Decisions & considerations in building your WebRTC app

on YouTube ↗21m

" summary

Twilio runs a global WebRTC service (750,000 devs, 50B API interactions) and treats failures as constant: tens of service alerts weekly but 100% uptime for two quarters and target 99.95% for the year. Production tooling includes lightweight Node signalling E2E tests plus headless Chrome/Firefox “robocallers” every ~5 minutes, Rollbar for aggregation, HipChat bots to trigger suites, a Raspberry Pi Network Throttler using netem, and getStats → Kinesis → Redshift for realtime and historical call-quality analysis. Common observable failure is one-way audio; mitigation includes balancer/gateway rate limiting and devs operating their own services.

AI-written summary · about

" chapters (8)

1. 0:00Intro
2. 0:54Scaling complexity
3. 2:34Things break
4. 3:52Test constantly
5. 8:29Plan for failure
6. 16:45Operational excellence
7. 18:10Communication tools
8. 18:30Questions

" related talks

• 33mWebRTC Change Communcations ForeverGreg Baugues · RailsConf 2014
• 24mMaking WebRTC awesomeHenrik Joreteg · CascadiaJS 2013
• 38mReal-time Communication on MobileThomas Gorissen · Meta Refresh 2015
• 19mDebugging Tools and Techniques for WebRTCPhilipp Hancke · Kranky Geek WebRTC Brazil 2016
• 27mWebRTC: Real-Time CommunicationsYlia Kasianenko · OdessaJS 2015
• 33mWebRTC Change Communications ForeverGreg Baugues · RailsConf 2014
• 30mHow to Build an End-to-End WebRTC ServiceDaniel Petersson · Kranky Geek WebRTC Brazil 2016
• 36mWebRTC Use CasesTsahi Levent-Levi · Kranky Geek WebRTC Brazil 2016