Dynamic file structure for partial match retrieval based on overflow bucket sharing

Abstract

A hashing-based dynamic file structure is introduced for partial match retrieval using overflow bucket sharing. The sharing of overflow buckets is dynamic in the sense that an overflow bucket is shared by a varying number of primary buckets according to the local conditions of the file. The use and sharing of overflow buckets defers splitting of the data buckets, thereby increasing the storage utilization. For the same reason, plus the fact that the sharing is dynamic, the growth of the directory is slowed down. Under the proposed organization, the records are stored more compactly in the data buckets, and for those partial match queries in which few attributes are specified, groups of neighboring directory entries have high probability of being referenced together, so that the retrieval costs for these types of partial match queries are reduced. This file organization is found to be space efficient and is also time efficient for queries in which the number of specified attributes is small.