Sequential vs. random retrieval of a large indexed file

Abstract

An indexed sequential (IS) file allows both sequential and random retrieval of its records. When a large IS file is stored in a disk pack with one assembly that must seek different cylinders to retrieve records, the total seek times will become a primary concern for choosing file access methods. While sequential retrieval requires some additional time for sorting the input keys, it may gain considerable reduction in seek times as the access arm does not move in and out to access different records. Once a sorting (internal or external) algorithm is chosen, the time can be estimated as sorting techniques are well developed and documented. In this paper, the average total seek times for both sequential and random retrieval of the file are derived. The result can be used to facilitate a tradeoff study between the two retrieval methods for a particular application.