The full chain load test in AliExpress Russia
summary
AliExpress Russia runs a full-chain load test that models QPS from daily peaks plus a campaign adjustment coefficient and separates read/write chains by DAU and per-datacenter QPS; data-prep pipeline generates realistic request files daily in about one hour, uploads them to Alibaba OSS, and executors run scenarios with a one-button click. Test traffic uses a special traffic tag parameter propagated through threads to distinguish synthetic from live traffic. Dashboards show ramped step increases and CPU/memory metrics. Real incidents: downstream calls spiked 4–6x after an endpoint refactor (fixed by reducing calls or scaling), and missing tags in threads lost requests (took two work days to root cause).
AI-written summary · about
chapters (8)
- 0:00Introduction and agenda
- 1:08Key terms and background
- 3:01Traffic model analysis
- 8:24Data preparation pipeline
- 10:59Load test script setup
- 13:39Monitoring and execution
- 15:52Classic problems and solutions
- 21:45Q&A and closing
related talks
23mAI-Driven Load Test Analysis: From Routine Toward Action
37mExploratory Load-Testing with AI Tooling - it's Time to Ramp Up!
31mUse Load Testing to Build a Proper Mental Model of Your Service
31mLoad Testing Kubernetes: How to Optimize Your Cluster Resource Allocation in Production [I]
46mChaos Experiments - Datacenter Stress Testing
44mSofts You Need to Run the Perfect Load Test
27mTesting Before You Scale & Making Friends While You Do It
01h 10mPerformance Testing Crash Course