6 Access Methods and Query Processing Techniques

The performance of a database management system (DBMS) is fundamentally dependent on the access methods and query processing techniques available to the system. Traditionally, relational DBMSs have relied on well-known access methods, such as the ubiquitous B-tree, hashing with chaining, and, in some cases, linear hashing [52]. Object-oriented and object-relational systems have also adopted these structures to a great extend. During the past decade, new applications of database technology — with requirements for non-standard data types and novel update and querying capabilities — have emerged that motivate a re-examination of a host of issues related to access methods and query processing techniques. As an example, a range of applications, like cadastral, utilities, shortest path finding, etc., involve geographic, or spatial, data, which are not supported well by existing technology, making it not only desirable, but plain necessary to examine access methods and query processing techniques afresh. Specifically, Oracle’s Spatial Data Engine uses Linear Quadtrees [2] at a conceptual level, but uses B-trees as the storage mechanism for the quadtrees at the implementation level. As another example, R-trees [34] have been implemented by Oracle. However, R-trees are mapped to B-trees [75], in order to not change other system components, such as the transaction manager, the recovery manager, the buffer manager, etc. Thus, R-trees do not support deletions physically, but only logically, because this is the practice of B-trees [77], which is the brand of B-trees implemented in commercial systems. When, as we have seen, DBMSs supporting only the traditional access methods fall short in supporting spatial data, it is not surprising that new access methods and query processing techniques are needed if DBMSs are to support the many and diverse emerging applications that call for the management of spatio-temporal data. While access methods and techniques exist that support time and, as discussed above, space, existing proposals are unable to simultaneously support time and space, either efficiently or at all.