David Mark's suggestion

This should not be difficult to do if you have adequate information
about the performance of the GIS. Dan Wagner (Univ. of South Carolina)
and I are working on the design and development of special test sets
which will permit measurement of performance to take place over the
entire performance envelope of the GIS. Based upon these data points,
we hope that valid statistical estimation can take place.

The only system I know of that did this kind of "up-front" estimation
was an urban planning system in Sweden some years ago. There it was a
matter of $$ since the interactive portion of the GIS merely created
a batch file which was subsequently launched. The planner's objected
to the original version of the system which generated bills of unknown
size. Alas, when realistic estimates of the costs of a run were
generated the most common reaction proved to be cancellation of the
request!

--
Duane F. Marble
Department of Geography E-Mail: dmarble@magnus.acs.osu.edu
The Ohio State University Phone: (614) 292-2250
Columbus, Ohio 43210 Fax: (614) 292-6213