Evaluating COTS component dependability in context

Abstract

The software industry has increasingly expanded its adoption of COTS components for complex, mission-critical applications. Using COTS products can shorten development and deployment time because they let system developers focus on creating domain-specific services. Selecting the right COTS component, however, is no easy task. We present a practical process that developers can use to empirically evaluate component dependability in their context. Our approach uses the unified model of dependability, a requirements engineering approach specially devised to capture dependability in context. This model clearly specifies the measurable characteristics the component must have to be dependable in a specific context. The model then serves as a reference, providing guidance on effectively designing experiments to compare similar components and interpret collected data. The process can be applied to any specific context and COTS component. Here, we describe our comparison of the dependability of real-time Java virtual machines (RT JVM) in the spacecraft software context.