Survivability approaches for multiple failures in WDM optical networks

Abstract

In this dissertation, we study survivability paradigms for surviving multiple link failures. We study the network design and upgrade problem in WDM backbone networks and formulate it using a simulated annealing based technique. This framework provides a cost effective way of upgrading the network by identifying how much resources to budget at each stage of network evolution. This results in significant cost reductions for the network service provider. We study protection reconfiguration using two different survivability paradigms, namely sub-graph routing and shared mesh protection. Initially sub-graphs are defined taking into account link, SRLG, or node failures, and could, in the event of an unrelated or subsequent multi-link failure, incorporate the reactive form of sub-graph fault tolerance. Reactive sub-graph fault tolerance employs a recursive approach for tolerating numerous sequential overlapping failures. It can tolerate simultaneous multiple link failures by simply serializing the handling of each individual fault. Connection re-routing and network reconfiguration is one of the primary challenges in sub-graph routing methodology. We propose a constrained subgraph routing methodology, which restricts the connections to be routed using the same trunk and channel in the subgraphs, thus minimizing reconfiguration. The subgraph based routing methodology is further explored to tolerate multiple link failures, in the form of shared-risk link groups and node failures. The generalized diverse routing problem, for finding two diverse routes between a source and destination has been shown to be NP-Complete. Recent studies have also proven the NP-completeness of the SRLG diverse routing. We propose a polynomial time graph transformation algorithm for solving the diverse routing problem for certain specific SRLG's, which includes link-sets incident on a common node. The proposed graph transformation methodology for diverse routing, is also studied for shared-risk node group (SRNG) failures. (Abstract shortened by UMI.)