An Approach to Software Evaluation.
Software evaluation is discussed within the context of RAND's experience in evaluating its computer system. Software has become more significant because of increased dependence on it for system operation and performance. At the same time, it has become more difficult to evaluate because of increased complexity, greater flexibility and functional capability, and sheer size. Any organization, undertaking an evaluation of its system, must consider two major areas: functional capabilities and performance based on the organization's requirements and goals. At RAND, workload and capacity requirements, desired throughput and accessibility, and cost constraints were used to determine an acceptable cost/performance range. Estimating software performance is the weakest link in an evaluation. If an organization's output is largely dependent on accurate performance prediction, it must select only from systems with working software. 12 pp.