摘要
空间索引是提高空间数据库查询性能的关键技术。空间数据具有海量、空间目标不规则、结构和关系复杂等特征,要动态地维护空间索引结构,传统R树的构建方法插入代价非常高。在深入分析空间索引批量加载算法基础上,面向多核处理器的新型硬件架构,基于OpenMP并行编程模型,实现Hilbert R树索引的并行批量加载算法。实验结果表明,相对于串行经典算法,该算法的并行效率接近50%,通过查询实验验证,并行加载算法保持了串行算法生成索引的优良查询性能。
Spatial index is a key technology for improving the inquiry performance of spatial database.As the spatial data has the characteristics of massive amount of data in the database space,irregular space target,complex structure and relationship,the inserting cost of the traditional R tree loading algorithm is very high to dynamically maintain the spatial index structure.Based on in-depth analysis of spatial index bulk-loading algorithm,the Hilbert R-tree index parallel bulk-loading algorithm based on multi-core hardware architecture was realized by the aid of OpenMP parallel programming model.Experimental data shows that,compared with the classic serial algorithm,the parallel efficiency of this algorithm is close to 50%.The results of inquiry experiment certify that the parallel bulk-loading algorithm maintains the good query performance of the serial algorithm.
出处
《现代电子技术》
2011年第22期90-94,共5页
Modern Electronics Technique
基金
国家自然科学基金资助项目(61070035
60902036
40801160)
高等学校博士学科点专项科研基金(20104307110017)
国家高技术研究发展计划("863"计划)课题资助(2011AA120306)
关键词
空间索引
批量加载
多核
并行加载算法
spatial index
bulk-loading
multiprocessor
parallel loading algorithm
