Randomized trial demonstrates enhanced anomaly detection in microservices, suggesting improved system reliability.
Key Points
This research aims to develop a novel method for detecting anomalies and identifying their causes in microservice architectures using advanced data integration techniques.
Proposed GRAFlow method combines logs, performance metrics, and traces through cross-modal feature fusion.
Utilized a multi-scale graph attention network to model dependencies among microservices.
Evaluated on two datasets, TrainTicket and SocialNetwork, to assess effectiveness.
GRAFlow outperformed existing methods in accuracy and F1 score.
Achieved significantly higher HR@K and NDCG@K metrics compared to state-of-the-art techniques.
Demonstrated robustness in complex system scenarios, enhancing reliability.