Quantitative Verification of Non-functional Requirements with Uncertainty

We focus on non-functional requirements, such as those concerning reliability, performance, or cost and examine how to support the transition from requirements to design models that can be analyzed formally in quantitative terms. We assume that the initial description is given in behavioral terms, using annotated UML Sequence Diagrams. Annotations are used to express environmental assumptions, which are subject to uncertainty, in probabilistic terms. We also assume that a set of requirements is expressed via Structured English statements, which provide predefined patterns to support specification of common probabilistic properties. We discuss how sequence diagrams can be automatically translated into formal models that support software engineers in reasoning about the application being developed. In particular, requirements are transformed into appropriate logic statements while sequence diagrams are translated into Markov models, which can then be analyzed by using probabilistic model checking.