Ranking software components using a modified PageRank algorithm including safety aspects

Automotive software is increasing rapidly in its size, amount or complexity and the demand for calculation power rises. According ISO26262, additional non-functional requirements for safety-relevant applications must be considered. Therefore, analyzing software networks is necessary to identify problems within an early development state. For future multicore systems, complex software structures probably impede a safe segregation of software components between cores. In this paper, we want to use social network measurement methods to analyze automotive software architectures. Here, the interconnection of single software components and their influence within the network is focused. We use a modification of Google's PageRank algorithm, where websites partly inherit the rank of their interlinked neighboring websites, and transfer these techniques to automotive needs. For our evaluation, we use data structures of an exemplary electronic control unit of BMW.