Scalable Temporal Anomaly Causality Discovery in Large Systems: Achieving Computational Efficiency with Binary Anomaly Flag Data
Researchers have developed a new approach called AnomalyCD to efficiently discover causal relationships in large-scale systems using temporal binary anomaly data. This method addresses the significant computational burden of traditional causal discovery techniques, making them more applicable to real-time and large-scale deployments. AnomalyCD incorporates strategies like anomaly data-aware causality testing and data compression to reduce computational overhead while maintaining or improving accuracy, as validated on datasets from CERN and an IT monitoring system. AI
IMPACT This research offers a more computationally efficient method for identifying root causes of system failures, potentially improving diagnostic capabilities in large-scale monitoring.