We are presently running a Suite of Cases which takes around 17 hrs to run. We cant run it overnight. So, is it advisable to divide the suite into 3 parts and then by using the Client-Server Technique, test each part on different machine through one machine on which Silk is loaded. Theoratically, I can divide the work load of one machine into 3 machines.
Is there any limitation of doing this ?
spawn and rendezvous are the way you do this... concurrency testing examples in help section are ok. I've had bad luck with this process though... It seems to be a bit buggy and overly complicated for something that could be done much easier by splitting up the test plan and running it in sections on different pc's. That way you dont have to manage all those threads. If one machine is down or one agent isn't started it becomes a problem quickly.
To be quite honest with you, after trying this concurrency testing for several weeks, I abandoned it entirely and just invoke silktest using a command line call to a plan file that's scheduled using windows scheduler. This is reliable and I can control the test distribution better than relying on a somewhat buggy concurrency test.