Managing trade-offs among architectural tactics using feature models and feature-solution graphs

Abstract

In Software Architecture, several approaches focus first on selecting architectural tactics to address quality attributes and later use the corresponding design alternatives to create application architectures. Regrettably, sometimes the alternatives used to improve some quality attributes inhibit or reduce the achievement of others. These conflicts, known as trade-off points, lead to trade-off decisions to solve them. Usually, detection of trade-off points and guidance for trade-off decisions rely on the expertise of software architects. The architect has to (1) identify and analyze the conflict on design alternatives, (2) determine which quality attributes and tactics motivated the selection of conflicting alternatives, and (3) decide about which set of non-conflicting tactics to use. This paper presents an approach based on feature models to help manage trade-offs. It is based on the specification of relationships between architectural tactics and design alternatives that describe, for each tactic, which combination of designs can be used or must not be used. When a set of tactics are selected to implement an architecture, these relationships serve to determine the set of alternatives to use. In addition, they aid to detect trade-off points and determine the tactics to consider in a trade-off decision. We present a formalization of our approach and illustrate it using a case study on tactics and patterns for database-based web applications.