php/enduser/std_adp.php?p_faqid=138 I hope this helps, Regards, Sergio