White Paper: Age-In-Place Storage: When Not To Tier And Migrate Active Data

By Randy Chalfant

We acknowledge the changes in file reference behavior are subtle, consistent, and complex. There are two ways to "look" at any file on a Unix file system4. The most common is with the file's path and name, which is exactly what people and most programs do when they open or close a file as we look at a directory structure. Even so, names are not used to mange files. For the operating system to manage a file, it must give each file a unique number. In this way, management is based upon the file names and the i-nodes, respectively. You can think of one as a hierarchical name space, à la the directory, and the other as a numeric index. To most efficiently manage a file system, both must be considered because these two views of the same file system activity often differ in significant ways.