HBase数据库模式设计准则

General Rules for HBase Schema Design

Hadoop数据库HBase作为基于HDFS的分布式大数据处理系统,其模式设计不能套用或模仿传统关系型数据库方式,而需要与自身特点相匹配,其性能的好坏主要取决于内部表的设计和资源的分配是否合理,如行键、列族的设计以及数据存储等。从行键设计、列族划分、数据量估算与控制、内存需求估算与配置等几个方面研究了HBase数据库模式设计应当遵循的准则,并且在读访问模式和写访问模式下应采用不同的行键设计方式。通过一个电商搜索平台的数据库设计实例说明如何应用这些准则。

Apache HBase of Hadoop database is based on HDFS,and a distributed and big data processing system.Engineers cannot do the schema design for HBase by the same as those for traditional RDBMS.HBase has its own characteristics.The performance of HBase system is mainly related to the table design and resource allocation internally,such as row key design,column family design,configurations for data storage,etc.This paper does research on the general rules for HBase schema design from the following aspects：row key design,column family design,data size estimation and control,memory estimation and configuration,etc.Different row key design rules are defined for read access mode and write access mode.Finally,a case design about an online retailer＇s search platform is described to show the method of adopting these rules when one does HBase schema design.