Survivability analysis and design of broadband communication networks

Abstract

The rapid development of high capacity transmission facilities such as fiber optic has rendered the structure of present-day communication networks much simpler than it used to be. Whereas the simplicity in network structure lessens the cost burden associated with network construction and maintenance, it in return makes networks more susceptible to severe network-service degradation under some component failures. On the other band, the performance standards imposed on the networks are getting stricter as the society itself becomes more dependent on the information provided via such networks. These technological and environmental changes, in turn, must call for the highly reliable and survivable network structure to be employed. Particularly, a two-level hierarchy with multiple homing of user nodes and general mesh-like structures are considered as most promising alternatives for present-day survivable networks. In accordance with this line of advances, this thesis first analyzes the network survivability with focus placed on lick failures. To authorize the more quantitative analysis than possible with the conventional connectivity-oriented approach, the network link survivability is defined as the relative portion of traffic amount still intact under a set of link failures. We then exploti its characteristics on the case of two-link failures to provide an efficient computation procedure. Based on a graphtheoretic result, the so-called cycle-factoring operation is devised, which reduces the network at hand into two simpler ones. We also develop a branch and bound procedure for two-link survivability of a general network, which successfully incorporates a sequence of cycle-factoring operations and some other computation-saving subprocedures. Next we investigate the design methodologies for the networks on which a certain kind of survivability conditions are imposed. With a two-level hierarchical network with full-mesh interconnection of backbone nodes, we show th...