Digital Twin Driven Hybrid Lstm And Isolation Forest Framework For Predictive Failure Detection And Autonomous Self Healing In Servers

Abstract

Cloud Server Infrastructures Form The Foundation For Technical Services, Where The Services Will Be Highly Available, Critical For Business Continuity And User Satisfaction. Traditional Systems Notify The User After The Failures Had Occurred. It Is Completely Based On Threshold-based Alerts And Results In Performance-degradation, Downtime, Inefficient Resource Utilization And Delayed Recovery. To Address This Problem, Our Proposed Framework Twins The Cloud Server Digitally To Predict The Anomalies And Self-heal The Cloud Server Environments. The Framework Continuously Mirrors Telemetry Data Such As CPU Utilization, Memory Usage, Disk I/O And Network Throughput Into A Virtual Replica Of The Physical Infrastructure. A Hybrid Anomaly Detection Model That Combines Long Short-Term Memory (LSTM) Networks And Isolation Forest Is Used To Identify Early-stage Performance Degradation. The Proposed System Results In Reduced Downtime, Efficient Resource Utilization When Compared To Traditional Monitoring Techniques. The Framework Demonstrates The Feasibility Of Autonomous And Intelligent Cloud Infrastructure Management.

Keywords