Question

Hypotetical use case of an interaction:

• client M log in to the server;
• client M connects to the server and create an object OBJ;
• client P1 log in to the server;
• client P1 connects to the server passing OBJ.id and other parameters;
• client P2, P3… do the same;
• client M connects to the server and destroy object OBJ;

Each client need to store session cookies.

How can I can replicate that interaction pattern many times (even concurrenty) and measure latency and failed requests?

Answer 1

We would do that one of two ways:

1. Create a scenario that contains all of the listed actions, including logins and logouts. It would be a long scenario, but would allow passing of any required object identifiers for OBJ from M to P1, P2, P3 and then back to M to destroy OBJ. The downside is that the scenario is long and complex. The upside is that it is fully self-contained.
2. Other option is to pre-populate the database with 2 sets of OBJs for P1…P3 to work on and for M to delete. During the test, there would be 3 scenarios. Scenario 1 would be M creating more objects. Scenario 2 would be P1…P3 operating on objects previously pre-populated (set 1). The 3rd scenario would be M deleting objects from set 2. This allows the three scenarios to work independently. The upside is that the scenarios are smaller and simpler. The downside is that data will need to be pre-populated.

All of that can be done fairly easily in the better load testing tools (I’ll plug our product, Load Tester).

Hope that helps!