Infinite improbability testing – execute all tests in parallel
summary
Claim: 100000 tests in 27 minutes achieved by distributing work across machines and balancing by past timings. The actual tool is Mesa open-source distributed test runner — single CLI, server + agents, portable .NET Core bundle, per-agent parallelism, retries for failing tests, result merging, and extensibility hooks. Measured speedups range 4–40× depending on test type. Practical notes cover test data pooling, agent cleanup, browser/emulator limits, and when native per-process parallelism is a better fit.
AI-written summary · about
chapters (17)
- 0:00Intro
- 2:52How Do People Run Their 'Heavy' Tests?
- 5:00When Do You Execute Your Tests?
- 5:40What is a Parallel Test Execution?
- 9:19The Bellatrix Story
- 11:54How Do We Use It?
- 17:20How Do You Run Your Tests in Parallel?
- 19:00Comparison
- 24:08Meissa Test Agent Mode
- 25:18Meissa Test Runner Mode
- 27:16Preparing Environment and Managing Data
- 29:22Keep The Clean State of Agents
- 30:06Video Recording
- 30:56Screenshots
- 31:40Port Already in Use
- 32:17Edge Driver Cannot Start Multiple Instances
- 34:54The Biggest Experiment
related talks
43mInfinite Improbability Testing- Execute All Tests in Parallel
33mFrom Hours to Minutes: Run Thousands of CI Tests in Just Minute
43mHyper-Parallelism - The Panacea of Automated Testing?
09mTest Time Execution! Why it Can't be Ignored?
31mPortable and Fast - How to implement a parallel test runner
31mTests Are What You Want, What You Really Really Want
54mGetting maximum mileage from a minimal number of tests
24mSupercharging CI: Test Suite Optimization for Speed, Stability, and Savings