Fault localization for next generation survivable all-optical networks

Title: Fault localization for next generation survivable all-optical networks
Authors: Khair, Mazen George
Date: 2008
Abstract: We propose a distributed fault localization protocol, called Parallel Limited Perimeter Vector Matching (P-LVM) protocol, for localizing multi-link failures in all-optical networks. The P-LVM protocol is based on the Limited Perimeter Vector Matching (LVM) protocol for localizing single-link failures. To handle multi-link failures, it tries to separate each failure in a small perimeter area after identifying each perimeter area with its corresponding failure and then localize the failures in parallel respectively in a distributed manner. Through simulation results, we show that the P-LVM protocol can effectively localize multi-link failures that occur simultaneously or at different times. We also study the optimization problems in applying the LVM protocol in static all-optical networks. We consider two optimization problems: one is to optimize the traffic distribution so that the fault localization probability in terms of the number of localized links is maximized and the other is to optimize the traffic distribution so that the time for localizing a failed link is minimized. We formulate the two problems into an integer linear programming problem, respectively, and use the CPLEX optimization tool to solve the problems. By optimizing the traffic distribution the fault localization probability can be maximized and the fault localization time can be minimized. Moreover, a heuristic algorithm is proposed to evaluate the optimization results through simulation experiments. Finally, we propose two different fault localization protocols for localizing single-link failures in multi-domain all-optical networks. These two protocols are based on LVM protocol mechanism, which restricts fault localization within a smaller perimeter area and can thus significantly reduce the time and space complexity for fault localization. The first protocol assumes the existence of power monitoring only at edge nodes, whereas the second protocol assumes no power monitors at the intermediate nodes or the edge nodes. The two protocols can localize both inter-domain and intra-domain link failures without exchanging any internal confidential domain-specific information between different domains. The analytical results show that both protocols can not only fast localize an inter-domain link failure between different domains but also localize an intra-domain link failure that affects inter-domain traffic faster than the Open-Shortest-Path-First (OSPF) protocol in a large network.
URL: http://hdl.handle.net/10393/29614
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
NR48657.PDF2.63 MBAdobe PDFOpen