Load testing distributed web services
summary
loveholidays serves 8,050 rps in peak January 2024 and hits 12.5k rps later while keeping p50 search latency ~35ms and p99 <200ms by running daily load tests in production that replays real production access logs using Ripley, an open‑source Go tool (https://github.com/loveholidays/ripley). Ripley preserves request timestamps, replays at multiples up to ~80x, required a wall‑clock drift fix and a repeatable harness, and proves fixes against real incidents without impacting customers.
AI-written summary · about
chapters (11)
- 0:00Introduction and scalability
- 3:18About Love Holidays
- 8:14Website speed and search challenges
- 12:26Why load testing matters
- 14:30System architecture overview
- 16:39Introducing Ripley tool
- 19:11Nightly production load tests
- 22:45Fixing Ripley performance
- 27:51Tips for load testing
- 32:54Outcomes and results
- 34:53Q&A session
related talks
52mLoad testing distributed web services
28mAccelerate with confidence through AWS Distributed Load Testing
33mHold My Beer - Load Testing. In Production. On Autopilot.
38mPerformance in the Wild
30mBeat the rush! Designing effective load tests for your web application
47mLoad Testing You Always Wanted
26mLoad testing websites following the steps of a user journey
26mIntroduction to Load Testing