Accuracy Estimation for Fault Classification in Virtual Machine using Deep Learning

A large-scale cluster network needs to be provided with a low probability of failure occurrence along with high service reliability and availability. Frequent faults are experienced in the cluster network, which results in job or task failure. Thus there is a need to detect and classify the fault in the High performance cloud (HPC). Therefore, the fault needs to be detected and classified in the HPC before failure occurs. This study proposes a new model for fault classification of the virtual machine in the HPC based on deep learning model. This research work is motivated by various existing fault prediction and classification techniques which uses different using machine learning approaches for fault classification. As the TabNet architecture deals with tabular data, the objective of the proposed model is to classify the fault in the HPC using deep learning based TabNet architecture. The tabular data from Los Alamos National Laboratory (LANL) is used to carried the experimental analysis in HPC. This data set includes the failure records of nodes in the HPC. The result obtained in terms of average accuracy advocates the strength of the proposed model. The Logistic Regression, Random Forest, Decision Tree, AdaBoost, Multi-Layer Perceptron, Gaussian Naive Bayes, K-Neighbors, Support Vector Classifier, Voting Classifier are taken into consideration for comparison purpose. The accuracy obtained by these model varied from 61% to 77%. The experiment demonstrated the effectiveness of the proposed model with the accuracy obtained more than 98%.