Use operational signals (metrics trends, event patterns, key access distributions) to support anomaly identification and faster debugging. Extract features from time-series data that can surface unusual patterns and suggest possible root causes.
This differs from threshold-based alerts : alerts fire when a known metric crosses a predefined line ("memory > 80%"). Anomaly detection identifies abnormal patterns without predefined rules, i.e. it finds things you didn't think to set a threshold for ("memory usage pattern looks unusual compared to the last 7 days")
Use operational signals (metrics trends, event patterns, key access distributions) to support anomaly identification and faster debugging. Extract features from time-series data that can surface unusual patterns and suggest possible root causes.
This differs from threshold-based alerts : alerts fire when a known metric crosses a predefined line ("memory > 80%"). Anomaly detection identifies abnormal patterns without predefined rules, i.e. it finds things you didn't think to set a threshold for ("memory usage pattern looks unusual compared to the last 7 days")