TencentDB for TDSQL is a distributed database service deployed in Tencent Cloud that supports automatic sharding (horizontal splitting) and the Shared Nothing architecture. With a distributed database, your business obtains a complete logical database table which is split and distributed evenly across multiple physical shard nodes on the backend. TDSQL deploys the master-slave architecture by default and provides a full set of solutions for disaster recovery, backup, restoration, monitoring, and migration, making it ideal for storing terabytes to petabytes of data.
TDSQL is a distributed database for OLTP businesses.
TDSQL is a distributed database that supports sharding.
Sharding is to spread the data of a table into multiple independent physical database servers according to a defined rule to form an "independent" database "shard". Multiple shards together form a logically complete database instance.
The Shared Nothing architecture can scale data and access capacity by simply adding more servers. Although it can meet the database capacity needs of most users, it is essentially a minicomputer + shared storage solution that is prone to capacity and performance bottlenecks and relatively expensive.
In principle, TDSQL uses a sharding scheme based on automatic horizontal splitting. Specifically, a modulo operation is executed on the shardkey, and then data is distributed into different databases through TProxy according to the specific range of values after modulo operation.
A relational database is a two-dimensional model. To shard data, it is usually necessary to find a shardkey field to determine the sharding dimension. Then, a rule needs to be defined to actually shard the database.
Below are some common shardkey options:
Based on date order, such as sharding by year (one shard for 2015 and another for 2016).
Based on user ID modulo, where fields in the specific range after modulo operation are spread across different databases.
Based on primary key modulo, where fields in the specific range after modulo operation are spread across different databases.
Before sharding multiple tables, the following options are available:
In terms of sharded data source management, there are currently two modes:
Faced with millions of users of the internet-based business, a standalone database will reach bottlenecks in data storage capacity, access capacity, and disaster recovery due to hardware and software limitations as the business grows.
Application-layer sharding highly couples business logic with database logic, which incurs heavy development workload over rapid iteration of the current business. Based on the imperceptible sharding scheme of TDSQL, your developers only need to modify the code during initial access without having to care much about the database logic during subsequent iterations, which can greatly reduce the development workload.
Choosing open-source or NoSQL solutions can also break through database bottlenecks at no or relatively low costs. However, you need to pay attention to the following issues with such solutions: