Automatic Discovery of Diverse and Dynamic Network Services

Common-mode failures have thus far been considerably overlooked in reliability studies. By their definition, common-mode failures correlate heavily, breaking the assumption of probabilistic independence. This article describes how monitoring a live, largescale network infrastructure provides a realistic environment for analyzing common-mode failures, while also introducing new problems not immediately obvious in analytical or simulation studies. Administrators perform changes to a system whenever new requirements, research projects, or unplanned failures manifest. To distinguish changes from failures, the process from network discovery to service monitoring must be streamlined and enabled to handle extremely diverse installations.