A Hybrid Indez Structure on Multi-core Cluster Architecture

摘要：

Since multi-core has been mainstream of processors, multicore cluster architecture become more important for many server appli cations, including high-dimensional data processing. In this paper, we present a hybrid-index structure for high-dimensional data on multicore clusters. To make full use of two-level parallelization of multi-core clusters, we design an index structure for high-dimensional data: HKDtree(Hybrid K-Dimensional Tree). An HKD-tree is combined by KDtree and LSH, which uses LSH in the leaf nodes of KD-tree. We paral lelizes operations(tree construction and query processing) of HKD-tree. We evaluate the performance of HKD-tree with real image dataset. Due to the experiment results, HKD-tree is more efficiently for query process ing on multi-core cluster architecture.

关键词： multi-core parallelization HKD-tree cluster high-dimensional data

作者: Bai Long GuangZhong Sun Guoliang Chen

作者单位: School of Computer Science and Technology University of Science and Technology of China Hefei,230026,China

会议类型: 国际会议

会议名称: The Second International Symposium on Parallel Architectures,Algorithms and Programming(第二届国际并行体系结构、算法和程序设计研讨会)

会议地点: 南宁

会议语种:英文

页码: 1-18

在线出版日期: 2009-12-04（万方平台首次上网日期，不代表论文的发表时间）