From Telemetry to Insight: Building AI-Powered Observability Pipelines with OpenSearch
summary
Shows how to stream logs, traces and metrics from an OpenTelemetry collector via the OTLP exporter into OpenSearch, enrich and time‑roll indexes, and run OpenSearch ML anomaly detectors to cut noisy alerts and lower MTTR. Demonstrates a working pipeline with buffering/DLQ, S3 for offline model workflows, and alerting to Slack/PagerDuty. Uses OpenSearch’s built‑in models (trained on live telemetry) to suppress predictable weekly CPU spikes and surface real anomalies for microservices, LLM token/GPU spikes, fraud and delivery‑time cases.
AI-written summary · about
chapters (10)
- 0:00AI observability intro
- 1:06Observability challenges & AI
- 4:03Observability fundamentals
- 6:12Real-time analytics pipelines
- 8:03Anomaly detection with OpenSearch
- 10:30Use cases in architecture
- 13:05Telemetry to intelligence
- 15:58OpenSearch dashboards & insights
- 19:30Microservices & LLM monitoring
- 20:16Q&A
related talks
42mObserve Smarter, Not Harder: Scaling AI-Powered Observability With O...
20mSeeing Is Believing: OpenTelemetry and OpenSearch in Next-Gen Observability
34mHow to Do Microservice Observability with OpenSearch
26mPlug-and-Play Observability: OTel Auto-Instrumentation for LLMs in OpenSearch
29mOpenTelemetry With OpenSearch: Use-Cases, Status and Roadmap
06mSponsored Keynote: OpenSearch - See Everything: Open Observability for Agentic AI
27mAlert Fatigue to Action: Observability as the Backbone of Practical AIOps with OpenSearch
21mKubernetes Under the Magnifying Glass: Observability With OpenSearch