Analysing Software Fault Trees via a Key Node Metric

Complex software systems for business critical and safety critical applications require tools for business performance managers to use, especially in software systems in which failure leads to major economic failure or loss of life. The software engineering community stands to benefit from metrics, analysis tools, and techniques that address software system safety from a design perspective. The design-time use of software fault trees in representing the structure of a software system allows designers to focus on business and safety critical aspects of software during early development stages thereby allowing business performance managers to provide cost-effective oversight of software development. This paper applies a technique for evaluating systems through the analysis of software fault tree “key nodes” that require multiple inputs to fail before the failure being considered will propagate to other parts of the software system. The metric provides business managers with a tool through which to control costs during software development. A heuristics-based Key Node metric is presented, and provides a design tool with which to compare fault trees without requiring a priori knowledge of component reliability. The ability of the metric to be applied without requiring a priori component reliability knowledge allows the metric to be used at design time where component reliability values for a type of component are often unknown. The Key Node metric allows designers to proactively improve the business or safety critical aspects of a system before final component selection or the completion of component reliability studies. The paper provides an application of the Key Node Metric, and discusses the results of applying the metric to a product line represented by a set of software design mutations.