ABSTRACT
Before computerised systems can be used in any kind of critical applications, evidences that these systems are dependable are required. This is desirable for most systems, but essential for systems which affect human safety and welfare, e.g. patient monitoring systems, fly-by-wire systems, railway signal systems, etc. Considering that most computerised systems are built as a structure of several software components, of which some might have been pre-developed and used in other contexts, there is a need for methods for assessing reliability of compound software 1 . In this paper, we investigate the main challenges of our component-based approach for assessing software reliability when we go from binary systems of binary components to multistate systems of binary components. Especially, we look at the effect of using only partial dependency information when assessing reliabilities and unreliabilities of multistate systems. To find the most important component dependencies in each system state, direct calculation and the Birnbaum measure are applied on an example from medical monitoring.
