Using network measure to reduce state space enumeration in resilient networks

Resilient network design has become a prominent research topic in the wake of recent disasters in Japan and the United States. Often communications networks that are designed to be highly reliable have significant outages when rare but important events occur. Finding rare events is generally difficult because it may require an exhaustive search of the network state space. Mission critical systems have specific qualities that enable enumerating the network space for important events more efficient. In this work, we use mission critical system qualities and requirements to create a tractable approach to evaluate a network for resiliency. With specific system requirements, the Self-Pruning Network State Generation algorithm is able to completely evaluate the failure modes of a network by examining approximately 90,000 events as opposed to an exhaustive search requiring 5.5 × 10¹¹ events. In addition, a network resiliency metric is proposed that considers not only failure mode probability but also system impact. These approaches are also flexible, working with multiple network measures and network types.