Structuring test designs and prioritizing your test effort for large and complex software systems are daunting tasks, ones that have beaten many, very good test engineers. If you add concurrency issues and a distributed system architecture to the mix, some would simply throw up their hands. At Microsoft, where Keith Stobie plies his trade, that is not an option. Keith and others have reengineered their testing, employing dependency analysis for test design, model property static checking, "all pairs" configuration testing, robust unit testing, and more. They employ coverage to successfully help select and prioritize tests and make effective use of random testing including fuzz testing security. Finally, models of their systems help them generate good stochastic tests and act as test oracles for automation.
- Test checklists for large, complex, distributed systems
- Appropriate use of code coverage with increased coverage for less effort
- Dependency analysis for improved test designs