A generic approach to bulk loading multidimensional index structures英文资料.pdfVIP

A Generic Approach to Bulk Loading Multidimensional Index Structures Jochen van den Bercken1 Bernhard Seeger1 Peter Widmayer2 1Fachgebiet Informatik, Universität Marburg 2Institut für Theoretische Informatik Hans-Meerwein-Str., D-35032 Marburg, Germany ETH-Zentrum, CH-8092 Zürich, Switzerland e-mail: {lastname}@informatik.uni-marburg.de e-mail: widmayer@inf.ethz.ch [Hen 96]. Common to all these applications is that one-di- Abstract: Recently there has been an increasing in- mensional index structures (e.g., B+-trees) are not appro- terest in supporting bulk operations on multidimen- priate indexing techniques. sional index structures. Bulk loading refers to the In order to support queries on multidimensional data sets process of creating an initial index structure for a efficiently, a huge number of multidimensional index struc- presumably very large data set. In this paper, we tures have been proposed during the last two decades. Mul- present a generic algorithm for bulk loading which tidimensional index structures support insertions, deletions is applicable to a broad class of index structures. and updates, as well as proximity queries (e.g., window Our approach differs completely from previous queries, nearest-neighbor queries). In addition to these tra- ones for the following reasons. First, sorting multi- ditional operations, there is currently an increasing interest dimensional data according to a predefined global in supporting b