Under the hood of fast traffic splitting for A/B testing
summary
Explains the internal design of a high-performance traffic splitting system for A/B testing, focusing on how to route users to experiments with low latency and high reliability. Covers technologies and techniques used to implement fast, scalable splitters in production (e.g., efficient hashing, routing logic, and integration with existing service stacks) and practical considerations for correctness and observability.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 4:49Service requirements
- 7:17Segment calculation optimization
- 13:19Garbage collector impact
- 15:23GIL issues and solutions
- 18:12Integration challenges
- 20:46Load balancing fixes
- 23:42Project lessons
- 25:08Q&A
related talks
39mDIY A/B Testing Architecture
40mReal-time A/B: how to get metrics for experiments
09mDemystifying the Dreaded A/B Test
33mA/B testing: performative experimentation at scale
19mA/B User Testing When Your CDN Cache Messes With Your Results
21mHandling Traffic Spikes Through Automated Prioritized Load Shedding
24mWe Raise Children: Or How Did Ivan Vasilievich Run the Experiments?
40mA Test Harness to Verify Traffic Router Performance