SL-trees: An indexing structure for object-oriented data bases

In an object-oriented data base, the scope of a query is often a class hierarchy rooted at a particular class. That is, a query may either request instances chosen only from a root class, or it may request instances which are chosen from any subclass of the root class. So, the data base indexing method must support efficient retrieval of instances from either a single class or from a hierarchy of classes. In this paper, we propose a new hierarchical indexing structure called the SL-tree. Like previous indexing methods for data bases, SL-trees are based on B+ trees. Like H-trees, SL-trees have a B+-like tree for each class of a class hierarchy, and these trees are nested according to their superclass-subclass relationships. However, the SL-trees have far fewer links than H-trees, leading to better performance because of lower storage overhead and fewer link updates. The simulation results indicate that SL-trees need very few nesting pointers and that SL-trees are a simple and efficient indexing structure for object-oriented data bases.