Automatic evaluation of complex design decisions in component-based software architectures
Max Scheerer, Axel Busch, Anne Koziolek · 2017
The quality of modern industrial plants depends on the quality of the hardware used, as well as software. While the impact on quality is comparably well understood by making decisions about the choice of hardware components, this is less true for the decisions on software components. The quality of the resulting software system is strongly influenced by its software architecture. Especially in early project phases a software architect has to make many design decisions. Each design decision highly influences the software architecture and thus, the resulting software quality. However, the impact on the resulting quality of architecture design decisions is hard to estimate in advance. For instance, a software architect could decide to deploy software components on a dedicated server in order to improve the system performance. However, such a decision may increase the network overhead as side-effect. Model-driven approaches have been shown as promising techniques enabling design-time quality prediction for different quality attributes such as performance or reliability. However, such approaches are limited in their automated decision support to simple design decisions like the exchange of one single component. In this paper, we present an approach that automatically evaluates complex design decisions in software architecture models. Such design decisions require the reuse of subsystems with many involved components coming with inhomogeneous architectures. We evaluate our approach using a real-world example system demonstrating the benefits of our approach.