J. van den Bercken, B. Seeger, and P. Widmayer. A generic approachtobulk loading multidimensional index structures. In Proc. 23rd Intl. Conf. on Very Large Databases, pages 406--415, 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....rebalancing, have been proposed [15, 26, 9, 18] Bulk loading an R tree with N rectangles using the naive method of repeated insertion takes O(N log B N ) I Os, which has been recognized to be abysmally slow. Several bulk loading algorithms using O( N B log M=B N B )I Oshavebeenproposed [24, 17, 14, 20, 28, 10]. These algorithms are more than a factor of B faster than the repeated insertion algorithm. Most of the proposed algorithms [24, 17, 14,20] work in the same basic way# the input rectangles are sorted according to some global one dimensional criterion (suchasx coordinate [24] the Hilbert value of ....

....[24] the Hilbert value of the center of the rectangle [17, 14] or using an order obtained from a rectangular tiling of the space [20] and placed in the leaves in that order. The rest of the index is then built recursively in a bottom up, level by level manner. The algorithm developed in [28] also builds the index recursively bottom up, but it utilizes a lazy buffering strategy. Since this buffer strategy is similar to the technique we will describe later, we will give a more detailed description of the algorithm of [28] in Section 2.3. Finally, the method proposed in [10] works ....

[Article contains additional citation context not shown here]

J. van den Bercken, B. Seeger, and P. Widmayer. A generic approachtobulk loading multidimensional index structures. In Proc. 23rd Intl. Conf. on Very Large Databases, pages 406--415, 1997.

....by use of bulk loading which is the process of buildingthe index for a set of objects withoutany intervening queries. In recent years, numerous bulk loading algorithms for spatial indexes have been introduced. Most of the attention has been focused on the R tree, and related structures (e.g. [1, 2]) Here we improve upon an algorithm previously developed by us for bulk loading a disk based PMR quadtree [7] Although our presentation and experiments are in terms of the PMR quadtree, our results hold for any variant of the quadtree. In fact, our approach can be adapted to speed up the ....

J. van den Bercken, B. Seeger, and P. Widmayer. A generic approach to bulk loading multidimensional index structures. In Proc. 23rd VLDB Conf., pp. 406--415, Athens, Greece, Aug. 1997.

.... of the first category treat non indexed inputs (e.g. when there is another operation, such as selection, before the spatial join) If there is an Rtree for only one input, processing can be done by (i) index nested loops, ii) building a second R tree for the non indexed input using bulk loading [54] and then applying an R tree based algorithm (see below) iii) the sort and match algorithm [40] iv) the seeded tree algorithm [24] which works like (ii) but builds the second R tree using the existing one as a skeleton (seed) v) the slot index spatial join [27] which is an improved version of ....

van der Bercken J., Seeger, B., Widmayer, P. A Generic Approach to Bulk Loading Multidimensional Index Structures. VLDB, 406-415, 1997.

....) where d is the dimension of the input) and a new dynamic insertion method. Packing Rtrees was successively proposed in [RL85] KF94] and [LEL96] The latter paper describes and experiments the STR variant presented in section 4. A new algorithm for fast bulk loading of R tree is described in [dBSW97] R trees were primarily designed for supporting point and range (window) queries. Some authors consider other operations. RKV95] study the problem of reporting the nearest neighbors of a point P using R trees. Given the mbb R of an object O, two distance functions of (R; P ) are defined, which ....

Jochen Van den Bercken, Bernhard Seeger, and Peter Widmayer. A generic approach to bulk loading multidimensional index structures. In Proc. Intl. Conf. on Very Large Data Bases, pages 406-- 415, 1997.

....an improved version of R tree that employs a sophisticated insertion algorithm, achieving optimal tree quality. The R tree and R tree are dynamic SAMs that build and maintain their structure incrementally, thus serving as efficient index methods for spatial data. Packing algorithms [RL85] KF93] vdBSW97] build optimal R tree structures for a static set of objects in space. The resulting packed R trees have full leaf nodes, and thus minimum number of nodes and height, leading to minimization of search time. Among the most important spatial queries is the spatial join, which retrieves from two ....

....value for modern disks [SKS97] In all cases IRS is substantially better than the other two policies, with SplitXL to perform very bad because of extensive replication. The inferior performance of SplitXL and SplitHC is due to the fact that the entries to be split have large spatial extents [vdBSW97] In the rest of the paper we adopt IRS as the standard SISJ partitioning policy. 12 0 20 40 60 GSxGR LBxMG U1xU2 G1xG2 LBxU1 MGxG2 G2xU2 SplitXL SplitHC IRS 0 500 1000 1500 2000 GSxGR LBxMG U1xU2 G1xG2 LBxU1 MGxG2 G2xU2 SplitXL SplitHC IRS 0 20 40 60 GSxGR LBxMG U1xU2 G1xG2 LBxU1 MGxG2 G2xU2 ....

van der Bercken J., Seeger, B., Widmayer, P. "A Generic Approach to Bulk Loading Multidimensional Index Structures". VLDB, 1997.

....time than SR tree for the reasons explained previously. In the current implementation, all three trees use point by point insertion. Therefore, based on the average performance of the tree generation, we can project that PK tree takes much less time for insertions than the SR tree and the X tree. [Ber97] describes a batch loading algorithm that can be applied to most indexing structures to improve the structure generation time. However, at this time, we are not sure what the effect of the batch loading algorithm would be on the three indexing structures. 8.2 K Nearest Neighbor Query ....

Jochen Van den Bercken, Bernhard Seeger, Peter Widmayer. A Generic Approach to Bulk Loading Multidimensional Index Structures. Proc. 23rd Intl. Conf. on Very Large Data Bases (VLDB), 406415, 1997.

....the records of the second relation are probed against the first relation. This step is performed on the index structure that has been created previously in the build phase. In order to support both phases efficiently, we use a technique recently proposed for bulk loading index structures [BSW 97] This technique is not limited to bulk loading, and a modified version of this technique can be applied to probing the second relation in bulk as well. Our work differs from previous approaches for the following reasons. First, our approach is generic, i.e. it can be applied to a broad class ....

....a data record t and a query q. Return true if t satisfies q. Note that these methods are in general available in a grow and post tree. 3 The Build Phase The build phase is based on creating an index in a way adapted from a recently presented generic approach to bulk loading index structures [BSW 97] In contrast to [BSW 97] the index of the bulk index join consists of large index nodes. The fan out of the index nodes (C) is chosen in such a way that an index node and a page for each of its child nodes fill up main memory almost totally. More formally, we assume that and C m hold. ....

[Article contains additional citation context not shown here]

J. van den Bercken, B. Seeger, P. Widmayer: "A Generic Approach to Bulk Loading Multidimensional Index Structures". Proc. VLDB 1997: 406-415.

No context found.

ACM, 20(5):339--343. van den Bercken, J., Seeger, B., and Widmayer, P. (1997). A generic approach to bulk loading multidimensional index structures. VLDB'97, pages 406--415, Athens, Greece.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC