Batched searching in database organizations (Q1113697)

Savings in the number of page accesses due to batching on sequential, tree-structured, and random files are well known and have been reported in the literature. This paper asserts that substantial savings can also be obtained in database organizations by batching the requests for records (in queries), and also by batching intermediate processing requests while traversing the database. A simple database having two interrelated files is used to demonstrate such savings. For the simple database, three variations on batching are reported and compared with the case of unbatched requests. New mathematical expressions have been developed for the batched cases as well as for the unbatched case, and the savings are demonstrated with some example problems. As an extension, larger databases will enjoy even greater savings due to batching. The paper also discusses several strategies for applying the batching approach to current databases, and the advantages of emerging very large main memories for the batching approach.