Empirically Derived Micromodels for Sequences of Page Exceptions

Abstract

Based on a statistical analysis of actual computer program address traces, some results are presented of a study aimed at deriving empirically valid stochastic models for program reference patterns in a computer system operating under demand paging. For the address traces examined, a semi-Markov model for the (univariate) point process of page exceptions is formulated and found to be an adequate characterization of the data.