Debunking the Myth of Video Delivery Monitoring on End-User Devices
summary
Claim: installing probes in customers' set‑top boxes rarely identifies responsibility and yields redundant identical data. The talk shows a probe that runs on Raspberry Pi class hardware, quantifies faults as roughly 15% per stage (encoder/packager/CDN/box) leaving 25% between packager and CDN, and gives a real case where zero-size audio chunks cause 20–30s dropouts until a caching optimization on origin is found. Monitoring belongs at ingest, transcoding, packaging, CDN and last mile, not just on devices.
AI-written summary · about
chapters (6)
- 0:00Video delivery monitoring intro
- 1:53Monitoring on set-top boxes fails
- 6:34Streaming data flow errors
- 10:52Real purpose of monitoring
- 12:41Proving delivery chain problems
- 15:57Customer assumptions and outcomes
related talks
17mMetrics Driven Obsession with Viewer Experience
10mMoving Live Video Quality Control from the Broadcast Facility to the Living Room
14mAnalyzing Video Metrics like Richard Feynman
29mSex, Lies & Videotapes: The Truth Behind Your CDN Logs
34mIdentifying Root Cause via Playback Metrics
37mA Bartender's Guide to Network Monitoring
13mSecurity Theater: Moving Beyond Performative Content Protection
29mGaining Insights from a Black Box System