OrEdge: Lightweight AI spots system glitches in real time

A team of researchers from York University, Honeywell, and IBM has unveiled **OrEdge**, a novel lightweight framework designed to revolutionize real-time anomaly detection in distributed software systems. Unlike traditional approaches that rely on computationally expensive attention- and graph-based architectures, OrEdge leverages **orthogonal-domain temporal representations** to achieve high accuracy with minimal computational overhead. By jointly analyzing heterogeneous monitoring data—such as logs, metrics, and traces—OrEdge identifies abnormal software behavior while capturing temporal dependencies and reducing redundancy across observability signals.

At the heart of OrEdge is **OrEdgeCore**, a compact orthogonal-domain reconstruction module that captures recurring temporal patterns while suppressing transient variations. In rigorous evaluations on three real-world microservice datasets (MSDS, SN, and TT), OrEdge demonstrated competitive detection performance while reducing the reconstruction model size to just **9.6K parameters**—a fraction of the 20K to 143K parameters used by existing methods. This ultra-compact design enables efficient deployment on resource-constrained edge devices, such as Raspberry Pi, where OrEdge achieves **sub-second inference** and reduces inference latency by **over an order of magnitude** compared to conventional approaches.