Shared Risk Link Groups of disaster failures

Abstract

Current backbone networks are designed to protect a certain pre-defined list of failures, called Shared Risk Link Groups (SRLG). A common assumption is to ignore the failures not part of an SRLG as they assume to be extremely rare events; therefore, the list of SRLGs must be defined very carefully. The list of SRLGs is typically composed of every single link or node failure. It has been observed that some type of failure events manifested at multiple locations of the network, which are physically close to each other. Such failure events are called regional failures, and are often caused by a natural disaster. The number of possible regional failures can be very large, thus simply listing them as SRLGs is not viable solution. In this study we focus only to every regional failure with a shape of disk that do not contain network nodes. We provide a fast systematic approach for generating a list of SRLGs the protection of which is essential to increasing the observed reliability of the networks. According to some practical assumptions this list is very short with O(|V|) SRLGs in total, and can be computed very fast, in O(|V|log|V|).