Research On Porbe Station Deployment Of Distributed Network Fault Management

As an important component of network management system, network faultmanagement is the key to maintain network reliability and the main security approachto control the normal operation of the network. The primary concern of network faultmanagement that is based on active probing is to select the suitable sites for probestations deployment. As the foundation of fault management, the quality of probestations deployment, a key role in network fault management system, affect probingefficiency, monitoring capability, and deployment cost.Based on the analysis of the research progress on fault management, thisdissertation has done the following works.First, combined with weakly connected node, a probe stations selection methodhas proposed based on k-path idea in the deterministic environment. Thecharacteristic of the weakly connected node has less independent paths to be find hasused to reduce the algorithm execution time. In addition, we reduce the total numberof the probe stations and solve the problem that the fault of the nodes with degree lessthan k can’t locate and monitor by limiting the probe stations have a lower probabilityof failture than the ordinary nodes. The experiment based on different scale networktopology has provided to verify the performance of the method. The experimentalresults show that the method has strong ability of fault location and cost-effectivelyand time-savingly.Second, proposed a probability routing method to simulate the dynamic routingenvironment in the deterministic environment, to specify the probability of next hopfor every node in the network and eventually determined the active probe routinginformation with probability. Then to find the k independent paths for every node inthe network and get the probe stations sets. The experiment based on different scalenetwork topology has provided to verify the performance of the method. Theexperimental results show that the method has strong ability of fault location andcost-effectively.Third, use bayesian network model to build the probability dependencyrelationship between network fault and its symptoms in the non-deterministicenvironment, and use binary symmetric channel to dispose the noise. To reduce probecost with the premise of high monitoring capability, this paper explores the probe staion deployment methods in the probability and noise environment. Based ondifferent scale network topology, with noise considerations and use fault propagationmodel to verify the performance of the methods. The experimental results show thatthe method has a strong ability of fault monitoring with fewer probes.