Analysis of File Reference Patterns and Migration Policies.

In this paper, we describe a statistical experiment that is used to characterize the file interreference time (the time between two consecutive file accesses), and the size of the accessed files in a general purpose computing environment. The data collected during the experiment are analyzed and used to evaluate file migration policies. Specifically, we use theoretical models to fit the empirical distributions of the file interreference time and the size of an accessed file. The association between the size of a file and its interreference time is studied. Furthermore, we compare the performance of two file migration policies.