Towards a Tactic-Based Evaluation of Self-Adaptive Software Architecture Availability

nowadays, several non-automatic or semi-automatic software architecture evaluation methods have been proposed to evaluate their quality attributes as availability. In spite of their applicability, they are not effective in self-adaptive software architectures due to their off-line properties; e.g., scenario-based methods. Since the architectural tactics provide a bridge between architectural designs and quality attributes, they have sufficient potential to resolve this problem. In this paper, we assume that the software architecture is completely composed of some architectural patterns. Then we propose an automated evaluation method which composes the architectural tactics and the patterns to measure the availability of software architectures. In this method, the composition of a few availability tactics and patterns are simulated with appropriate probability distribution functions. To predict the availability of patterns, a data mining approach is applied to these simulated models to generate training models for each combination of tactics and patterns. Furthermore, a utility function is defined to compute the availability of systems by these models in O(n) where n is the number of patterns of systems. This method improves the data gathering and analysis activities of the SASSY (Self-Architecting Software SYstems) framework. To validate our method, we have applied it to the Rapidminer case study. KeywordsAvailability, Self-Adaptive Architecture, Architectural Tactic, Architectural Pattern, Data Mining.