面向PM优化的数据库存储引擎设计  

Design of Database Storage Engine for PM Optimization

作　　者：陈河堆 闫宗帅 CHEN Hedui;YAN Zongshuai(Nanjing Zhongxing New Software Co.,Ltd,Nanjing Jiangsu 210012,China)

机构地区：[1]南京中兴新软件有限责任公司,江苏南京210012

出　　处：《信息与电脑》2022年第5期175-178,共4页Information & Computer

摘　　要：近年来新型存储硬件技术快速发展,出现了一种介于DRAM与磁盘之间的颠覆性存储介质,称为非易失性内存(Non-Volatile Memory,NVM),具有字节可寻址、非易失性等特征,也叫持久内存(Persistent Memory,PM)。它弥补了内存和磁盘之间的性能鸿沟,对传统DBMS软件设计提出新的挑战。本文针对数据库存储引擎的3个关键技术点,包括B+tree索引、缓存页、微事务日志,探讨在PM新型存储介质下的实现机制优化方法。实验结果表明,本文所采用的优化方法及其实现机制可充分挖掘新硬件潜力,有效提升数据库处理性能。In recent years, with the rapid development of new storage hardware technology, a subversive storage medium between DRAM and disk has emerged, which is called Non-Volatile Memory(NVM). It has the characteristics of Byte addressable and nonvolatile, also known as Persistent Memory(PM). It makes up the performance gap between memory and disk, and puts forward new challenges to the traditional DBMS software design. Aiming at the three key technical points of database storage engine, including B +tree index, cache page and micro transaction log, this paper discusses the implementation mechanism and optimization method under the new PM storage medium. The experimental results show that the optimization method and its implementation mechanism used in this paper can fully tap the potential of new hardware and effectively improve the database processing performance.

关 键 词：数据库 存储引擎 NVM PM 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]