Anomaly Detection with Apache Pinot & ThirdEye
summary
A concrete pipeline that catches real incidents: anomaly detection runs on Apache Pinot queried by ThirdEye, ingesting events from Kafka (Telegraf in the demo) on a k3s cluster, with detector templates (threshold, mean/variance, Holt‑Winters) and root‑cause heatmaps. Claims include handling 1 million events per second workloads, delivering sub second queries for OLAP lookups, reducing detection from twelve hours (example loss of ~$10k) down to one minute detection and interactive RCA.
AI-written summary · about
chapters (13)
- 0:00Anomaly detection intro
- 2:48What is anomaly detection
- 5:23Black hole problem
- 7:34Pinot real-time analytics
- 10:23Pinot vs competitors
- 12:57Detecting performance issues
- 14:41ThirdEye alerting templates
- 17:13ThirdEye root cause analysis
- 19:02Anomalies with alerting
- 20:30Handling special events
- 21:11Demo setup
- 22:20Live Pinot & ThirdEye demo
- 29:57Conclusion and key takeaways
related talks
49mIntroduction to Real-Time Analytics with Apache Pinot
34mReal-Time Analytics: Going Beyond Stream Processing with Apache Pinot
47mReal-time analytics for your users with Apache Pinot
44mUSER-FACING, REALTIME ANALYTICS USING APACHE PINOT
54mAn Introduction to Apache Pinot
47mReal-time Analytics for Your Users with Apache Pinot
46mOptimizing Speed and Scale of User-Facing Analytics w Kafka and Pinot
37mCan Apache Pinot replace your OLTP database?