Using real options to select stable middleware-induced software architectures

The requirements that force decisions towards building distributed system architectures are usually of non-functional nature. Scalability, openness, heterogeneity, and fault-tolerance are examples of such non-functional requirements. The current trend is to build distributed systems with middleware, which provide the application developer with primitives for managing the complexity of distribution, system resources, and for realizing many of the non-functional requirements. As non-functional requirements evolve, the “coupling” between middleware and architecture becomes the focal point for understanding the stability of the distributed software system architecture in the face of change. We hypothesise that the choice of a stable distributed software architecture depends on the choice of the underlying middleware and its flexibility in responding to future changes in non-functional requirements. We devise an option-based model to value such flexibility and guide the selection. We empirically evaluate the model using a case study that adequately represents a medium-size component-based distributed architecture. We report on how a likely future change in scalability could impact the architectural structure of two versions, each induced with a distinct middleware: one with CORBA and the other with J2EE. Our hypothesis is verified to be true for the given change. We conclude with some observations that could stimulate future research in the area of relating requirements to software architectures.