Reasoning about software architectures with contractually specified components

One of the motivations for specifying software architectures explicitly is the better prediction of system quality attributes. In this chapter we present an approach for determining the reliability of component- based software architectures. Our method is based on RADL (Rich Architecture Definition Language), an extension of DARWIN [16]. RADL places special emphasis on component interoperation and, in particular, on accounting for the effects of interoperation on system reliability. To achieve this, our methods use a notion of design-by-contract [19] for components, called parameterized contracts [26]. Our contracts involve finite state machines that allow software architects to define how a component's reliability will react to a deployment environment. We show how a system, built from contractually specified components, can be understood in terms of Markov models, facilitating system reliability analysis. We illustrate our approach with an e-commerce example and report about empirical measurements which confirm our analytical reliability prediction by means of monitoring in our reliability testbed.