Analysis of indexed sequential and direct access file organizations

Abstract

A computer systems designer is faced with a decision concerning the organization of data files. He often would like to design a file so that sequential and random processing can both be performed efficiently. Two file organizations often proposed for these processing requirements are indexed sequential and direct. Another technique called “batch random” has been proposed for the batch updating of direct access files. Discrete simulation models of these file organizations and access methods in both transaction and batch processing modes are developed and used to analyze the file organizations. A general guideline is derived to indicate whether batch or transaction processing should be performed based on the percentage of the file which is to be processed. Based on these results, a methodology for selecting the “best” file organization is developed for a given set of criteria. Not all the parameters for describing a file organization are incorporated into the methodology and therefore “best” does not necessarily mean optimal.