RESUMEN

Developers typically address dependability concerns in the late phases of system development. However, two trends are compelling developers to consider dependability earlier, at the architectural level. First, emerging applications are increasingly complex. Second, to address this first trend, developers are increasingly attempting to build dependable systems from existing undependable components. A new architectural approach employs exception handling to represent and analyze fault-tolerant software systems. It partitions architectural elements into normal and exceptional parts, thus promoting a clear separation of concerns regarding how to detect and handle errors.This article is part of a focus section on software architecture.

RESUMEN

In contemporary societies, individuals and organizations increasingly depend on services delivered by sophisticated software-intensive systems. Dependability has become a key systems property, which needs to be engineered and guaranteed regardless of continuous, rapid, and unpredictable technological and context changes. But, what is dependability? Dependability relates strictly to the specific goals the system's users want to achieve. So, in different circumstances, dependability implies different system attributes (for example, availability, real-time response, and ability to avoid catastrophic failures and resist adverse conditions), as well as different levels of adherence to such attributes. Achieving and maintaining dependability can't happen without firmly understanding its meaning. In this perspective, the authors introduce the Unified Model of Dependability, a modeling language that lets you reason about dependability and turn it into clearly defined and implementable system properties. Experience in the NASA High Dependability Computing Program provides a case study for UMD.