Recent advances in optical wavelength division multiplexing (WDM) and optical switching have enabled the development of next generation computer networks, to operate at several Terabits per second. A side impact of this development is that fault tolerance has become even more important, since a single failure may now result in much greater loss of information. The proposed research will investigate critical problems in the design of survivable optical wavelength division multiplexed (WDM) networks that are based on mesh topology. The goal is to provide reliable connections, based on deterministic and probabilistic requirements, to the clients using the network.

This topic constitutes an important component of the nation's critical infrastructure protection, given that optical WDM networks are fast becoming part of the high-speed national wide-area backbone infrastructure. The differentiating factors of the proposed research include:

(i) wavelength sharing, where each session requests partial wavelength capacity, and multiple sessions share a single wavelength,

(ii) dynamic network routing, where the routes are determined on a dynamic basis, considering current network status information,

(iii) decision making based on partial information, where a given node constructs its routing table based on current available (semi-local) information, as compared to a network with centralized decision that uses full global information.

The specific research objectives are:

(i) developing techniques for the design of survivable optical wavelength division multiplexed networks;

(ii) capacity minimization and revenue maximization using reconfiguration techniques during various operational phases, for dynamic traffic;

(iii) deriving practical and tractable optimal and near-optimal solutions for networks with large number of nodes (50 - 100), with several fibers per link and more than 100 wavelengths per fiber; an additional objective is achieving sub-millisecond restoration times, compared to the few-seconds solutions of today;

(iv) developing dynamic routing techniques based on partial network information; this will use intelligent WDM-specific link cost functions to determine suitable routes; and (v) developing multiple backup path techniques to reduce the computational complexity associated with computing node- and link-disjoint paths when backup paths are shared among several sessions.