Modelling the Asynchronous Dynamic Evolution of Architectural Types

Self-adaptability is a feature that has been proposed to deal with the increasing management and maintenance efforts required by large software systems. However this feature is not enough to deal with the longevity usually these systems exhibit. Although self-adaptive systems allow the adaptation or reorganization of the system structure, they generally do not allow introducing unforeseen changes at runtime. This issue is tackled by dynamic evolution. However, its support in distributed contexts, like self-organizing systems, is challenging: these systems have a degree of autonomy which requires asynchronous management. This paper proposes the use of asynchronous dynamic evolution, where both types and instances evolve dynamically at different rates, while preserving: (i) type-conformance of instances, and (ii) the order of type evolutions. This paper describes the semantics for supporting the asynchronous evolution of architectural types (ie. types that define a software architecture). The semantics is illustrated with PRISMA architecture specifications and is formalized by using typed graph transformations.