A formally founded framework for dynamic software architectures

Abstract

Software architectures play a significant role in the development of software-intensive systems in order to allow satisfying both functional and non-functional requirements. In particular, dynamic software architectures have emerged to address characteristics of the contemporary systems that operate on dynamic environments and consequently subjected to changes at runtime. Architecture description languages (ADLs) are used to represent software architectures, producing models that can be used at design time and/or runtime. However, most existing ADLs have limitations in several facets: (i) they are focused on structural, topological aspects of the architecture; (ii) they do not provide an adequate support for representing behavioral aspects of the architecture; (iii) they do not allow describing advanced aspects regarding the dynamics of the architecture; (iv) they are limited with respect to the automated verification of architectural properties and constraints; and (v) they are disconnected from the implementation level, thus entailing inconsistencies between architecture and implementation. In order to tackle these problems, this thesis proposes formally founded framework for dynamic software architectures. Such a framework comprises: (i) π-ADL, a formal language for describing software architectures under both structural and behavioral viewpoints; (ii) the specification of programmed dynamic reconfiguration operations; (iii) the automated generation of source code from architecture descriptions; and (iv) an approach based on statistical model checking (SMC) to formally express and verify properties in dynamic software architectures. The main contributions brought by the proposed framework are fourfold. First, the π-ADL language was endowed with architectural-level primitives for describing programmed dynamic reconfigurations. Second, architecture descriptions in π- ADL are translated towards implementation source code in the Go programming language, thereby contributing to minimize architectural drifts. Third, a novel logic, called DynBLTL, is used to formally express properties in dynamic software architectures. Fourth, a toolchain relying on SMC was built to automate the verification of architectural properties while striving to reduce effort, computational resources, and time for performing such a task. In this work, two wireless sensor network-based systems are used to validate the framework elements.