Process support for the transition from software requirements to software architectures

Abstract

The quality of software products heavily depends on requirements engmeenng and software architecting. Both phases are the foundation for subsequent activities, such as project planning, implementation or maintenance. In recent years, significant progre has been made in requirements engineering and software architecture research and practice, but mainly in isolation. Currently, the transition from requirements to architectures is performed in an ad-hoc manner and greatly relies on experience and skills of the individuals performing the transition. Thus, the research put forward in this thesis outlines TRAM (Transition between Requirements and Architectures using a Modularized Framework), a method that aims at facilitating the transition from requirements to architectures. TRAM consists of five modules, starts from a generic requirements engineering process and creates different architectural perspectives. TRAM is distinguished from other approaches by it integration with the requirements engineering process, its consideration of developer and customer needs and by its ad-hoc assessment of architecture quality. To achieve this, several approaches are developed and used within TRAM: RESTAR structures requirements from an architecture perspective using a clustering technique; SYSAS supports the selection of software architecture styles; SQUAD supports ad-hoe architecture assessment. All three approaches can also be used independent of TRAM. To validate TRAM, we apply it to a case study and also collect expert opinion . The results show the potential advantage and usefulness of TRAM. Also, we perform a theoretical evaluation by checking how TRAM compares to previous methodologies that claim to address the requirements-architecture gap. In order to demonstrate the feasibility of tool support for TRAM, the thesis proposes TRAM-PT (TRAM-Prototype Tool). This prototype tool has been used when conducting the case study. The thesis concludes with a discussion of future work to improve TRAM, how TRAM fits into other software engineering paradigms (such as agile development), and a presentation of potential research directions for further improving the requirements architecture transition.