Reducing MTTR and False Escalations: Event...
summary
LinkedIn builds an Event Correlation Engine that maps dependencies for 900 applications and 2,200 internal APIs into a call graph in a graph database, and runs two engines: a metrics engine using k-means clustering (three clusters, predict + trend + week‑over‑week scores) and an alert engine that polls alerts every minute, weights call count/alert count/deviation, and a recommendations layer that returns the responsible service and confidence. The pipeline feeds automation that avoids false escalations and reports examples like 91% confidence for escalations, cutting MTTR.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:47False Escalations
- 2:23Problem Statement
- 3:49Project Goals
- 5:30Architecture Considerations
- 8:27Platform Overview
- 10:21Correlation Engine Overview
- 11:48Learning Curve
- 13:25How do we map service
- 18:15K-Means: How it works
- 18:54Predict score
- 19:55Typical Workflow
- 21:32in Visualize Assumptions
- 26:58Recommendation Engine
- 27:47Ecosystem Integration
- 28:44Key Takeaways
related talks
46mEvent Correlation: A Fresh Approach towards Reducing MTTR
37mWeathering the Storm: How Early Warnings Save the Farm
20mSpike Detection in Alert Correlation
42mMachine Learning In Observability Systems: From Insights To Action
19mFrom Alerts to Action: AI-Powered Incident Response Systems
19mLeveraging Analytics for Technical Efficiency and Enhanced User Experience
10mHow ML-Driven Slack Integration Cut MTTR by 88.5%
15mIntelligent Incident Management in DevOps