Sometimes it is better to think about splitting the data rather than the execution itself - you
therefore can run the same code over different parameter values, or different subsets of the data.
Whether this is possible, or whether it gives the same answer as doing it all in one go will
depend on the job itself.
There are many many scenarios for this, so some clues ast to what field you are working in would help.
Condor works at its best when handling many independent tasks (such as Monte Carlo or parameter
sweep). It can be used for closer-coupled MPI-style parallel tasks, but often HPCs or smaller clusters
are more appropriate for that.