The software empiricist

Abstract

The advent of software and hardware monitoring technology has presented us with a flood of data, without bringing commensurate understanding by which to interpret it. Thus, the most important problem before us in the field of computer measurement is to discover the relationships between the variables we measure and the overall system properties of interest. Particularly, we wish to be able to predict system behavior and performance from a knowledge of the values of factors under our control. In this way, not only will we understand the meanings of these variables, but we shall learn how to design our systems to perform as we wish them to. The latter is a prime goal of software engineering, the former the rational of what has been called software physics. In this section of the Review we are and shall be interested in the empirical development of such an understanding, and the experimental aspects of computer measurement. Our intent is to assist in the building of a solid body of knowledge by providing a publication vehicle for empirical and experimental data. That is, we have little interest in publishing theory, which can normally be done elsewhere. Our goal is to publish experimental data to support or refute theory, and empirical data from which theory builders may take their inspiration.