Date of Award

2012

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical Engineering

First Advisor

Bikdash, Marwan

Abstract

Survivability is the ability of a system to continuously deliver essential services despite attacks, failures, or accidents that damage a significant portion of the system. Most of existing survivability measures evaluate global impact of faults on a network by averaging the loss in performance over the whole network. These approaches undermine the fact that a fault is not expected to impact all parts of the network equally. Impacts of some faults are contained in local neighborhoods while others have a global reach. In addition, graph algorithms are often computationally intensive, and even polynomial time algorithms get impractical for a fair sized network. In this dissertation, the interplay between geographic information about the network and the principal properties and structure of the underlying graph are used to quantify the structural and functional survivability of the network. This work focuses on the local aspect of survivability by studying the propagation of loss in the network as a function of the distance of the fault from a given origin-destination node pair. Geographic-based partitioning and graph-based representations of the interactions of the partitions are used to build a coarsened network. The partitions are designed to behave as a subnetwork. A complexity reduction in network computation is achieved by performing the desired computation on the subnetworks and the coarsened network. The overall network parameters are determined by merging the probability distributions in the subnetworks with the parameters of the coarsened network.

Download

Share

COinS