The Dependency Model: A Tool for Calculating System Effectiveness

The Dependency Model (DM) is a mathematically unsophisticated but useful and practical tool for measuring the effectiveness (reliability, availability, maintainability, efficiency, etc.) of a complex system. Contrary to classical models, the DM does not require arbitrarily chosen statistical distributions or expensive simulations for its application. The large computer storage and running time requirements, the need for an accurate random number generator, and the inherent stochastic inaccuracy associated with such Monte Carlo techniques are thereby avoided. However, the DM does require the user to assign relative weights to the elements of the system, and to measure the performance of the lowest level elements considered. Common models such as GEM, TIGER, and REX assume a series-parallel structure. Although this might be sensible when applied to the lowest replaceable units (LRUs) in a hardware structure, it is unreasonable to apply it to components further up in the hierarchy, or to software. Thus, redundancy is not clearly defined for software. If there is an error in the code, placing identical programs in parallel does not eliminate that error. The series-parallel logic structure allows only for each component to be up or down, with no gradation. In contrast, the DM measures the degree of failure of a component or system, in addition to the probability and method of failure that can be determined from classical models. The elementary structure and flexibility of the DM yield additional benefits.