Performance measurement of models specified through component-based software architectural styles

Abstract

Designing scalable and executable architectural models has a particular importance in software system development, because it allows evaluating different functional and non-functional properties of a designed system at a high level of abstraction. Measuring the performance related properties at the architectural level and before implementation, is very important while designing complex software systems. In this paper, using component-based software architectural styles specified through graph transformation systems, we propose an approach to performance measurement and analysis of software architectural models. To do so, we enrich the style with performance and timing information. Then, we propose the necessary algorithms for transforming architectural models into PEPA (Performance Evaluation Process Algebra). PEPA provides the measurement of performance metrics such as utilization of different software components and throughput. This paper proposes an efficient algorithm for model transformation. The proposed approach solves the state space explosion problem which was occurred in our previous work by generating PEPA models from the host graph instead of state space. The accuracy of this approach and other concepts are illustrated through ‘e-map router’ case study.