Optical networks employing wavelength-division multiplexing (WDM) provide an ideal infrastructure to satisfy the huge (and increasing) bandwidth demands. However, the Future Internet (FI) is expected to accommodate many new services, which are not only bandwidth-intensive, but also characterized by heterogeneous quality-of-service (QoS) requirements. To capture this QoS heterogeneity and achieve high customers' satisfaction (by meeting Service-Level Agreements, SLA) as well as resource efficiency, service-centric provisioning is desirable. This dissertation studies novel models, algorithms, and architectures that will help achieve service-centric provisioning for next-generation optical WDM networks.;We first study dynamic provisioning/reprovisioning schemes to satisfy SLAs. We assign an Urgency Level (UL) to each service indicating its status and relative importance (technically and economically). We propose an adaptive provisioning scheme that tracks a service's status, and reallocates resources from low-UL services to higher-UL services, which are more at risk of violating their SLAs or have higher penalties.;Next, we find that SLA is always at risk because of the dynamism of network failures. We show that SLA Violation Risk may vary between paths, and is affected by many factors (e.g., failure rate, connection holding time, etc.). We formulate the risk-aware routing problem in optical mesh networks, and present an efficient routing scheme that computes routes that are likely to successfully accommodate the SLA-requested availability.;To accommodate customers' extra resilience requirements for specific time periods, we propose a novel SLA framework that supports Critical Windows (CW) to address these time-differentiated resilience demands. We identify opportunities for backup resource sharing in a time-domain multiplexing manner, and propose efficient schemes for connection assignment to minimize backup resources.;We propose to back up services using primary paths of other services. We present a new provisioning framework called Service Cluster (SC) that uses no explicit backup resources but still provides protection. We devise an event-triggered scheme to support dynamic resource allocation in an SC.;Typically, a certain amount of Operational Power is needed for service provisioning. We analyze the constituents of Operational Power, and discuss the opportunities for energy saving following a Traffic Engineering (TE) approach. Capturing the energy efficiency of provisioning strategies (i.e., optical bypass vs. traffic grooming) using a novel auxiliary graph, we present a Power-Aware provisioning scheme to minimize the Operational Power.
展开▼