Tailoring an architectural middleware platform to a heterogeneous embedded environment

Embedded systems are rapidly growing in size, complexity, distribution, and heterogeneity. As a result, the traditional practice of developing one-off embedded applications that are often rigid and unmanageable is no longer acceptable. Recent studies have suggested that an effective approach to developing software systems in this domain is to employ the principles of software architecture. However, for software architectural concepts to be truly useful in a development setting, they must be accompanied by support for their implementation and evolution. This has motivated our work over the past several years on an architectural middleware, called Prism-MW, that provides implementation-level support for the development of software systems in terms of the software architectural constructs (e.g., components, connectors). Prism-MW was initially developed in Java and used in several domains. Recently, as part of an on-going project, we were required to implement Prism-MW in ANSI C++. This experience proved to be more challenging than we initially anticipated, mainly due to the inherent heterogeneity of the computing substrate. As a result of this experience, we had to reconsider some of our earlier assumptions of what constitutes an architectural middleware and its role in the software development process. In this paper, we provide an overview of our experience and the lessons we have learned along the way.