Cloud-Based AI Systems: Leveraging Large Language Models for Intelligent Fault Detection and Autonomous Self-Healing

TL;DR

The paper addresses scalable, real-time fault detection and autonomous self-healing in cloud environments by proposing a multi-level framework that fuses ML-based fault detection with Large Language Models to interpret logs and streams. It uses time-series embeddings from LSTMs and text embeddings from BERT, integrates supervised $L_{SVM}$ with autoencoder-based anomaly detection and variational autoencoders, and adds reinforcement-learning-driven self-healing with a joint objective $L_{total} = L_{SVM} + L_{AE} + L_{VAE} + L_{DNN} + L_{RL}$. Experimental results on the VT SDK IoT dataset show the approach achieving about 92% accuracy and faster recovery than Transformer-, Graph Neural Network-, Deep Reinforcement Learning-, and BERT-based baselines, along with higher stability. The work demonstrates a practical, scalable path for autonomous cloud fault management, and suggests extending to more complex, real-time, and heterogeneous cloud environments.

Abstract

With the rapid development of cloud computing systems and the increasing complexity of their infrastructure, intelligent mechanisms to detect and mitigate failures in real time are becoming increasingly important. Traditional methods of failure detection are often difficult to cope with the scale and dynamics of modern cloud environments. In this study, we propose a novel AI framework based on Massive Language Model (LLM) for intelligent fault detection and self-healing mechanisms in cloud systems. The model combines existing machine learning fault detection algorithms with LLM's natural language understanding capabilities to process and parse system logs, error reports, and real-time data streams through semantic context. The method adopts a multi-level architecture, combined with supervised learning for fault classification and unsupervised learning for anomaly detection, so that the system can predict potential failures before they occur and automatically trigger the self-healing mechanism. Experimental results show that the proposed model is significantly better than the traditional fault detection system in terms of fault detection accuracy, system downtime reduction and recovery speed.

Figures (3)

• Figure 1: Accuracy Comparison of Fault Detection Methods.
• Figure 2: Fault Recovery Time Comparison with Model Complexity.
• Figure 3: System Stability Comparison