Variability Mining of Technical Architectures

Technical architectures (TAs) represent the computing infrastructure of a company with all its hardware and software components. Over the course of time, the number of TAs grows with the companies' requirements and usually a large variety of TAs has to be maintained. Core challenge is the missing information on relations between the existing variants of TAs, which complicates reuse of solutions across systems. However, identifying these relations is an expensive task as architects have to manually analyze each TA individually. Restructuring the existing TAs poses severe risks as often sufficient information is not available (e.g., due to time constraints). To avoid failures in productive systems and resulting loss of profit, companies continue to create new solutions without restructuring existing ones. This increased variability in TAs represents technical debt. In this paper, we adapt the idea of variability mining from the software product line domain and present an efficient and automatic mining algorithm to identify the common and varying parts of TAs by analyzing a potentially arbitrary number of TAs in parallel. Using the identified variability information, architects are capable of analyzing the relations of TAs, identifying reuse potential, and making well-founded maintenance decisions. We show the feasibility and scalability of our approach by applying it to a real-world industrial case study with large sets of TAs.