Tencent Distributed SQL

You can automatically split a large table horizontally (i.e. sharding) in the database by setting the shardkey when the table is created. The TDSQL system will automatically distribute the written data evenly to different physical shards based on the hash scheme, and queries will also be automatically aggregated to return results. The sharding process is imperceptible to users, they simply see the complete table and do not need to concern themselves with the backend workings.

TDSQL is compatible with frequently used MySQL syntax elements, including MySQL's language structures, character sets, time zones, data types, common functions, preprocessing protocols, sorts, joins (JOIN), indices, partitions, transactions, control commands as well as common DDL, DML, and DCL commands and database access APIs.

Due to the particularity of the distributed architecture, it is virtually impossible to balance both advanced functionality and high performance for databases. To address this issue, TDSQL provides three table creation solutions, featuring (distributed) transaction characteristics, globally unique numerical sequences, and JSON format support, effectively filling the distributed architecture gaps and enabling more flexible development solutions.

At present, one single shard can sustain up to 6 TB of storage. If the performance or capacity is insufficient, you can easily scale up with just a few clicks in the console. You do not need to worry about data migration, balancing, and route switching within the distributed system during the migration process, the IP stays unchanged, and the service will be interrupted for just a few seconds upon automatic switch. You only need to make sure that your business has a reconnection mechanism.

TDSQL adopts a master-slave architecture by default, which ensures over 99.95% availability. It supports strong synchronous replication for strong data consistency where an application transaction is responded to only after the data written by the business system is synced to the slave, ensuring complete master-slave data consistency and eliminating the possibility of data loss or corruption. At present, the performance of strong synchronous replication is basically equal to that of asynchronous replication.

Based on deeply customized and optimized MySQL kernel, TDSQL greatly outperforms MySQL. It supports read-write separation in three schemes, effectively ensuring development flexibility while providing read extensibility. It boasts optimized thread pool scheduling algorithms with better heavy load performance. In addition, it is equipped with PCI-E SSDs that feature an IO throughput three times of SATA disks, making it easier to meet the performance needs of your business.

A production-ready TDSQL instance can be started and connected to in minutes with just a few clicks in the TDSQL console. The console comes with frequently used database OPS operations that facilitate refined management of your database. Moreover, it is capable of displaying common system monitoring data and performance analysis data to help you quickly identify and troubleshoot database issues.

TDSQL supports native JSON types. Specifically, it doesn't store JSON values as strings; instead, it uses an internal binary format that allows for quick reading of document elements, which will automatically validate JSON texts when they are inserted or updated in JSON columns. In other words, you can think of TDSQL as a NoSQL database compatible with the MySQL protocol.

RocksDB is a storage engine based on LSM tree. Its data compression rate is over 60% higher than that of the InnoDB engine. Generally, it only needs about 1 GB storage space for 10 million rows of data after compression, which greatly reduces the database usage costs. Plus, it features high write performance, making it suitable for space-sensitive business scenarios where there are more writes than reads, such as IoT and logging. RocksDB support is currently in beta.