Availability requirements in survivable transport networks depend on the type of costumers using the network and the supported services. Nowadays, a variety of services with different protection guarantees, also called Quality-of-Protection (QoP), are proposed through the same network at different rates. In this work, we propose a framework for optimized design of multiple quality-of-protection classes including single and dual link failure scenarios under arbitrary SRLG scenarios in survivable WDM networks that use pre-configured protection structures (i.e., p-structures).
We develop compact optimization models and propose a scalable solution method based on Column Generation (CG) where these two classes of QoP are guaranteed. Contrary to classical optimization techniques, where the shapes and protection capabilities of the potential protection structures are decided ahead of the optimization process, in our CG based approach, these characteristics are dynamically decided during the optimization process in order to effectively meet the QoP requirements of the supported users and traffic.
We test the proposed design method under several R2 levels, and compare the optimal capacity designs of the p-structures with the p-cycles in order to gain an insight about how the shape of the protection building blocks affect the performance of the protection scheme. Furthermore, the shape and protection capability of the p-structures are studied in order to illustrate the most appropriate structures for each QoP level. The computation results show that, in some test cases, and depending on the network connectivity, an up to 150% of protection capacity can be saved throughout the use of p-structures rather than limiting the protection structures to p-cycles. This illustrate the potential of adapting the shape of the protection structures in order to meet different QoP requirements. These design methods and results can be used by network planners to evaluate the availability, flexibility, and cost of the different capacity design strategies using pre-defined shape structures (e.g., p-cycles, p-trees).
Paru en février 2010 , 19 pages