Research on a Parallel BD-tree Index Structure
College of Information Science and Engineering, Guangxi University for Nationalities, Nanning 530006, China
Journal of Computational Information Systems, Vol. 9 (12) : 4831-4838, 2013
@article{liu2013research,
title={Research on a Parallel BD-tree Index Structure},
author={LIU, Yong and LUO, Liping and JIA, Lianyin and HUANG, Jinwang and WANG, Jingxing and HUANG, Dongpin},
journal={Journal of Computational Information Systems},
volume={9},
number={12},
pages={4831–4838},
year={2013}
}
The BD-tree is an efficient database index structure which has good random access performance like hashing methods and can also provide range search and key sequential access like the B+-tree. In order to further improve the operational performance of the BD-tree, we adapted the traditional BD-tree hash function to realize the BD-tree parallel processing. By analyzing the mapping relationships between each key of internal node and corresponding leaf node of the BD-tree, we proposed a parallel solution of quickly creating and querying of the BD-tree on GPU. Comparing among the test results on various platforms, the results show that the proposed solution is about 27 times faster on batch creating and 140 times on batch querying of the BD-tree than traditional CPU solutions, as the solution make full use of the high parallel throughput of GPU.
June 14, 2013 by hgpu
