Associative searching in multiple storage units

Abstract

A file maintenance model, called the multiple random access storage units model, is introduced. Storage units can be accessed simultaneously, and the parallel processing of an associative query is achieved by distributing data evenly among the storage units. Maximum parallelism is obtained when data satisfying an associative query are evenly distributed for every possible query. An allocation scheme called M -cycle allocation is proposed to maintain large files of data on multiple random access storage units. The allocation scheme provides an efficient and straightforward indexing over multidimensional key spaces and supports the parallel processing of orthogonal range queries. Our analysis shows that M -cycle allocation achieves the near-optimum parallelism for processing the orthogonal range queries. Moreover, there is no duplication of records and no increase in insertion/deletion cost.