Identifying Root Cause via Playback Metrics
summary
Playback metrics identify root cause fast when backed by concrete signals and tools: an in-house playback metrics pipeline plus session-level tooling lets engineers go from anomaly alert to the offending embed in about three minutes, reveal an infinite logging loop, and fix it; multi‑CDN and live/VOD/HLS/progressive breakdowns expose a CDN edge drop and segment availability faults driving rebuffering; anomaly detection and scored sessions surface issues but require calibration to avoid false positives and misreporting.
AI-written summary · about
chapters (14)
- 0:00Intro
- 2:16Overview
- 4:20Standards
- 7:07Platform Unification
- 8:36Vimeo Playback Metrics
- 11:13Problems with Metrics
- 14:01Typical Flow
- 15:20Example
- 18:55Product Impact
- 21:42Time
- 26:55Pros and Cons
- 28:53Rebuffering and Live
- 29:58Summary
- 31:59Next Steps
related talks
33mThe Challenges of Testing Video Playback at Scale
26mLogging, and Errors, and Metrics Oh My!
29mSex, Lies & Videotapes: The Truth Behind Your CDN Logs
14mAnalyzing Video Metrics like Richard Feynman
46mWhy People Really Abandon Playbacks
17mMetrics Driven Obsession with Viewer Experience
18mDebunking the Myth of Video Delivery Monitoring on End-User Devices
20mDistributed Request Tracing for the Streaming Video Ecosystem