A computer performance prediction model
Robert W. Otto, Mark Auerbach · 1976
This paper presents a model which departs from the classical queuing theory approach and uses probability to predict computer system throughput. The probability a program desires computer resources (CPU or channels) is obtainable from system measurements or can be estimated. Knowing these probabilities, the multiprograming throughput of a system concurrently executing a number of programs is calculable. The multiprograming model presented is easy to use and requires only the probabilities denoted above. The serial model, reviewed briefly in this paper has been previously published.