Gandalf: An Intelligent, End To End Analytics Service for Safe Deployment in Large Scale
summary
Gandalf is a Microsoft Azure analytics service that monitors diverse fault signals during software rollouts and correlates them with all concurrent deployments using spatial and temporal correlation. It combines an ensemble ranking algorithm to identify likely culpable rollouts and a binary classifier to assess impact, runs on a lambda architecture for real-time and long-term monitoring, and drives automated stop-or-proceed decisions; deployed in production with reported precision and recall metrics for data-plane and control-plane rollouts.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:22Azure Architecture
- 1:56Azure Compute Architecture
- 3:19Existing Practice of Sale Deployment: Pre-qualification Test and Safe Deployment Policy
- 5:17Design Considerations
- 11:23Spatial Correlation: Which Components & Faults Are Co-Located
- 18:55Why Safe Deployment is Challenging?
related talks
16mLeveraging Azure AI Foundry to Build Intelligent Release Agents
25mObservability In ArgoCD/Rollouts Using Streaming ML For Reducing MTTR
13mAchieving Seamless O11y Across Cloud Providers
16mEvolution of Aegis: Fault Diagnosis for AI Model Training Service in Production
22mAnalysis in Action: Real-World Metrics Driven Argo Roll...
33mAutomating Data Analytics: Collect, Transform and Deploy - Lead With Change
14mPlatform Engineering with Self-Healing AWS Infra
26mImproving fault detection & real time detailed analytics with Telemetry