How spacey can they get? Space overhead for storage and indexing with object-oriented databases

The impact of physical storage model choice on the performance of an object-oriented database is studied. It is examined how each of six physical storage models´ space overhead reacts to changes in database parameters such as directed acrylic graph (DAG) shape, number of instances per class, and distribution of instances over the DAG. Home-class, leaf-overlap, and split-instance physical models consistently require the least storage, whereas repeat-classes, universal-class, and value-triple models require the most. For all models, the depth of the DAG has the strongest impact on space overhead. A set of analytic formulas is developed that allow a database designer to estimate database size for each physical model