An analytic resource model for large-scale software development

摘要

Recent work conducted by members of the Purdue Software Metrics Research Group has focused on the complexity associated with coordinating the activities of persons involved in large-scale programming efforts. A resource model is presented which is designed to reflect the impact of this complexity on the economics of software development.The model is based on a formulation in which development effort is functionally related to measures of product size and manloading. The particular formulation used is meant to suggest a logical decomposition of development effort into components related to the independent programming activity of individuals and to the overhead associated with the required information flow within a programming team.The model is evaluated in light of acquired data reflecting a large number of commercially developed software products from two separate sources. Additional sources of data are actively being sought.Although strongly analytic in nature, the model's performance is, for the available data, at least as good in accounting for the observed variablility in development effort as some highly publicized empirically based models of comparable complexity. It is argued, however, that the model's principle strength lies not in its data fitting ability, but rather in its straight forward and intuitively appealing representation of relationships involving manpower, time, and effort.