Analysis of retrieval performance for records and objects using optical disk technology

Abstract

In this paper we examine the problem of object and record retrieval from optical disks. General objects (such as images, documents, etc.) may be long and their length may have high variance. We assume that all the components of an object are stored consecutively in storage to speed-up retrieval performance. We first present an optical disk model and an optimal schedule for retrieval of records and objects which qualify in a single query on a file stored on an optical disk device. We then provide exact and approximate analytic results for evaluating the retrieval performance for objects from an optical disk. The analysis provides some basic analytic tools for studying the performance of various file and database organizations for optical disks. The results involve probability distribution of block accesses, probability distributions of span accesses, and probability distribution of seek times. Record retrieval is an important special case. This analysis differs from similar ones in database environments in the following respects: (1) the large size and large variance of the size of objects; (2) crossing of track boundaries by objects; (3) the capability for span access that optical disks provide (e.g., when the optical assembly is located in a given position, information can be read from a number of consecutive tracks (span) with a small additional cost).