Sex, Lies & Videotapes: The Truth Behind Your CDN Logs
summary
Claims that 95% of errors originate in the CDN and that routine 4:00 a.m. maintenance spikes explain odd overnight failures. Recommends multi‑CDN + redundancy, establish alarm levels and separate notification channels, prefer real‑user monitoring if budget is limited, and run synthetic stress tests against providers and production. Notes CMCD headers carry per‑chunk telemetry, cites a 4.1M concurrent viewers switch that succeeds unnoticed by users when multi‑CDN failover is in place.
AI-written summary · about
chapters (16)
- 0:00Introduction to CDN logs
- 1:28What is a CDN
- 2:08Common CDN industry issues
- 3:29Hourly and daily patterns
- 4:34The 4 a.m. dark hour
- 6:20Distance and ISP truths
- 8:53First five CDN commandments
- 12:25Test everything thoroughly
- 13:46Next five CDN commandments
- 16:14Where CDN errors originate
- 17:58CMCD protocol insights
- 19:26Fixing errors without downtime
- 20:31Real-life use cases
- 23:19Conclusion and industry truths
- 24:10Q&A: Small vs large content scaling
- 27:22Q&A: Testing external CDNs
related talks
09mCDN Logs: Not just for breakfast anymore
23mContent delivery sucks, and what you can do about it
37mA Bartender's Guide to Network Monitoring
44mA System for Collecting Detailed Statistics on CDN Node Performance
28mReliability Lessons from Running High Traffic Services for More Than 10 Years
16mCDN Challenges of HTTP-based Low Latency Live Streaming Delivery
33mThe Challenges of Testing Video Playback at Scale
53mCDNs 101: An Introduction to Content Delivery Networks