R. Orlandic, J.L. Pfaltz, and Y. Lepouchard (USA)
Scientific databases, multi-dimensional access, query transformation, data dimensionality.
Contemporary research on multi-dimensional databases fo cuses on the efficient representation and retrieval of point objects. However, for scientific database applications, ef fective representation of extended (regional) data is also important. In these applications, regional data typically appear through aggregation and/or clustering of points. A common problem of contemporary retrieval techniques de signed for extended multi-dimensional objects, which are often referred to as spatial access methods, is their inability to cope with data of widely varying size. In this paper, we investigate the idea of segregating the regional objects into separate search trees according to their size. While the ap proach is appealing for many different spatial access meth ods, the paper investigates the idea in the context of a re cently designed spatial access method for high-dimensional data, called QSF-trees. The experimental evidence shows that, when the volumes of objects vary widely and the num ber of large objects is proportionally small, separating ob jects by size may significantly improve the performance of multi-dimensional selections.
Important Links:
Go Back
