Trust and Costs for Evolving Architectural Performance Models: A Survey : Technical report

Architectural performance models allow assessing a system's performance during different development phases by predicting performance characteristics. At the same time, by representing a software's architecture, they can support software developers in various scenarios. Due to time constraints and frequent changes, their application is hindered. Thus, researchers proposed approaches, which continuously update architectural performance models during the development. However, we do not know how software professionals perceive the trust and costs in such approaches and if they address current challenges in performance management. Therefore, we conducted a survey among software professionals to investigate these topics.

Based on 110 responses, our findings indicate that performance prediction approaches are less used and gain less trust than monitoring tools or performance tests. We also found out that certain properties can improve the trust. Additionally, the results suggest that software professionals consider different cost factors and are willing to spend some time when adopting a new tool.