Challenges and Experiences with MLOps for Performance Diagnostics in Hybrid-Cloud
summary
ML-based performance diagnostics for VMware hybrid-cloud builds a cloud-hosted pipeline that ingests anonymized telemetry via Apache Spark, trains unsupervised models (autoencoder, isolation forest, PCA, regression) and uses decision-tree rules for root cause and remediation suggestions. Validation uses manual labels, known-bug traces and synthetic perturbation datasets; feature sets come from domain experts while the pipeline auto-tests preprocessing chains (normalization, Box‑Cox, outlier filters) and hyperparameters. Deployed across thousands of clusters, it detects real regressions early, monitors detection-rate drift, and requires QA before model rollout.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:44Performance Troubleshooting for Hybrid-Cloud Deployments
- 4:23ML-Based Performance Diagnostics Service
- 6:08Building Performance Diagnostics Service
- 6:54VMware Analytics Cloud
- 8:18ML-Performance Issue Detection and RCA
- 10:59Ops-Feature Selection
- 15:39Application / Feedback
- 17:37Performance Drift Monitoring
- 21:23Handling New Performance Issues
related talks
33mBuilding & Managing MLOps Pipelines in the Cloud
13mMLOps at Scale
53mExperience with an end-to-end MLOps pipeline: Practical insights and 3D architectural perspectives
36mMLOps from the edge to on demand visualization
10mML-based Log Analysis for Faster Debugging
10mFrom Model to Production: Building Scalable MLOps Pipelines
25mBuilding Scalable MLOps Pipelines That Actually Work
05mThe Evolution of MLOps