【作者】 冯玉才； 梁俊杰； 曹忠升；

【Author】 Feng Yucai~1,Liang Junjie~(1,2),and Cao Zhongsheng~1 1(College of Computer Science & Technology,Huazhong University of Science and Technology,Wuhan 430074) 2(Faculty of Mathematics & Computer Science,Hubei University,Wuhan 430062)

【机构】 华中科技大学计算机学院； 湖北大学数学与计算机科学学院；

【摘要】 主存多媒体数据库系统性能深受处理器缓存未命中的影响,缓存感知型主存索引是提高数据检索效率的有效手段.针对SA-Tree不适用于主存存取的缺点,提出它的变体CSA-Tree.CSA-Tree利用PCA降维技术,将树的各层节点采用不同的维度来表示,这样不仅提高缓存空间的利用率,还降低了CPU负载,从而提高了索引查询效率.大量实验证明,CSA-Tree在主存环境中具有良好的高维数据检索性能.更多还原

【Abstract】 In main-memory databases,the number of processor cache misses has a critical impact on the performance of the system.Cache-conscious indices are designed to improve performance by reducing the processor cache misses that are incurred during a search operation.Considering the disadvantage of SA-Tree inefficient for main memory access,its variant called CSA-Tree is presented,which is a multi-level structure where each level of the tree represents the data space at different dimensionalities by using principal component analysis.Each level of the tree serves to prune the search space more efficiently as the reduced dimensions can better exploit the small cache line size.Moreover,the distance computation on lower dimensionality is less expensive.Extensive experiments are conducted to evaluate the proposed structures against other methods.The results show that the CSA-Tree is superior in most cases.更多还原