The three fundamental rules of performance testing
Violating any of these rules will make performance test results worthless for the prediction of production performance:
1. Use test hardware that resembles production
Performance is the result of dynamic interaction of systems with different roles and strengths. Having one component significantly different from production has ahigh risk of shifting performance characteristics! So rent or buy hardware similar to the production system for getting realistic behaviour!
2. Organize test data that resembles production data
Provide your system with data in a similar amount and variance of data as your production system has. If production has 20 million users, your test system must have the same number too! Use benerator∞
to extract and anonymize production data or generate synthetic data!
3. Create user load and concurrency with production characteristics
Simulate a realistic number of concurrent users and applications that. LoadRunner
, JMeter and Grinder can be used for this!