Software Engineering Institute

Architecture evaluation has been well established in industrial practice for over a decade, but one assumption that all evaluation methods have made is that there is an architecture to evaluate. In ultra-large-scale (ULS) systems, the architecture may not yet exist for portions of the system or there may be competing architectures for the same functionality in different portions of the system.

We describe an application of the principles underlying traditional architectural evaluation methods to the problem of analyzing an architecture landscape: a broad set of architectural decisions representing a spectrum of potential architectures. Why would we want to analyze an architecture landscape, rather than a concrete architecture? We are motivated to do this when there are many architectural decisions to be made, many stakeholders, and many systems to be built, and when the architectural decisions are non-trivial with far-reaching consequences. Smart Grid demand-response systems are examples of such landscapes.

This presentation was given at SATURN 2011 in Burlingame, CA.