Survivability of gmpls-based ip-over-optical networks

As the optical cross-connects have become cost-effective, Internet service providers (ISPs) have begun to consider carrying IP traffic directly in the optical transport domain, bypassing any intermediate layers, such as SONET and ATM. The generalized multiprotocol label switching (GMPLS) standard has been developed to enable these opportunities. By combining the revenue-generating features of the IP layer with the bandwidth-rich advantages of the optical layer, the IP-over-optical network promises a long-term solution to the problem of sustaining the ever-growing Internet traffic. Before this promise can be realized, however, various issues concerning the integration of the IP and optical layers need to be addressed. One of the critical issues concerns survivability, or resiliency of networks against failures. This dissertation addresses the survivability of the IP-over-optical networks from two directions: data plane survivability and control plane dependability. Data plane survivability is ensured via dynamic connection provisioning. The end-to-end survivable provisioning problem is tackled from two perspectives: intra-domain provisioning and inter-domain provisioning. For the intra-domain provisioning, we propose a scheme for dynamic provisioning of protection-guaranteed EP services in IP-over-optical networks. Our proposed provisioning scheme takes advantage of the synergy between the IP and optical layers through two novel path selection algorithms. These path selection algorithms utilize knowledge of topology, resource, and protection in both the EP and optical layers to compute a path or a pair of paths that meets a user's survivability requirement while minimizing resource consumption. We also study how this new synergistic provisioning approach impacts the design and planning of survivable IP-over-optical networks. For the inter-domain provisioning, we introduce three policies for coordinating the end-to-end survivability guarantee across multiple autonomous ISPs. We show how existing ISPs can benefit from these policies by analyzing potential performance gain, as well as potential financial profitability. Control plane dependability is guaranteed via a new GMPLS fault tolerant paradigm. The new paradigm utilizes existing network management objects to validate control plane actions against data plane traffic activity. This new paradigm provides an added level of fault tolerance against issues encountered in today's GMPLS protocols, such as signaling deadlocks, unauthorized disconnections, and memory leaks.