AI-DRIVEN DYNAMIC DEPENDENCY GRAPH GENERATION FOR PREDICTIVE OBSERVABILITY IN DISTRIBUTED SYSTEMS
Anusha Reddy Narapureddy · INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & TECHNOLOGY · 2025
Modern distributed systems present intricate interdependencies that challenge traditional observability tools.This paper introduces an AI-driven approach to dynamic dependency graph generation, leveraging time-series telemetry data and graph neural networks (GNNs) to model complex system interactions.By dynamically mapping dependencies in real-time, this technique identifies latent bottlenecks, predicts cascading failures, and offers actionable insights for proactive system management.The proposed solution includes a pipeline for telemetry preprocessing, feature extraction, and dependency inference, validated against real-world distributed architectures.Experimental results demonstrate improved accuracy in dependency mapping and significant reductions in mean time to resolution (MTTR) for incident management.This study highlights the transformative potential of AI in advancing predictive observability and operational resilience.