Font Size: a A A

Data network recovery from multiple failures

Posted on:2008-08-25Degree:Ph.DType:Thesis
University:The University of Texas at DallasCandidate:Wu, KaiFull Text:PDF
GTID:2458390005979902Subject:Engineering
Abstract/Summary:
As communications networks play deeper role in people's life, new network applications and services are ever growing. Network architecture has also been undergoing significant changes. The Internet presents an easily accessible, ubiquitous platform for converged services. It is hence important to maintain a highly reliable network while accommodating various needs.; Recovery schemes, under the assumption of single failure scenarios, have long been studied, where each single failure can be an independent failure of a single link, node or shared risk link group (SRLG). These schemes include both protection and restoration. However, today's network also subject to multiple element failures that may occur concurrently. However, coping with all the possible failure scenarios becomes a challenge as they are more disruptive and may require more bypass tunnels to be dealt with, when compared to single failure solutions.; The objective of this thesis it to explore and propose new solutions to deal with multi-failure scenarios. The main focus is on pre-planned protection using local recoveries. Two major challenges are to provide fast detection of failure status of non-adjacent network elements and to reduce the use of extra network resources, including bypass tunnels. The proposed solutions should also retain the simplicity and speed of today's (independent single failure) solutions as much as possible.; This objective is achieved by using concept of Probable Failure Pattern (PFP)'s to represent multi-failure scenarios. Various expected failure patterns may be grouped into one or more clusters in order to simplify both detection and bypass tunnel pre-planning.; Using the statistic model of PFP's, resource efficiency could be further improved by trade complexity such as bypass tunnel length with network performance such as recovery ratio.; The identification and characterization of multiple failure patterns is also discussed in detail. A tool package that is used on a field network to collect and analyze failure statistics.; Finally, the restoration realm of network recovery is investigated. Restoration schemes can deal with multiple failures by nature, since restoration path are sought on the fly. The proposed schemes could also provide differentiated reliability to achieve better network resource utilization under stress conditions.
Keywords/Search Tags:Network, Failure, Recovery, Multiple, Schemes
Related items