Anomaly Detection in Streaming Logs: Building Reliable Alerting Pipelines at Scale

As distributed systems continue to grow, spotting unusual patterns in log streams becomes essential for maintaining reliable services. This workshop teaches practical techniques for detecting anomalies in streaming log data, drawn from real-world examples and designed for large-scale systems. Participants will build a real anomaly detection pipeline step by step: rule-based log parsing, sliding-window features, and a range of statistical and machine learning methods for scoring events on a data stream. We will also address some of the hardest parts of running anomaly detection in practice, including threshold tuning, concept drift, and keeping alert volume manageable. By the end of the session, attendees will have a reusable pipeline and clear criteria for choosing methods that hold up at scale.