Efficient Disk Allocation for Fast Similarity Searching

As databases increasingly integrate non-textual information it is becomingnecessary to support efficient similarity searching in addition to rangesearching. Recently, declustering techniques have been proposed for improvingthe performance of similarity searches through parallel I/O. In this paper, wepropose a new scheme which provides good declustering for similaritysearching. In particular, it does global declustering as opposed to localdeclustering, exploits the availability of extra disks and does not limit thepartitioning of the data space. Our technique is based upon the cyclicdeclustering schemes which were developed for range and partial match queries.We establish, in general, that cyclic declustering techniques outperformpreviously proposed techniques.