一个高性能Key/Value数据库XDB的设计与实现

传统SQL数据库在管理大数据、性能和扩展性等方面的不足使得其不能胜任如大规模数据挖掘、商务智能及社交网络等应用场景,而NoSQL数据库由于其在上述应用领域优良的特性近年来得到很快发展。本文在分析现有开源NoSQL数据库产品Kyoto Cabinet和Tair的不足基础上,设计并实现了一个高性能的Key/Value数据库XDB。XDB分段流水并发处理Key/Value请求,并采用一致性算法保证在并发环境中对Key/Value记录读写的相关性,在随机写入和顺序写入实验中,相比于Kyoto Cabinet,XDB的性能分别提高6倍和15倍。

Design and Implementation of XDB:A High Performance Key/Value Database

The NoSQL DB has been developing fast in such area as Bigdata Mining, Business Intelligence, Social Network System etc., in which the traditional SQL DB is not qualified enough because of its inherently shortcomings like poor capability of handling Bigdata, low CURD performance, and bad scalability. In this paper, we first analyze the deficiency of some popular open-source Key/value databases such as Kyoto Cabinet and Tair, then introduce the design and implementation of the XDB, a high performance Key/value database system. The XDB adopts a request-processing pipeline architecture and handles all key/value requests in full parallel way, moreover, a consistency-keeping algorithm is implemented in the XDB to keep read and write correlation of the key/value requests in parallel environment. The XDB outperforms KC 6X and 15X in sequential write and random write experiment respectively.