Leveraging ML to Detect Application HotSpots [@scale, of Course!]
summary
LinkedIn builds a daily pipeline that trains per-service regression models in Python (caption notes a "salary framework") using 15 days of training data, stores models in MySQL, and runs a daemon that triggers detection after each real-traffic load test. Hotspots are service+endpoint regressions where observed latency or error exceeds model prediction; scores combine latency, error, page‑rank and chain‑grouping filters. The hotspot‑ranker surfaces problems early — top‑20 hotspot rank covers 90% of true hotspots — which reduces failed load tests and shrinks remediation scope.
AI-written summary · about
chapters (5)
related talks
39mMethodology of ML in Performance Testing (How to Find Bottlenecks Automatically)
14mPerformance Optimization in High-Volume Transaction Systems
26mProactive Failure Prediction Through Anomaly Detection
31mRunning Machine Learning Analytics On Traces
41mInfrastructure for ML Applications
30mReducing MTTR and False Escalations: Event...
26mML Monitoring & Observability
10mML-based Log Analysis for Faster Debugging