Design of Microservice Anomaly Propagation Judgment and Root Cause Tracing Algorithm Based on Dynamic Graph Neural Network
Year: 2026
Type: conference-paper
Abstract: To address the challenge of cascading fault localization caused by high-frequency microservice topology evolution and fragmented multi-source operation and maintenance (O&M) data in power business middle platforms, this paper proposes a multi-modal dynamic graph neural network algorithm for anomaly root cause tracing. The algorithm constructs time-varying attribute graphs by fusing metric, log and link tracing data, and encodes topological evolution features via a temporal graph attention network. An edge-level classifier is deployed to identify anomaly propagation paths, and node-level root cause probabilities are calculated using a random walk with restart model. Experiments on a real desensitized dataset show that the algorithm achieves a Top-5 root cause localization accuracy of 86.4%, effectively improving fault troubleshooting efficiency in complex O&M scenarios.
Institution: State Grid Corporation of China (China)
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7220693255