This project investigates the static and dynamic partitioning of optical networks into domains or clusters and addresses the problem of providing survivable end-to-end service provisioning for unicast and multicast traffic over such multi-domain networks. A large optical network, even if administered by a single operator, needs to maintain its state information in a distributed and multi-domain hierarchy for purposes of scalability and fault-tolerance. Moreover, dynamic reconfiguration is often necessary to respond and adapt to sudden unexpected events or changes in network conditions. Other significant challenges in such a heterogeneous multi-domain environment are to implement end-to-end routing and survivability mechanisms, which are complicated by the limited amount of state information shared across domains due to factors related to business, policy, security, and other issues. This project enables a self-organizing framework for survivable optical networks. The project is addressing new methods for defining and reconfiguring domains based on service and performance requirements, introduces several new mechanisms for providing survivability in multi-domain networks, and is introducing new algorithms and protocols for provisioning survivable end-to-end unicast and multicast services in a scalable and cost-efficient manner. The broader impact of this work is that it can lead to a more robust and dynamic optical control layer that will enable the deployment of a greater range of end-to-end optical services to support emerging applications. It envisions greater intelligence in the control plane of optical transport networks, potentially leading to a network that is more autonomous than networks in existing ASON and GMPLS frameworks.
The emergence and growth of high-bandwidth network services have prompted the need for dynamic optical networks that are capable of automatically provisioning end-to-end services on demand. As optical networks continue to grow and as requests become more dynamic, it becomes more difficult to manage the required state information for implementing various routing and provisioning protocols in a single large optical network domain; therefore, for purposes of scalability and ease of management, large optical networks may often be partitioned into multiple domains, with the domain boundaries determined by factors such as administrative boundaries, geographic regions, equipment technologies, or other policy-based criteria. The potential diversity of capabilities among domains and the possibility of limited state information availability between domains introduce significant challenges in provisioning end-to-end services across multiple domains. A critical issue in emerging multi-domain optical networks is that of network survivability. Survivability schemes ensure that the network is able to continue carrying traffic in the event of network failures.While survivability mechanisms have been extensively studied for single-domain networks, work on survivability for multi-domain networks is more limited. Implementing survivability mechanisms in multi-domain networks is complicated by the limited amount of state information that is exchanged among domains. This project focused on the investigation and evaluation of methods for supporting survivability in muti-domain optical networks. One primary focus of the project was the problem of how to provision working and route-disjoint back-up resources for a connection that may span multiple domains. Typically, each domain provides only limited information regarding connectivity within the domain due to scalability issues or administrative policies. In this project, we developed and evaluated several new topology aggregation mechanisms in which each domain provides additional disjointness information in order to facilitate the provisioning of disjoint paths. Our results demonstrated that the proposed approaches can significantly increase the chance of provisioning survivable connections across multiple domains. Another problem is that of survivable virtual optical network (VON) mapping over a multi-domain optical network. In this problem, virtual optical nodes must be mapped to physical nodes within domains, and virtual edges in the VON topology must be mapped to inter-domain paths. As a first step, we considered the VON mapping problem within a single domain and developed a heuristic for mapping nodes and edges while satisfying certain physical constraints. We then investigated the problem of VON mapping over a multi-domain optical network and the problem of survivable virtual topology mapping over a multi-domain optical network. Our work has lead to the development of new techniques for mapping a virtual topology to multi-domain physical topology while ensuring survivability and minimizing cost. In terms of broader impact, many of the proposed techniques can be incorporated into existing or emerging optical network control planes, enabling network operators or service providers to provide a greater range of reliable services in a scalable and cost-effective manner. Furthermore, the project has provided valuable research training experience for several graduate students.
