A Novel Fault Diagnosis and Recovery Mechanism Based on Events Prediction in Distributed Network

Abstract

AbstractIn current growth, the detection of faults in computer networks has fascinated wide attention due to its significance in network administration. Most existing problem-solving techniques depend on dynamic search methods that are quickly and accurately diagnose the occurrence of faults. However, these techniques undergo traffic constraints, particularly on highly scalable networks. Fault-tolerant distributed system components have to be able to perfectly identify which system elements are defective and which are normal. This paper intends to propose a novel fault diagnosis and recovery mechanism (FDRM) for distributed networks based on events logs identification at nodes in completely linked networks. The observation of events of a node can define the status of the node as standard or vice versa. The novel FDRM algorithm can achieve the diagnosis of execution events till the node remains in a certain condition for long enough to allow all required estimation to detect a node status. The experimental evaluation was performed considering a hierarchical evaluation strategy where all nodes are normal in the form of the hypercube. The simulation results have a significant reduction in average latency and false alarm rate with high detection accuracy and when evaluated to other related approaches to enhance the fault diagnosis.KeywordsFault diagnosisRecoveryEvents identificationDistributed network