Dynamic file organizations for partial match retrieval based on linear hashing

Abstract

Two new file organizations based on Linear Hashing are proposed for partial match retrieval. The first organization introduces a load-balancing scheme whereby overflow records are stored temporarily in other primary buckets so that the allocation of overflow buckets are deferred. The second organization defers the physical splitting of underflow buckets, so that the records belonging to underflow buckets can be retrieved together. These two techniques are then combined together to form a new variation of Linear Hashing. Compared with the original scheme, the performance of these organizations for partial match retrieval are improved, both in terms of storage utilization and retrieval time.