The Design of Adaptive File Classifying in File system

To improve performance and manage file systems, it must understand important properties (e.g., size, access pattern, life time,) of various files. This paper designs a properties-based adaptive file classifying and describes how systems can automatically learn to classify the properties of files and predict the properties of new files as they are created, by using the associations between a file's properties and the names. Decision tree classifiers can automatically recognize and model such associations. Such predictions can be used to select storage policies (e.g., replication factors and disk allocation schemes) for individual files. Further, changes in associations can reveal information about applications, helping autonomy system components differentiate growth from basic change.