Graph Neural Networks with Causal Inference for Root Cause Analysis in Microservice Architectures
International Journal of Computational and Biological Sciences · 07.12.2025 · 2 (02) · 202-220
DOI: 10.66238/ijcbs26 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computational and Biological Sciences, online erschienen am 07.12.2025. Verfasst von Mateo Rinaldi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Microservice architectures have revolutionized modern cloud computing by enabling scalable, flexible, and independently deployable services. However, the distributed nature and complex interdependencies of these systems pose significant challenges for root cause analysis when performance anomalies occur. Traditional monitoring approaches often fail to distinguish genuine causal relationships from spurious correlations, leading to prolonged diagnosis times and incorrect remediation actions. This paper proposes an integrated framework that combines graph neural networks with causal inference techniques to address these challenges systematically. Our approach operates through a hierarchical architecture that performs both offline causality graph construction and online root cause inference. The offline component employs change point detection and causal discovery algorithms to build multi-level causality graphs capturing both service-level dependencies and metric-level causal relationships. The online component leverages these pre-constructed graphs along with graph neural network embeddings to perform real-time root cause localization when service level objective violations are detected. We introduce a novel graph pooling and unpooling mechanism within our GNN architecture that enables hierarchical reasoning across different granularities of system abstraction. Experimental evaluation on benchmark microservice applications demonstrates that our framework achieves superior diagnosis accuracy with precision exceeding eighty-five percent while significantly reducing diagnosis latency compared to correlation-based and pure machine learning approaches.
Zitieren
APA
Rinaldi, M., & Sørensen, A. (2025). Graph Neural Networks with Causal Inference for Root Cause Analysis in Microservice Architectures. International Journal of Computational and Biological Sciences, 2(02), 202-220. https://doi.org/10.66238/ijcbs26
Vancouver
Rinaldi M, Sørensen A. Graph Neural Networks with Causal Inference for Root Cause Analysis in Microservice Architectures. International Journal of Computational and Biological Sciences. 2025;2(02):202-220. doi:10.66238/ijcbs26
BibTeX
@article{Rinaldi2025,
author = {Rinaldi, Mateo and Sørensen, Anika},
title = {Graph Neural Networks with Causal Inference for Root Cause Analysis in Microservice Architectures},
journal = {International Journal of Computational and Biological Sciences},
year = {2025},
volume = {2},
number = {02},
pages = {202--220},
doi = {10.66238/ijcbs26}
}
EUVIMED