Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
Explicitly considering the software architecture supports efficient assessments of quality attributes. In particular, Architecture-based Performance Prediction (AbPP) supports performance assessment for future scenarios (e.g., alternative workload, design, deployment, etc.) without expensive measurements for all such alternatives. However, accurate AbPP requires an up-to-date architectural Performance Model (aPM) that is parameterized over factors impacting performance like input data characteristics. Especially in agile development, keeping such a parametric aPM consistent with software artifacts is challenging due to frequent evolutionary, adaptive and usage-related changes. The shortcoming of existing approaches is the scope of consistency maintenance since they do not address the impact of all aforementioned changes. Besides, extracting aPM by static and/or dynamic analysis after each impacting change would cause unnecessary monitoring overhead and may overwrite previous manual adjustments. In this article, we present our Continuous Integration of architectural Performance Model (CIPM) approach, which automatically updates the parametric aPM after each evolutionary, adaptive or usage change. To reduce the monitoring overhead, CIPM calibrates just the affected performance parameters (e.g., resource demand), using adaptive monitoring. Moreover, CIPM proposes a self-validation process that validates the accuracy, manages the monitoring and recalibrates the inaccurate parts. As a result, CIPM will automatically keep the aPM up-to-date throughout the development time and operation time, which enables AbPP for a proactive identification of upcoming performance problems and evaluating alternatives at low costs. CIPM is evaluated using three case studies, considering (1) the accuracy of the updated aPMs and associated AbPP and (2) the applicability of CIPM in terms of the scalability and the required monitoring overhead. The findings confirmed the accuracy of updated aPMs and the applicability for realistic cases.