An Evaluation of Physical Disk I/Os for Complex Object Processing

In order to obtain the performance required for nonstandard database environments, we need suitable storage structures for complex objects. In this paper we use an hierarchical complex object model with object references. We describe several storage models for these complex objects, as well as a benchmark to evaluate their performance. We develop a cost model for analytical performance evaluation, and validate the analytical results by means of measurements on the DASDBS complex object storage system. The results show which storage structures for complex objects are the most efficient under which circumstances. 1 Introduction Data-intensive applications such as geographic information systems, robotics, and CAD/CAM, the socalled non-standard database applications, require a high performance in retrieving and processing complex objects. Complex objects are data objects that are both highly structured, and large in size. These large clusters of structured data are a unit of manipulatio..