Sub-library sub-table presentation
With the rise of the micro-architecture of this service, we have a complete application from the application of large, cut into many small applications can provide services independently. Each application has its own database.
Cut into two data points:
Vertical segmentation: be segmented according to the service module, different modules of the cutting table into different databases.
Level segmentation: The large table in accordance with a certain degree of segmentation rules, in accordance with the line cut into different tables or different libraries.
MyCAT Introduction
Official Website: http://www.mycat.org.cn/
What is MyCAT?
Simply put, MyCAT is:
- A completely open source, enterprise-oriented application development "big database cluster."
- Supports transactions, ACID, can replace the enhanced version of Mysql database
- Enterprise-class database can be viewed as a "Mysql" Clusters, Oracle Clusterware used to replace expensive
- A memory cache fusion technology, Nosql technology, HDFS large new SQL Server data
- The new combination of traditional database and distributed data warehouse generation of enterprise-class database product
- A new database middleware products
MyCAT objectives are: low cost would smooth migration of existing stand-alone databases and applications to the "cloud" side, resolve data bottlenecks in the rapidly growing data storage and scale of business situations.
Key Features of MyCAT
- Supports SQL 92 standard
- Mysql support clusters, can be used as a Proxy
- JDBC connection support ORACLE, DB2, SQL Server, MySQL Server will use it to simulate
- Support galera for mysql cluster, percona-cluster or mariadb cluster, high availability cluster data pieces
- Automatic failover, high availability
- Support separate read and write support Mysql from multiple double main, and a mode from a plurality of master
- Support global table, the data to be fragmented plurality of nodes for efficient table associated with the query
- Support partitioning strategy based on the unique relationship of ER, enables efficient table associated with the query
- Multi-platform support, simple deployment and implementation
MyCAT architecture
MyCAT core concepts
- Schema: specifies the logical database
- Table: logic table
- DataNode: real storage node
- DataHost：真正的数据库主机
Mycat存在的问题
跨库join问题
- 通过业务分析，将不同库的join查询拆分成多个select
- 建立全局表（每个库都有一个相同的表）
- 冗余字段（不符合数据库三范式）
- E-R分片（将有关系的记录都存储到一个库中）
- 最多支持跨两张表跨库的join
分布式事务（弱事务）
- 强一致性事务（同步）
- 最终一致性事务（异步思想）
分布式主键
- redis incr命令
- 数据库（生成主键）
- UUID
- snowflake算法
1.1 分片策略
MyCAT支持水平分片与垂直分片：
- 水平分片：一个表格的数据分割到多个节点上，按照行分隔。
- 垂直分片：一个数据库中多个表格A，B，C，A存储到节点1上，B存储到节点2上，C存储到节点3上。
MyCAT通过定义表的分片规则来实现分片，每个表格可以捆绑一个分片规则，每个分片规则指定一个分片字段并绑定一个函数，来实现动态分片算法。
- Schema：逻辑库，与MySQL中的Database（数据库）对应，一个逻辑库中定义了所包括的Table。
- Table：表，即物理数据库中存储的某一张表，与传统数据库不同，这里的表格需要声明其所存储的逻辑数据节点DataNode。在此可以指定表的分片规则。
- DataNode：MyCAT的逻辑数据节点，是存放table的具体物理节点，也称之为分片节点，通过DataSource来关联到后端某个具体数据库上
- DataSource：定义某个物理库的访问地址，用于捆绑到Datanode上
Mycat读写分离
MyCat的读写分离是建立在MySQL主从复制基础之上实现的。
数据库读写分离对于大型系统或者访问量很高的互联网应用来说，是必不可少的一个重要功能。对于MySQL来说，标准的读写分离是主从模式，一个写节点Master后面跟着多个读节点，读节点的数量取决于系统的压力，通常是1-3个读节点的配置
Mycat读写分离和自动切换机制，需要mysql的主从复制机制配合。