Metadata changes in large file systems: a metadata querying perspective

Understanding metadata changes is crucial to the design and implementation of file-system-wide metadata querying systems, which are important tools for administrators. However, existing file system trace studies pay much attention to optimizing the design and implementation of file systems themselves, but little attention to supporting applications. In this paper we present the analysis of two large-scale file system traces from enterprise-class file servers deployed in the NetApp data center. Different from most existing file system trace studies, we did our study from the perspective of supporting storage management metadata querying. We took an in-depth look at the relationships between changes of different attributes of the same file and changes of different files (such as those in the same directory or a file and its parent), focusing on how such relationships could be used to predict change patterns. We found several characteristics that are considered important in the design and implementation of file system utilities, especially file-system-wide metadata querying tools. Furthermore, we also give some suggestions on how our findings can be used.