A synthetic job for measuring system performance
William J. Buchholz · IBM Systems Journal · 1969
In the absence of a theoretical definition of performance, we will here describe performance quantitatively in terms of the running time of a given job, or rather the reciprocal of this time so as to associate the faster marhine with the larger number. To make performance a reproducible measure, we will, as far as possible, exclude hard-to-control variables, such as manual set-up and error recovery times. These important aspects of overall performance must be stated as factors reducing the ideal performance of a perfect system.