User bandwidth demands have continued to expand at an unprecedented rate over the past two decades. This growth has been driven by the emergence of many new applications across a wide range of sectors, including commercial, private, and scientific computing. As a result, network carriers and operators have deployed a wide range of high-speed technologies to meet their growing needs. In particular, these solutions include state-of-the-art Internet Protocol (IP) and Ethernet systems as well as optical wavelength division multiplexing (WDM) networking platforms. In particular, the latter solutions provide unmatched terabit-per-second speeds and are commonly used to provide underlying "lightpath" connectivity between IP and Ethernet devices.;Now researchers have developed a range of schemes for lightpath provisioning and survivability in WDM networks. Most notably these solutions include optimization and heuristic-based strategies to solve the routing and wavelength assignment (RWA) problem. However, as WDM deployments have expanded, there is a further need to provision lightpath connections across multiple network domains. For example, these domains can be delineated in a variety of manners, including administrative ownership (intra- and inter-carrier), vendor or technology type, geographic, etc. Given the above, multi-domain (optical) network provisioning and survivability has become a key focus area. Indeed, this is a rather challenging problem as scalability and privacy concerns limit the amount and type of information that can be shared across domain boundaries, i.e., particularly in inter-carrier settings. Hence researchers have developed various solutions, with most using distributed graph-based heuristics to resolve connection routes with partial (dated, inaccurate) network state.;Nevertheless, it is well-understood that heuristic schemes are sub-optimal in nature and cannot provide any bounds on network performance. As a result, most multi-domain studies have used other heuristics for comparison purposes. In light of this, it is very difficult for network carriers to gauge the true achievable performance of their multi-domain networking setups. However, optimization-based methods offer a very effective means of formally analyzing network performance under idealized conditions with full a-priori knowledge of user demands. Moreover, these schemes have been widely-used to bound lightpath RWA performance in single-domain settings. Nevertheless, the further application of such methods in multi-domain network settings has not yet been considered.;To address these concerns, this dissertation presents a comprehensive optimization-based study of lightpath routing and survivability in multi-domain optical networks. First, a novel (two-stage) hierarchical model is introduced to optimize lightpath routes over inter-/intra-domain topologies pursuant to several traffic engineering (TE) objectives, i.e., including throughput maximization, resource minimization, and load balancing. Next, this model is extended to implement lightpath protection recovery for single-link failures by adding link-disjointness constraints at both the intra- and inter-domain levels. Finally, a novel optimization formulation is also developed to implement probabilistic lightpath protection for multiple correlated failures, i.e., as occurring during large-scale disaster events. The performance of these differing schemes is also tested for several multi-domain network configurations and compared against some advanced heuristic strategies, i.e., including regular (working-mode) provisioning, single-link protection, and probabilistic protection. Overall, the detailed findings from this effort show that the new optimization schemes give significantly better results, thereby providing an invaluable benchmark reference from which to develop improved solutions.
展开▼
