The client computers (2) receive commands from the console computer that instruct when to start running test scripts and with what frequency to run them. Running the test scripts executes a predetermined sequence of actions within the client (browser) side of the PT application (see Using Test Scripts). Client computers vary the test workload only when they receive specific commands from the console computer — "simulate 15 users," "simulate 20 users," and so on.
All the data published for this test (see PT Application Performance Test Results) reflect traffic between the client computers (2) and the PT application server computer (3) or on the client computers themselves if a client-side interaction results in a transaction (see About Transactions).
Note Tool-software vendors corroborate the fact that increasing the number of computers in the client-computer test cluster gives more accurate results. Because these computers generate enough client-side actions to simulate hundreds of virtual users, vendors consider three computers (as used to test the PT application) to be the minimum size for such a cluster. Not only does a larger test cluster reduce the number of packets each computer must send, but it also lets the test architecture more closely approach the real-world situation of one user per thread.
For specific information about these client computers, see Client Computer Statistics (First Computer), Client Computer Statistics (Second Computer), and Client Computer Statistics (Third Computer).