The original project is a good starting point for a straight forward jmeter project but I wanted a bit more detail in my results and a bit more flexibility in the models that I can run.
I also want automated results analysis, particularly aimed at CI perf testing with pass fail criteria based on transaction 95th percentile limits.
Ultimately I would like LoadRunner like scenarios but in the short term I have compromised: I provide information to the jmeter script regarding the injector we are on and thus allow the script to handle input data segregation (for example splitting input data uniquely across all users). And I use scripting techniques to control ratios of calls to produce scenario based load models (various percentages of users following particular workflows)
I will outline all this below. I have had to follow some coding guidelines to enable script and controller synchronization - for example results files must follow the naming convention in the script and injector id is placed in a particular file on the injectors for the script to pick up from.