AI-Driven Predictive Auto-Scaling for Cloud-Native Systems with Real-Time Anomaly Detection
Year: 2024
Type: article
Abstract: Cloud-native architectures demand highly dynamic resource scaling to handle fluctuating workloads efficiently. Traditional reactive scaling methods often lead to over-provisioning, under-utilization, or performance degradation. This paper introduces an AI-driven predictive auto-scaling framework that leverages machine learning-based observability data to anticipate resource demand proactively. By integrating real-time anomaly detection, this approach minimizes system failures due to unexpected surges or resource misallocations. Our proposed solution utilizes Long Short-Term Memory (LSTM) networks for predictive analytics and an unsupervised anomaly detection model to optimize AWS-based cloud infrastructures. Experimental results demonstrate improved cost efficiency, reduced latency, and enhanced system resilience, outperforming conventional auto-scaling mechanisms.
Author: Mahender Singh
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7216873745
