一种基于V3模型的内存数据库性能分析研究

摘要

To satisfy the ever-increasing performance demand of Big Data and critical application’ s operation,the data management needs to offer the flexible schema,high availability,light weight replica,high volume and high scalability features,the 19 kinds of main memory database has carried on the comparison and analysis from storage type,system structure,scale,concurrency,availability and scalability in the memory database. The V3 performance model is proposed considering the velocity,volume and scalability comprehensively,classifying the main memory database,and selecting the representative in-memory database to conduct the analysis and testing in the high frequency of quantitative trading environment. Test results clearly demonstrate that NewSQL is better at dealing with high-frequency trading mod-els. To increase the parallel processing speed of memory database under the condition of multitasking,the design and optimization for memory database of the multi-core environment are analyzed,and the optimization process is divided into fetching,concurrent accelera-tion and data classification mode,and development of memory database is discussed.%针对大数据时代各种复杂业务对数据处理日益增长的性能要求,以及对数据管理中：模式自由、高可用、轻量级复制、大容量水平可扩展等方面的需要,文中从内存数据库的存储类型、体系结构、规模、并发性、可用性与可扩展性等方面对19种主流内存数据库进行了对比分析。提出了一种综合考虑处理速度、规模与可扩展性的V3性能模型,对主流内存数据库进行了分类,并选取了有代表性的内存数据库在高频量化交易测试环境进行性能分析与测试。结果表明,NewSQL数据库有较好的综合性能。为提高内存数据库在多任务并行情况下处理的速度,文中对多核环境中内存数据库的设计与优化进行了分析,将优化过程分为访存、并发加速和数据划分模式,并对内存数据库发展进行了展望。