Dual-Failure Availability Analysis of Span-Restorable Mesh Networks

Modern networks are typically designed to be survivable for single-failure scenarios only. While there does exist some inherent dual-failure survivability, resource limitations generally mean dual failures are not fully protected or restored; there is some nonzero level of unavailability arising due to dual failures. Understanding the complexities of dual-failure related availability, therefore, is an important issue in network design. There are a number of current approaches for evaluating network availability, whether directly, by calculating service path unavailability, or indirectly, via several widely used availability metrics. However, their performance is not as effective as widely believed. After investigating current availability methods, this paper proposes a new method to calculate network service unavailability. An ILP model framework is also built to qualify the new method and evaluate current and new availability analysis methods. Experiments are performed on four test-case networks of various topologies and scales.