Reconstruction of J2EE-based system architectures

[jbaragry]

G'Day,

I'm trying to apply some of the recent research techniques of software architecture recovery to J2EE technology but I need some help to make sure its as useful as possible.

The plan is to first represent systems using the JSR 26 spec (UML/EJB Mapping Specification) plus relevant aspects of the UML profiles for EDOC and EAI. Then to use different techniques to make the diagrams more readable. That is, turn a diagram with potentially hundreds of classes on it into something that shows a lot fewer architecture components, patterns, etc - which hopefully will be easier to read/understand/use.

The main problem is what to show when you go from low level details to higher level architecture. So, would you find such an architecture recovery tool useful? and What would you use such a tool for (or what would you like to see in the architecture representation of a J2EE-based system)? thanks for any help.

Jason
---
Jason Baragry
Distributed Systems and Software Engineering Research Norwegian Computing Center.
www.nr.no/~jason

[davidjencks]

I'm very curious as to how this research measures "useful". I've found sketching pictures useful to clarify my own thinking, and using a uml tool useful when I had to publish some pictures, but I have yet to find any uml or other modeling techniques very useful when the picture is larger than one sheet of paper.

Is there a well established measure of usefulness in this research community?

[davehawley]

Is the idea to infer the framework as well, or does the framework need to be modelled before analyzing the application architecture?

Considering the usefulness of framework-specific naming conventions and patterns for human understanding, the former problem might prove very hard.