The GeekNarrator Testing Distributed Systems the right way ft. Will Wilson
Jul 19, 2024
Will Wilson, Engineer and co-founder of Antithesis, dives deep into the world of deterministic simulation testing for distributed systems. He breaks down the limitations of traditional methods, showcasing how his company's approach improves software reliability. Key discussions include optimizing bug detection strategies, the significance of simulated workloads, and the challenges posed by third-party APIs. Real-world examples like chat applications illustrate how effective testing can reveal hidden issues, making this an essential listen for tech enthusiasts and developers.
AI Snips
Chapters
Transcript
Episode notes
Limitations of Conventional Testing
- Conventional testing methods struggle to uncover unknown issues because tests are written for anticipated scenarios.
- Distributed systems make this worse; their complexities and timing dependencies introduce fragility.
Fragile Test Leading to Production Outage
- A fragile test correctly flagged a problematic change, but was dismissed due to its history of false positives.
- Ignoring the test led to a production outage, highlighting the danger of flaky tests.
Deterministic Simulation Testing
- Deterministic simulation testing shifts from specific input-output checks to validating general properties, aligning with fuzzing and property-based testing.
- It virtualizes non-deterministic parts of a system for controlled testing, making randomized testing possible in distributed systems.