Failure Localization Via a Central Controller

TL;DR: The chapter is on the m-trail allocation problem by introducing algorithms and approaches in presence of single and multiple link failures, respectively, including Random Code Swapping (RCA–RCS) for single-link failures, and a suite of heuristics for general topologies.

Abstract: To achieve fast Unambiguous Failure Localization, an essential problem for the network operators is to determine how to efficiently probe the network elements such that the number of probes is the minimum. By launching a set of m-trails, the transmitter of each m-trail constantly probes the health of the links along the m-trail, and the monitor at the receiver issues an alarm once detecting any irregularity. A failure may interrupt multiple m-trails which incurs a set of alarms. The m-trails should be allocated such that the network controller can uniquely and precisely localize the failure state according to the issued alarms. The chapter is on the m-trail allocation problem by introducing algorithms and approaches in presence of single and multiple link failures, respectively. With single-link failures, an essentially optimal construction for m-trail allocation is provided for lattice topologies. For general topologies, a suite of heuristics are presented, including Random Code Swapping (RCA–RCS) for single-link failures, Adjacent Link Failure Localization and Link Code Construction for adjacent link failures, and Greedy Code Swapping (CGT-GCS) for dense-shared risk link group failures based on combinatorial group testing.