Analytical Frame work for Fault Management in Communication Networks

Abstract

Fault management systems plays an exceptional role in the design of high performance communication networks especially in real time network applications to ensure continuous reliability and guaranteed availability of resources. Fault management system must detect, isolate and fix up the network problems, a task which is complex for large networks and requires more number of resources. So there is a need for having a learning and prediction based fault tolerant method which manages faults in diverse aspects. This demanded an effective distributed fault detection method to enable speedy decision-making in the early detection and diagnosis of defects. newlineThe Congestion in the network plays a vital role in defining the Quality of Service offered by the network. It is always desirable to reduce the overloading conditions as much as possible. For small networks, finding the fault is easy using the self-learning algorithms which works on dynamic environment. As the users of the network increasing rapidly, an alternate method must be invented. The best solution for this can be provided using machine learning algorithms which has the capability to predict the network parameters using historical relationships between data. Fault tolerant networks need to employ different techniques at connectivity phase and infrastructure phase. This thesis contributes analytical approaches in each of these phases to optimal QoS parameters. newlineThe thesis proposes a reward based algorithm at connectivity phase, to distribute the workload across the network and hence to reduce the probability of overloading conditions to achieve fault tolerant system. The protocol suggests routing the packets through that path which has the optimum rewards. Further, load balancing in software defined network (SDN) is deployed at connectivity phase. To avoid single point failure, decentralized architecture approach is employed. Also a Hybrid algorithm for load balancing in SDN is designed and developed which outperforms the existing algorithms. newline