| The significance of resilient communication networks in the modern society is well established. Resilience and survivability mechanisms in current networks are limited and domain specific. Subsequently, the evaluation methods are either qualitative assessments or context-specific metrics. There is a need for rigorous quantitative evaluation of network resilience. We propose a service oriented framework to characterize resilience of networks to a number of faults and challenges at any abstraction level. This dissertation presents methods to quantify the operational state and the expected service of a network using functional metrics. We formalize resilience as transitions of the network state in a two-dimensional state space quantifying network operational characteristics and service parameters. One dimension represents the network as normally operating, partially degraded, or severely degraded. The other dimension represents network service as acceptable, impaired, or unacceptable. Our goal is to initially understand how to characterize network resilience, and ultimately how to guide network design and engineering toward increased resilience. We apply the proposed framework to evaluate the resilience of three ISP backbone network topologies and the topologies generated using a new realistic topology generator. Furthermore, we quantify the resilience of MANETs at multiple layer boundaries. We develop new predictive routing algorithms for weather disruption-tolerant networks, which are shown to be more resilient than the existing protocols. Lastly, we develop a new routing algorithm for highly-mobile ad hoc airborne networking. |
