Traffic Steering using Rum DNS @ LinkedIn
summary
LinkedIn describes using real user monitoring (RUM) from member browsers combined with triple-vendor DNS and Big Data to steer traffic to the nearest POP or edge in real time. The system uses geolocation to direct members (example: Mumbai users to a Mumbai POP) and automatically stops resolving to unreachable POPs within seconds based on browser-reported last-mile failures, with a multi-vendor approach for redundancy and tools for cross-vendor consistency, fail-out, global monitoring and status boards.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:51Problem Statement
- 2:42Multi CDN Strategy
- 4:40POP (Point of Presence)
- 10:55Traffic Steering
- 21:01POP RUM Steering
- 31:09DNS Tool
- 31:59CDN Regression tool
- 36:33Closing thoughts
related talks
53mReal User Performance Monitoring at Netflix scale
05mHow real is real-user measurement?
37mA Bartender's Guide to Network Monitoring
29mUsing the Internet as Your Load-Balancer
29mDNS: Old Solution for Modern Problems
46mNoise Canceling RUM
23mAutomatic Traffic Scheduling for Internet Connectivity Failures
39mLoad Balancing Building Blocks