Modelling of Software Failures

Software is crucial in the provision of communication services. Most functions related to control, management and operation are realized in software. With the ongoing virtualization and shift of network functions to new software platforms, the role and criticality of software for ordinary operations as well as handling of disasters increase significantly. As discussed in the introductory section, malfunctioning software may cause disasters as well. The objective of this chapter is to provide insight into the software failing process and give an introduction to quantitative models to predict the impact of software failures to communication systems and services. These models have been applied to software-defined networking (SDN). It is shown how the effect of various kinds of software failures and their rectification may be modelled for a controller by using stochastic Petri nets/stochastic activity networks. Furthermore, the software failure intensity may be predicted by non-homogeneous Poisson process (NHPP) reliability growth and fault correction models. It is demonstrated how these models have been successfully applied to multi-release developments with real data from the ONOS and OpenDaylight projects.