From Domain Models to Architectures
January 1994 • White Paper
This whitepaper was presented at the Workshop on Software Architecture, USC Center for Software Engineering, Los Angeles, 1994, by Paul Clements.
Software Engineering Institute
A software system can be evaluated against criteria in two broad categories:
- functional and performance attributes: how well does the system, during execution, satisfy its behavioral, functional, and performance requirements? Does it provide the required results? Does it provide them in a timely enough manner? Are the results correct, or within specified accuracy and stability tolerances?
- non-functional attributes: how easy is the system to integrate, test, and modify? How expensive was it to develop?
These two categories are orthogonal; systems that unfailingly meet all of their requirements may or may not have been prohibitively expensive to develop, and may or may not be impossible to modify. Highly modifiable systems may or may not produce correct results.
Given a set of requirements for a system, the developer must choose an architecture that will allow the implementation of the system to proceed in a straightforward manner, producing a product that meets its functional and non-functional requirements. How is that done?