The simulation of rich star clusters presents challenging problems of several kinds, including the design of suitable hardware and software, and numerous theoretical problems in stellar dynamics and stellar physics. Great progress has been made possible in recent years through the widespread use of GRAPE hardware. Simulations are, however, still too small to be applied to real star clusters without scaling. How this is done is partly an issue of stellar dynamics, and it has thrown into focus a number of fundamental theoretical problems in this field.

10 pages, postscript, also available at http://www.maths.ed.ac.uk/people/douglas/preprints.html To appear in T. Ebisuzaki, ed, Proc. Int. Symp. on Supercomputing - New Horizon of Computational Science, Tokyo, Japan, 1-3 September, 1997. Kluwer, Dordrecht