Teaching Software Architecture in Industrial and Academic Contexts: Similarities and Differences

Introduction

What is the role of the software architecture inside a large mission-critical system? How is it created? How is it managed? How can the concept foster software reuse and productivity? These questions are quite relevant for engineering companies, which produce families of software intensive systems (Buschmann, 1996). As software systems become larger, more complex, and more expensive, companies - in particular system integrators - feel an increasing need for improving their productivity exploiting sound and effective techniques for the definition, analysis, and evaluation of software architectures. This is what we observed in a number of cooperations between academia and industry, and that motivated our study of how Software Architecture can be taught.

The initial question from which we started our study was: “how can one introduce and teach software asset reuse and software architecture evaluation to engineers who have been designing systems for years without explicitly dealing with these concepts?” Then we added a related question: “how can our experience in teaching Software Architecture in an industrial context be imported in an academic context of Computer Science students?”

When the field of Software Architecture emerged, it was argued it should be considered a discipline in its own, separate from Computer Science and possibly encompassing Software Engineering (Clements, 2010). After almost twenty years the corpus of the scientific works in the field has developed consistently, but there is still a large gap between this body of knowledge and what is actually needed in academic and industrial settings. Many of the achievements in the field have not matured enough; an example are Architecture Definition Languages (ADLs), that have not replaced - and are not likely anymore to replace - standard modeling languages (Clements, 2012). Some others achievements are more mature, for instance some tools for architectural analysis: in (Bernardo, 2001) is described a tool for performance evaluation of a software architecture, whereas in (Sterling, 1996) is described a tool for architecture animations. However, these tools still need to be tailored to specific software systems, and even more to internal industrial practices.

The increase of the size and complexity of contemporary software-intensive systems raises critical challenges to the engineering companies which build them to be integrated into larger systems of systems. Production and management problems with software intensive systems are well known and related to requirements engineering, software design, systems’ families management, and their continuous testing integration and evolution.

Thus, teaching software architectures in an industrial context requires to meet a company’s expectations in terms of mature knowledge, special competences, and best practices transferred to practitioners, that they can subsequently turn into the engineering life cycle of complex systems.

This is not easy to achieve, as architecting large-scale complex software systems - having tens of thousands of requirements and millions of lines of code - requires very high abstraction and modeling skills. A number of methods and solutions to these problems are based on the introduction of software architecting in the industrial practice (Bass, 2012). However, to become effective an architecture-centric development approach needs to be assimilated and put in everyday practice by the company personnel, who need architectural education and training.