The potential of a single network failure to severely disrupt the economy and security of an entire nation makes survivability and fault tolerance key requirements for any core datacom or telecom network. With today's technological prowess, there is no reason why a network cannot be planned to be fully-survivable. For example, Deutsche Telekom uses 1 + 1 route diversity to protect every link in its core network. More recently, dual-ring networks have been the solution of choice because they provide single-fault full-survivability and flexible bandwidth allocation among all nodes.
Many existing core networks are based on the dual ring architecture because of their simplicity in planning and operation. However, they do not scale up well. In networks with large numbers of nodes the capacity utilization is very low, which results in very-high cost per unit of bandwidth. In addition, while mesh networks have been known to be scalable and resilient to faults, they are difficult to plan for and to operate.
Over the last few years, intelligent optical network technology has advanced very rapidly, including the release of a set of new International Telecommunication Union standards (ITU-T) for the optical network control plane called Automatically Switched Optical/Transport Network (ASON/ASTN). Together with the ITU-T's next generation transmission standards, the intelligent optical network promises to enhance the flexibility and controllability of a core network by enabling it to be provisioned in response to traffic demands in a matter of seconds.
Besides enabling a quick response to customer demands and higher reliability, ASON/ASTNs can be operated on mesh-based networks with a much better capacity-utilization than the ring-based networks. This increased efficiency significantly reduces both capital expenses (CAPEX) and operating expenses (OPEX) for network operators. Because of these overwhelming advantages, many carriers such as AT&T have already begun to replace all their current static core equipment with intelligent optical network technology.1
Despite all the advantages offered by the intelligent optical network, the very things that give ASON/ASTNs their flexibility make it difficult to guarantee their uninterrupted operation when traffic demand is changed. Simply said, the flexibility conflicts with the survivability requirement. To resolve the conflict, we have proposed a new network concept called the Generalized Survivable Network (GSN). 2–5 The main conceptual framework for GSNs has already been defined and the computation modules are available for use.
The GSN can be considered a generalization of the fully-survivable dual ring network to the mesh-based arbitrary topology network.5 The two important properties of the dual-ring network, full-survivability and dynamic traffic demand provisioning, are both preserved in a GSN. Furthermore, the mesh-based topology reduces the spare capacity required for protection and restoration and allows the network to scale much more easily. By incorporating the GSN into the intelligent optical network with the ASON/ASTN control layer, the core network can be guaranteed to be fully-survivable as well as able to meet any allowable dynamic demands.
GSN is a generalization of the classical non-blocking network to the optical case. Here, it is assumed that—given the source node has the available outgoing capacity and the destination node has the available incoming—a path is guaranteed to be found. It is also assumed that, because of the capacity constraints at the input and output of each node, the traffic demands from one node to all the others must satisfy the input/output constraints. This is the foundation for capacity planning and routing assignment in GSN. The algorithms for finding the routing paths can be further classified into strictly, wide-sense, and rearrangeably non-blocking algorithms. By adding the full-survivability constraints onto the generalized non-blocking network, the GSN is formed. A rigorous mathematical formulation can be found that defines these networks' characteristics.3,4
In terms of cost, the following simple example illustrates how GSN can reduce the CAPEX of a core network.5 It compares a four-node fully-survivable dual-ring network (Figure 1) with a four-node GSN mesh network (Figure 2). Both can fully recover from any single fault and can support arbitrary traffic demands among all source-destination pairs. Table 1 shows a breakdown of the components required for the two networks. Excluding fibers (most of which are buried in the ground), the cost saving of GSN over dual-ring is almost 100%.
Figure 1. A four-node survivable dual-ring network.
Figure 2. An optimal four-node fully-survivable GSN.
Table 1. Cost comparison between dual-ring and GSN mesh.
The GSN architecture allows easy traffic redeployment and load balancing by service providers while guaranteeing full survivability of the network during reconfiguration. The simple four-node network example illustrates the cost advantages of a mesh-based GSN. Their design for arbitrary topologies is very difficult but the problem has been partially solved. 4 The GSN concept can be expanded beyond optical backbone networks such as ASON/ASTN and applied to networks running Internet Protocol (IP) and virtual private networks. This makes these architectures potentially, and deserving of further exploration. Some emerging issues within GSN include the need to develop a better understanding of the different classes of routing algorithms and survivability mechanisms that can be implemented, and the behavioral properties they display.
Michael Kwok-Shing Ho and Kwok-Wai Cheung
The Chinese University of Hong Kong
Sha Tin, Hong Kong
Kwok-shing Ho received a M.Phil. degree in information engineering from the Chinese University of Hong Kong in 2002. Since then, he has been a Ph.D. candidate in information engineering at the Chinese University of Hong Kong, where he is active in survivable-network planning research. His current research interests include the design and evaluation of survivable optical networks. He received the best student paper award at APOC 2005.
Kwok-wai Cheung is currently a professor in the Department of Information Engineering at the Chinese University of Hong Kong. He received his B.S., M.S., and Ph.D. from the University of Hong Kong, Yale University, and California Institute of Technology respectively. From 1987 to 1992, he worked as a member of Technical Staff at Bell Communications Research (now Telcordia Technologies), New Jersey, USA. He joined the Chinese University of Hong Kong in 1992. From 1996 to 2001, he served as the founding Director of the Centre for Innovation and Technology. He has published over a hundred journal and conference papers, holds five U.S. patents, eight international patents, and has over twenty others pending. He was awarded an Electronics Letters Premium in 1992.