An adaptive dependability model of component-based software

Reviewer: Cristiano di Flora

When designing complex and heterogeneous systems, it is fundamental to consider dependability issues from the early phases of the system lifecycle. When dealing with component-based systems, where heterogeneity may increase dramatically, this issue becomes crucial. A design-based approach to dependability may allow for the capture of diverse dependability requirements, and even for the production of an effective failure model for the involved components. The approach proposed by Lin and Youngsen in this paper supports both of these advantages, by taking into account how dependability requirements vary from one context to another. Using Bayesian belief networks (BBNs) to model the varying conditions of the operational environment, system requirements, component models, and components is quite an interesting idea, even though the advantages of adopting BBNs instead of other modeling techniques are not sufficiently clarified here. The effectiveness of this work could probably have been leveraged by briefly comparing the proposed approach to other approaches mentioned in the literature, namely, the approaches based on a combination of stochastic activity networks and unified modeling language (UML). Regardless of the paper's shortcomings, the approach is really interesting, and the way in which BBNs are used contains interesting insights. In particular, the BBN-based adaptation of dependability requirements to a specific application is the most interesting feature of the proposed approach, and I think all dependability-addicted readers will appreciate this insight. Online Computing Reviews Service