Long-Range Dependencies and Statistical Self-Similarity in Computer Memory System

Abstract

This paper shows the problem of possible existence of statistical self-similarity and long-range (long-term) dependencies in behavior of computer memory systems. It will be shown, that the hierarchical structure of memory and the dispersion of time constants during accessing successive levels of such a structure, lead to "memory effect" and self-similarity. These effects have not been taken into account so far in computer systems design, management and description. A few methods were used to calculate the Hurst H coefficient, which reflects the statistical self-similarity in time series and the d parameter that represents the problem of long-range dependencies.