Fault classification in the architecture of virtual machine using deep learning

Area of Science:

• Computer Science
• Network Engineering
• Machine Learning

Background:

• Network performance is critically dependent on fault probability and service availability.
• Frequent cluster network faults can lead to service detection and identification failures.
• Proactive fault detection and classification are essential to prevent system failures.

Purpose of the Study:

• To develop and evaluate a model for accurate fault classification in cluster networks.
• To predict potential failures in virtual machines to ensure service continuity.
• To improve the overall reliability and performance of network services.

Main Methods:

• A novel model incorporating feature selection, an attention transformer, and a feature transformer was developed.
• The model was designed to process tabular data using neural networks.
• Experimental analysis was conducted on a tabular dataset from the Telstra cluster network.

Main Results:

• The proposed model achieved high performance metrics: 98.3% accuracy, 98.3% precision, 97.4% recall, and 97.8% F1 score.
• The model effectively identified and classified fault occurrences, including service disruption events and connectivity interruptions.
• Trace-driven experiments validated the model's efficacy in predicting virtual machine failures.

Conclusions:

• The developed model demonstrates significant success in predicting network failures.
• The findings validate the research objective of failure occurrence prediction in virtual machines.
• The proposed approach offers a robust solution for enhancing network fault management and service availability.