We now move towards the construction of a D-GRAID prototype underneath a block-based SCSI-like interface. The enabling technology underlying D-GRAID is semantic knowledge [44]. Understanding how the file system above utilizes the disk enables D-GRAID to implement both graceful degradation under failure and quick recovery. The exact details of acquiring semantic knowledge within a disk or RAID system have been described elsewhere [44]; here we just assume that a basic understanding of file system layout and structures is available within the storage system. Specifically, we assume that D-GRAID has static knowledge of file system layout, including which regions on disk are used for which block types and the contents of specific block types, e.g., the fields of an inode.