Fine-grained fault diagnosis of microservices based on Hidden Markov Model and Granger

Fault diagnosis in microservice systems is increasingly challenging due to the complexity of service dependencies and the dynamic nature of runtime environments. Traditional techniques often struggle to accurately locate root causes and capture the paths of fault propagation across layers. To address this, we propose a fine-grained diagnosis approach that combines Hidden Markov Models (HMM) for modeling latent service states with Granger causality analysis for uncovering inter-service causal relationships. Our method enables both precise root cause localization and reconstruction of fault propagation paths. Extensive experiments in controlled and real-world environments show that the proposed framework achieves $91 \%$ accuracy, 88% precision, 87% recall, 87% F1-score, and an average diagnosis time of 87 seconds. These results demonstrate that our approach significantly improves the accuracy and efficiency of fault diagnosis in complex microservice architectures.