Architecture knowledge management during system evolution: observations from practitioners

It is widely accepted that awareness of architectural decisions enables better management and planning of system evolution, refactoring, and modernization efforts. In this paper we report data from interviews with software architects about how practitioners utilize architecture during system evolution. Our results show, despite the widely shared view that long-lived systems are better off with strong architectures; basic architecture-centric practices are not followed systematically. The key gap we observe is in correct and timely communication of architectural issues. This overall finding is not surprising. However, our data also contributes to how architecture knowledge management activities can be focused for most benefit throughout a system's lifespan. While the often-referenced problem is lack of time spent on documentation and design practices, our interviews show that lack of quality attribute reasoning early on, and during the lifespan of the system is a key contributor to failing to use architecture knowledge effectively during evolution.