Automated Root Cause Analysis with RAG/LLM
summary
Swisscom builds automated RCA from distributed traces into metrics (calls, errors, P95) with two months retention and a machine learning forecast over about 300,000 time series to flag anomalies and compute health scores that propagate from functions to components to ~100 top business processes across 1,300 resources. Detection hits near 30 seconds, and root cause ranking uses graph similarity plus RAG/LLM to surface past incidents and investigator comments (no model training, vector retrieval + LM summary).
AI-written summary · about
chapters (6)
- 0:00Swisscom monitoring platform
- 4:37Monitoring business processes
- 12:18Incident lifecycle and RCA
- 18:20Graph similarity for issues
- 23:17RAG/LLMs for automated RCA
- 27:08Q&A
related talks
32mGenerative AI for Observability: Automating Root Cause Analysis in Modern IT Systems
07mAccelerating RCA with AI
28mBeyond Dashboards: Architecting AI Agents for Autonomous Observability
46mLeveraging AI and Automation in DevOps
27mAI-Powered Root Cause Analysis at Scale: From Theory To Production...
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
40mWhen Incidents Fix Themselves: AI SRE in action
31mRAG at Scale: Logging, Traceability, and the Architecture for Control