Efficient operational management of enterprise file server with user-intended file access time

Abstract

Toward efficient management of enterprise file server, listing potentially deletable files is expected to be effective. For this purpose, we propose a method of extracting frequent file access pattern to sort out files accesses by operating system or of file system, because such access patterns are supposed to be highly frequent. In our method, each access sequence of one user ID at short time intervals is processed in three steps: (1) path is encoded on the basis of relative position in folder tree structure, (2) accesses are clipped into independent sub-trees, and (3) statistical significance of number of observation is calculated according to occurrence probability. In this paper, we apply our method to file server access logs of a company and demonstrate that it can distinguish system-driven access accurately by utilizing our method to calculate user-intended access time, it can contribute to make convincing list of deletable files.