Agentic Generative AI Framework for Predictive Fault Detection in Self-Healing Cloud Environments

Authors: Ripusoodan Sharma, Dr. Kriti Jain

DOI Link: https://doi.org/10.22214/ijraset.2026.78761

Abstract

The increasing complexity of cloud-native systems has made fault detection and recovery a critical challenge for modern IT operations. Traditional AIOps approaches rely on static rules and data-driven models, which often lack adaptability in dynamic and large-scale environments. To address these limitations, this paper proposes the ARCH (Autonomous Reasoning and Contextual Healing) framework, an intelligent self-healing architecture that integrates Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) for context-aware fault detection and autonomous remediation. The proposed framework adopts a layered design consisting of perception, cognition, knowledge, and action components, enabling continuous monitoring, intelligent reasoning, and automated recovery. By leveraging agentic reasoning strategies such as Chain-of-Thought and action-oriented decision-making, the system analyzes telemetry data and identifies root causes with minimal human intervention. The integration of RAG enhances contextual awareness by incorporating historical incident knowledge, thereby improving diagnosis accuracy and reliability. In addition, the framework supports predictive fault detection by utilizing historical telemetry patterns to anticipate potential failures. The performance of the ARCH framework is evaluated using key metrics, including Mean Time to Repair (MTTR), Autonomous Success Rate (ASR), and system efficiency. Experimental results demonstrate that the proposed approach achieves up to 82% reduction in MTTR and 89.5% autonomous success rate compared to baseline approaches. These results highlight the effectiveness of LLM-driven agentic architectures in enabling scalable, intelligent, and autonomous self-healing cloud systems.

Introduction

The ARCH (Autonomous Reasoning and Contextual Healing) framework is a novel AI-driven system for intelligent cloud self-healing. It integrates Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) in a layered architecture—Perception, Cognition, Knowledge, and Action—to enable predictive and reactive fault detection, context-aware reasoning, and autonomous remediation. Experimental evaluation shows that ARCH significantly reduces Mean Time to Repair (MTTR), improves Autonomous Success Rate (ASR), and enhances overall cloud system efficiency compared to traditional AIOps and deep learning approaches.

Conclusion

In this paper, an intelligent and autonomous self-healing framework, ARCH (Autonomous Reasoning and Contextual Healing), has been proposed to address the challenges of fault detection and remediation in modern cloud environments. The framework integrates Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) to enable context-aware reasoning, adaptive decision-making, and automated recovery. The proposed architecture adopts a layered design consisting of perception, cognition, knowledge, and action components, supported by a closed-loop feedback mechanism. This design allows continuous monitoring, accurate fault diagnosis, and efficient remediation with minimal human intervention. In addition, the incorporation of agentic reasoning strategies enhances the system’s ability to handle complex and dynamic failure scenarios. Experimental evaluation demonstrates that the ARCH framework significantly improves system performance, achieving notable reductions in Mean Time to Repair (MTTR) and higher Autonomous Success Rate (ASR) compared to traditional approaches. The integration of predictive capabilities further strengthens system resilience by enabling proactive fault detection based on historical telemetry patterns. Overall, the proposed approach highlights the potential of combining LLM-based reasoning with RAG-driven knowledge retrieval for building scalable and intelligent cloud self-healing systems. The results indicate that the ARCH framework can serve as a foundation for next-generation autonomous cloud operations. Future work will focus on improving real-time performance, reducing computational overhead, and enhancing security mechanisms to ensure safe and reliable deployment in production environments.

References

[1] A.-M. Tanas?, S.-V. Oprea, and A. Bâra, “Designing an Architecture of a Multi-Agentic AI-Powered Virtual Assistant Using LLMs and RAG,” Electronics, vol. 15, no. 2, p. 334, 2026. [2] K. Parthasarathy et al., “Engineering LLM Powered Multi-agent Framework for Autonomous CloudOps,” arXiv preprint arXiv:2501.08243, 2025. [3] Y. Chen et al., “AIOPSLAB: A Holistic Framework to Evaluate AI Agents for Enabling Autonomous Clouds,” Microsoft Research, 2024. [4] S. Author, “Building AI Agents for Autonomous Clouds: Challenges and Design Principles,” arXiv preprint arXiv:2407.12165, 2024. [5] Z. Zhang, “Foundations of GenAI Orchestration: RAG, MLOps, and LLMOps,” Washington University, 2024. [6] H. Liu et al., “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” in Advances in Neural Information Processing Systems, 2024. [7] J. Wei et al., “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models,” arXiv preprint arXiv:2201.11903, 2022. [8] T. Schick et al., “Toolformer: Language Models Can Teach Themselves to Use Tools,” arXiv preprint arXiv:2302.04761, 2023. [9] S. Yao et al., “ReAct: Synergizing Reasoning and Acting in Language Models,” arXiv preprint arXiv:2210.03629, 2023. [10] OpenAI, “GPT-4 Technical Report,” arXiv preprint arXiv:2303.08774, 2023. [11] M. Chen et al., “Evaluating Large Language Models Trained on Code,” arXiv preprint arXiv:2107.03374, 2021. [12] S. Zhang et al., “A Survey on AIOps: Intelligent IT Operations Using Machine Learning,” IEEE Access, vol. 10, pp. 112345–112360, 2022. [13] X. Xu et al., “AI-Based Fault Detection and Self-Healing in Cloud Systems,” IEEE Transactions on Cloud Computing, vol. 11, no. 3, pp. 1456–1468, 2023. [14] L. Wang et al., “Autonomous Cloud Management Using Reinforcement Learning,” IEEE Cloud Computing, vol. 9, no. 2, pp. 34–43, 2022. [15] P. Chen et al., “Machine Learning for IT Operations (AIOps): A Review,” ACM Computing Surveys, vol. 53, no. 5, pp. 1–35, 2021.

Copyright

Copyright © 2026 Ripusoodan Sharma, Dr. Kriti Jain. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Download Paper

Paper Id : IJRASET78761

Publish Date : 2026-03-25

ISSN : 2321-9653

Publisher Name : IJRASET

DOI Link : Click Here