内存数据库索引研究  

Research of index structure for main memory database

作　　者：王莹[1] 王艳[2] WANG Ying;WANG Yan(College of Computer Science and Technology,Harbin Engineering University,Harbin 150001,China;Heilongjiang Computing Center,Harbin 150028,China)

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150001 [2]黑龙江省计算中心,哈尔滨150028

出　　处：《齐齐哈尔大学学报（自然科学版）》2019年第1期23-26,共4页Journal of Qiqihar University(Natural Science Edition)

摘　　要：内存数据库将数据存储在内存中进行操作,与基于磁盘的数据库相比在性能上得到提升,但是由于MMDB与DRDB的性能瓶颈的差异,用于DRDB的B+树索引算法已经不再适用于MMDB。为适应内存数据库的性能,T树,CSS树,CSB+树等索引算法相继被提出,但是这些算法没有考虑到TLB失效会影响索引的性能。本文通过对CSB+树进行改进,得到一种更适用于内存数据库的索引算法。实验证明,该索引结构可以降低TLB失配,提升处理器缓存的利用率,提高查询效率。MMDB stores data in memory for operation,which improves performance compared with disk-based database.But due to the difference in performance bottleneck between MMDB and DRDB,the B+tree indexing algorithm for DRDB is no longer suitable for MMDB.In order to adapt to the performance of main memory database,indexing algorithms such as T-tree,CSS tree,and CSB+tree have been proposed,but these algorithms do not take into account that TLB failure will affect the performance of the index.This paper improves the CSB+tree and obtains an indexing algorithm that is more suitable for in-memory databases.Experiments show that the index structure can reduce the TLB mismatch,improve the processor cache utilization,and improve query efficiency.

关 键 词：内存数据库 索引 TLB CSB+树 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]