The importance of the Internet and its services to society make it evident that it should be
made resilient to failures. This awareness has instigated a large body of research on how to
protect networks, although they typically consider a single simplistic failure model in which the network is represented by a graph consisting of nodes and links. In order to protect the Internet against failures, we believe that it is essential to understand what kinds of failures exist, the impact they have, and the frequency at which they occur; in other words, a taxonomy of Internet failures. Even though large-scale Internet incidents have been reported in the media, and some papers include a brief list of several such failures, a taxonomy of key Internet failures showing the cause, duration, range, and effect does not yet exist.