摘要
随着主存技术的发展,系统级分布式主存多媒体数据库存储规模不断扩大,查询越来越复杂,分布式并行索引以其高性能而逐渐成为解决此类复杂查询问题的有效手段。文章提出一种新的适合于分布式并行索引的树结构——DPHI-Tree。该索引树以B树为基础,利用PCA降维技术,将树的各层节点采用不同的维度表示,提高了缓存空间的利用率,降低了CPU负载。另外,采用标识根到节点路径的二进制编码来对节点索引,在查询数据量急剧增加时,PHI-Tree对基于树的数据随机访问具有良好性能。
Aim.Pointing out what we believe to be the deficiencies of past research at the beginning of the full paper,we attempt to put forward a new DPHI-tree method.Fig.2 in section 1 of the full paper gives the block diagram of the structure of the DPHI-tree.The core of subsection 2.2 is:using the top-down method to construct the DPHI-tree of a certain data set;the DPHI-tree constructed through principal component analysis(PCA) can reduce the number of the sub-nodes of every node,greatly reducing the node space.Se...
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2009年第4期436-442,共7页
Journal of Northwestern Polytechnical University
关键词
信息索引
树
组件原理分析
分布式高维索引树
主存访问
indexing(of information)
trees(mathematics)
principal component analysis
Distributed Parallel High-dimensional Index(DPHI) tree
main memory access