Benchmark program on multi-core system

we want to benchmark Natural on a Linux platform with multiple cores.

The idea is based on two parts.

A. There is a Natural program which simulates a “base workload” of N user sessions, where N=100, 200, 300, …
The program should be able to “fork” independent sessions, as many as the parameter N will tell.
It is expected that Linux will assign them to different cores.
Within each session, a “fragment” of our application will be called with given parameters.
This fragment will execute a representative chunk of Natural code and also do some heavy Adabas operations.
It does not matter what output is produced by that code since we do not want to render or otherwise use the results in any way.
We do not even want to time those operations.
In sum: we just want to give all available cores something to do, which we control, and possibly “stress” them up to their capacity limits.

B. In addition, there is session (N+1) in which a dialog program will be executed.
It will be doing some “normal” user operation and its timing behaviour will be recorded.

Can you please advise us how we can implement such a workload generator?
Thanks to all.

I guess the simplest process would be to use a shell script which takes the number of sessions as input
and then start the requested number of natural instances in BATCHMODE