Computing the availability of entire networks can be cumbersome, since in practice networks are large. The availability of the entire network does not provide much information, since mostly certain paths in the network are used. In our research, we consider a communication network. The system is modelled by an undirected graph. Each vertex can fail, it is repaired upon failure and it is characterized by its mean time between failures and its mean down time. For a path in the network, connecting two arbitrary nodes, the availability, the mean time between failures, and also the mean down time of the path can be computed. For the mean down time, an approximate formula is given. Using theoretical data, a prediction for the availability characteristics for paths can be given. In the example studied, theoretically computed characteristics for paths are compared with observed data. The research has shown, how availability of network components can be verified based on data for paths in the network, rather than using data for components. The latter is sometimes complicated since many components in a network only rarely fail and statistical data are hard to obtain and as a consequence, computed characteristics have very large confidence bound.