Today's screencast features Oded Noy, CTO of PATH Communications, who demonstrates the PATH Application Manager (P.A.M) -- a tool that instruments and analyzes the behavior of large Java (or C/C++) applications. The label most folks would attach to this product is application performance management, but Oded favors the term application behavior management. The demo begins with an explanation of how P.A.M. injects its instrumentation into your code, and then shows how to correlate observed problems ("the application is running slowly") with underlying causes ("somebody is calling 382 database queries per second").
As you'll see, the tool creates common ground between the operations team who will notice problems, and the development team who will solve them. In the emerging world of composite applications that rely on local or remote components and services, this cross-disciplinary style becomes increasingly vital.
Another interesting aspect of our discussion revolves around pattern recognition. In the realm of source code analysis (see our recent feature on the topic), the patterns of interest are textual and the analysis is syntactic. With P.A.M, patterns are temporal and analysis happens in the frequency domain. The technique, Oded explains, was inspired by the seismological background of PATH Communications' directory of technology, Jason McBride.
Here's the screencast (30 min, 18MB). I've also included just the audio track as an RSS 2.0 enclosure. In this case, the ratio of discussion to screen activity is fairly high, so the audio alone might be of use to the iPodding drive-time crowd.
Former URL: http://weblog.infoworld.com/udell/2004/12/16.html#a1135