Link failure protection and restoration in WDM optical networks
Is Version Of
In a wavelength-division-multiplexing (WDM) optical network, the failure of fiber links may cause the failure of multiple optical channels, thereby leading to large data loss. Therefore the survivable WDM optical networks where the affected traffic under link failure can be restored, have been a matter of much concern. On the other hand, network operators want options that are more than just survivable, but more flexible and more efficient in the use of capacity. In this thesis, we propose our cost-effective approaches to survive link failures in WDM optical networks. Dynamic establishment of restorable connections in WDM networks is an important problem that has received much study. Existing algorithms use either path-based method or link-based method to protect a dynamic connection; the former suffers slow restoration speed while the latter requires complicated online backup path computation. We propose a new dynamic restorable connection establishment algorithm using p-cycle protection. For a given connection request, our algorithm first computes a working path and then computes a set of p-cycles to protect the links on the working path so that the connection can survive any single link failure. The key advantage of the proposed algorithm over the link-based method is that it enables faster failure restoration while requires much simpler online computation for connection establishment. Tree-based schemes offer several advantages such as scalability, failure impact restriction and distributed processing. We present a new tree-based link protection scheme to improve the hierarchical protection tree (p-tree) scheme  for single link failure in mesh networks, which achieves 100% restorability in an arbitrary 2-connected network. To minimize the total spare capacity for single link failure protection, an integer linear programming (ILP) formulation is provided. We also develop a fast double-link failure restoration scheme by message signaling to take advantage of the scalable and distributed processing capability of tree structure.