Using planning techniques to assist quality-driven architectural design exploration

A software architecture design captures the main decisions regarding the quality-attribute requirements for a system. When constructing these designs, the architect normally evaluates and combines solutions with different quality-attribute tradeoffs. This exploration of the design space can be seen as a searching problem, in which the architect's expertise is what directs the search towards a "good-enough" solution. Nonetheless, given complexity of this search, assisting the architect with adequate tool support becomes indispensable. In this context, we have investigated the utility of a planning-based tool approach called DesignBots to explore design alternatives. Specifically, the approach considers quality-attribute issues as goals achievable by combinations of architectural mechanisms, which are generated by a mixed-initiative and hierarchical planning engine. Our experiences with DesignBots show that the planning approach effectively helps architects to explore design alternatives productively.