Experience in Metrics and Measurements for N-version Programming

The N-Version Programming (NVP) approach applies the idea of design diversity to obtain fault-tolerant software units, called N-Version Software (NVS) units. The eeectiveness of this approach is examined by the software diversity achieved in the member versions of an NVS unit. We deene and formalize the concept of design diversity and software diversity in this paper. Design diversity is a property naturally applicable to the NVP process to increase its fault-tolerance attributes. The baseline design diversity is characterized by the employment of independent programming teams in the NVP. More design diversity investigations could be enforced in the NVP design process, including diierent languages, diierent tools, diierent algorithms, and diierent methodologies. Software diversity is the resulting dissimilarities appearing in the NVS member versions. We characterize it from four diierent points of view that are designated as: structural diversity, fault diversity, tough-spot diversity, and failure diversity. Our goals are to nd a way to quantify software diversity and to investigate the measurements which can be applied during the life cycle of NVS to gain conndence that operation will be dependable when NVS is actually employed. The versions from a six-language N-Version Programming project for fault-tolerant ight control software were used in the software diversity measurement.