Decision support for supercomputer acquisition

This paper reports on a decision support system for computer capacity planning. The system was developed for Sandia National Laboratories to support planning in an environment characterized by large-scale scientific computing needs, governmental budgetary limitations and specific planning documentation required by the funding source. The implementation of the system is based directly on the concepts of relational model management and logic-based model representation and manipulation. The implementation vehicle for this system is an extended version of Prolog, which allows interactions with externally defined computational routines and sources of data. This vehicle provides for the integration of various data sources with a set of models for estimating capacity requirements and optimizing different objective functions subject to budgetary constraints.