Metacomputing in a Regional ATM-Testbed - Experience with Reality

Metacomputing, here used meaning "heterogeneous supercomputing", offers a great potential for improving the performance of a program. While some work is done in using heterogeneous clusters of workstation, only little experience has been obtained in doing metacomputing. This paper describes the experiences and results which were gained in the project "Distributed massively parallel computer", which was part of the "Regional Testbed North-Rhine-Westphalia".