首页 | 本学科首页   官方微博 | 高级检索  
     

面向Flash Memory的高性能数据存储引擎的研究
引用本文:周晓云,覃雄派,徐钊.面向Flash Memory的高性能数据存储引擎的研究[J].工矿自动化,2009,35(6).
作者姓名:周晓云  覃雄派  徐钊
作者单位:1. 中国矿业大学信电学院徐州师范大学计算机科学与技术学院,江苏,徐州,221008
2. 中国人民大学信息学院,北京,100872
3. 中国矿业大学信电学院
摘    要:传统的数据存储引擎对Flash Memory数据的修改是通过页内更新技术实现的,这将导致FlashMemory的性能下降及其磨损加剧。针对该问题,文章提出了一种面向Flash Memory的采用页外更新技术的多版本数据存储引擎MV4Flash。该数据存储引擎采用多版本存储和垃圾回收机制,所有数据的更新和修改都通过文件追加的方式进行,适应了Flash Memory先擦除后写入的特点,延长了设备寿命。采用NDBBench对该数据存储引擎进行测试的结果表明,MV4Flash与传统的InnoDB相比,事物处理性能有较大的提升,更适合于数据规模大、实时性要求高的应用系统。

关 键 词:Flash  Memory  数据存储引擎  页内更新  页外更新  多版本  垃圾回收  NDBBench

Research of a Data Storage Engine with High Performance Oriented Flash Memory
ZHOU Xiao-yun,QIN Xiong-pai,XU Zhao.Research of a Data Storage Engine with High Performance Oriented Flash Memory[J].Industry and Automation,2009,35(6).
Authors:ZHOU Xiao-yun  QIN Xiong-pai  XU Zhao
Abstract:Traditional data storage engine updates data by in place update technique which would induce performance reduction and wear aggravation of Flash Memory.Aiming at the problem,the paper proposed a multi version data storage engine MV4Flash oriented Flash Memory which used out place update technique.The data storage engine uses the mechanism of multi version storage and garbage collection and updates and modifies all data by file superaddition mode,which is suitable for the characteristic of erase before write of Flash Memory and can prolong its life.The result of performance test for the data storage engine with NDB Bench showed that compared with traditional InnoDB,the MV4Flash has higher performance of transaction processing,so the MV4Flash is more suitable for application system with large data volume and high real-time demand.
Keywords:Flash Memory  NDB Bench
本文献已被 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号