Would very much like to see some of that “AIOps” funding directed towards unsupervised anomaly detection with seasonal decomposition for ops data. I’ve always found static alert threshold (and negative event) configuration to lag behind/drift from reality.
“It specializes in detecting anomalies in multi-dimensional time series data sets”
https://www.uber.com/en-IN/blog/uvitals-an-anomaly-detection-alerting-system/