Computational models of biology typically incorporate a number of parameters that can
significantly impact the model's behaviour.  Due to both real and computational time
limits, the values for these parameters are often determined manually through ad hoc
trial and error.  There are two good reasons for exploring parameter space more systematically.
First, optimal values may lie outside the range tested.  Second, there may be interesting
interactions between parameters that can provide insight into both the model's behaviour
and the underlying biological system.

This paper describes Dsweep, an easy to use and practical method for visually comparing
pairs of parameters in arbitrary models, without a priori knowledge of the expected output.
Since each experiment is an independent work unit, parameter sweeps are an ``embarrassingly
parallel'' problem, and simulations are sent to worker machines to dramatically improve
performance.  The executables generated by Dsweep are completely self-contained, so no
configuration of worker machines is required.  The benefits of this method of parameter
analysis are illustrated with a promoter model of Bacillus subtilis for transcription start
site prediction.