In this paper we present results from our research work on the parallel processing of knowledge-base systems, for which a new computational model has been developed. The focus here is on the preprocessing aspect of this model, which plays a crucial role in reducing the communication overheads and establishing a suitable environment for runtime processing. Algorithms have been developed for three operations: the clustering of hierarchies, their mapping to a processor network, and the determination of eflcient communication paths between hierarchies assigned to adjacent and non-adjacent processors. Throughout these operations speed is a primary goal, and the algorithms developed are designed to give good solutions in reasonable time, rather than to give optimal solutions.