Issues in Predicting the Reliability of Composed Components

Availability is one of the most frequently specified quality attributes for computerized systems and the computation of availability requires knowledge about the reliability of the system. Although much research has been devoted to software system reliability, much work remains to be done in identifying ways to predict reliability of assemblies of components. We are designing an experiment for use as a foundation for creating a reliability prediction-enabled component technology (PECT), which is to be used to produce systems that are predictably reliable by construction; in the course of that work we have recognized the need to evolve combinatorial reliability models for use in computing reliability of assemblies based on the reliabilities of constituent components. In this paper, we describe and discuss aspects of current models that need to be adapted and how they affect the design of our experiment.