A direct hashing directory for fast inode lookup

Abstract

In a conventional file system, the directory tree is traversed to find the inode number of a file. The mode lookup performance degrades as the size of the directory tree increases. In this letter, a new directory scheme, called direct hashing directory, is proposed. The mode number of a file is the cyclic redundancy code of the file's absolute path name such that the inode number can be computed directly. The average number of disk accesses for inode lookup is 1.08, which is order of magnitude faster than the conventional directory schemes such as hashing, B tree, and sequential directory.