Experience Building a Representative, Effective Model to Randomly Generate Valid Sequences of Web Page Visits for Load Testing

摘要

1. Have used table-driven program design to build load test scripts that randomly generate valid sequences of page visits 2. Combined compiler, simulation, and modeling techniques to make the resulting implementation flexible and extensible 3. Experience shows ease of extensibility as new pages added 4. Page visit pattern can be altered by modifying the transition probability matrix, without changing any code 5. Can use model to predict change in visit pattern when transition probabilities are modified, without changing any code 6. Math supports the method: augmenting the Markov chain model to make it irreducible and more easily solvable yields the same page visit pattern as the original reducible model 7. Experience with the method has been favourable.