Realistic abductive reasoning-based fault and performance management in communication networks

Abstract

Abductive reasoning is identified as a suitable candidate for solving network fault and performance management problems. A method to solve the network fault diagnosis problem using realistic abductive reasoning model is proposed. The realistic abductive inference mechanism is based on the parsimonious covering theory with some new features added to the abductive reasoning model. The network diagnostic knowledge is assumed to be represented in the most general form of causal chaining, namely, hyper- bipartite network. As many explanations may still be generated by the realistic abductive reasoning model, we propose a probabilistic method to order them so as to try out the diagnostic explanation in the decreasing order of plausibility until the hard failure-like faulty device is isolated and replaced/cop-ected.

In contrast, performance degradation in communication networks can be viewed to be caused by a set of faults, called soft failures. owing to whkh the network resources like bandwidth cannot be utilized to the expected level. An automated solution to the perfonnance management problem involves identifying these soft failures and use/suggest suitable remedies to tune the network for better performance. Abductive reas.oning model is used again to identify the network soft failures and suggest remedies. Common channel signalling network fault  management and Ethernet performance management are taken up as case studies. The results obtained by the proposed approach are encouraging.