Cost-benefit evaluation of scientific computing services
Donald N. Streeter · IBM Systems Journal · 1972
Evaluation techniques for measuring computing system efficiency continue to be refined as the variety of systems and their applications expand. System measures often tacitly assume that the highest overall efficiency is achieved when a certain critical system parameter, such as throughput, is optimized. Recognizing that these measures may be quite valid when the computing system essentially models the business, this paper does not differ with such criteria. Used in a research environment for supporting individual research workers, however, system measures and policies should attempt to be more flexible, and should strongly consider the human costs and benefits involved. The reason for this is that research projects are so varied that a single system policy may not effectively model the “business” of research. Therefore, we attempt to evaluate and compare system costs and user benefits under a variety of user policies.