Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Real-time tracking of evolving software architectures

: Rost, Dominik
: Knauber, Peter; Knodel, Jens

Mannheim, 2007, IX, 114 S.
Mannheim, Hochschule für Technik und Gestaltung, Dipl.-Arb., 2007
Fraunhofer IESE ()
architecture compliance checking; Eclipse; SAVE; software architecture; conflict detection; evolution monitor; real-time computation

Practical experience shows that the architecture of a software system degenerates during its evolution. A key reason for this effect is often an insuffcient awareness of the specifications, decomposition rules and constraints, defined in the target architecture of a software system. This creates the need for an improvement of architecture-centric development of software products.
This thesis contributes to this goal by providing the concepts and a tool to support developers and architects in distributed teams during the realization of a software project. This tool, the evolution monitor, is a plug-in for the Eclipse SDK on the basis of SAVE. It supports the implementation process by enabling the developers to perform architectural analyses in real-time and presenting the results integrated in the IDE. Analyses of three different categories are supported: context, impact and conflict, whereas context refers to the architectural context of elements, impact, to the architectural impact of modifications and conflict, to the identification of possibly conflicting changes implemented by different developers. These computations can be performed on-demand and in real-time during the development of the system by developers and the architect, and thus, providing instant feedback about the impact of implemented changes to the architecture of the system.
To validate the idea of the evolution monitor and its concept, a controlled experiment and a case study have been conducted. The experiment showed an increased architecture compliance in components, that were realized by teams supported by the evolution monitor. Additionally the experiment demonstrated a stable monitoring for a period of 4 weeks and the case study showed that the evolution monitor scaled up to a medium-size system.