Probabilistic Macro-Architectural Decision Framework

Experience with system-level concerns demonstrates that fitness for context is a consideration that is equally significant in making architectural decisions as is fitness for purpose. This requires architects to consider contextual factors in making decisions. These decisions are probabilistic in nature and they represent the subjective belief of the architect or the prior probability which is likely to change as new evidence becomes available during the course of the system design. They serve as recommendations and directional inputs to other decisions in the design process. In this paper, we introduce a macro-architectural decision framework we developed to enable the architect for a software-reliant system to model and reason about contextual factors. At the core of our framework is an adaptation of a Bayesian belief network that is augmented with decision and utility nodes. The framework captures contextual factors and their influence on decisions and utilities. We applied our approach in the study of a software system implementation at a healthcare company. The results show promise that such decision support tools help explore the space of factors involved in decision making and provide sensible suggestions for making architectural decisions.