This is the third article I do cloud native database architecture structure: Tencent Cloud TDSQL-C architecture analysis.
Cloud Native Database TDSQL-c (tdSQL-c) is short for TDSQL-c. The name of the Cloud Native Database was CynosDB. In order to unify the name, we all use TDSQL-c. Tdsql-c is a new generation of high performance and high availability enterprise distributed cloud database developed by Tencent Cloud. It integrates the advantages of traditional database, cloud computing and new hardware technologies, and is 100% compatible with MySQL and PostgreSQL. It realizes the high throughput of more than mega QPS and massive distributed intelligent storage of 128TB, ensuring the security and reliability of data.
Because external data of TDSQL-C is relatively small, here we will directly quote an official picture of customer analysis to see the architecture of TDSQL-C:
Architecturally, TDSQL-c is divided into two layers: computing and storage. MySQL and PostgreSQL are 100% compatible with both MySQL and PostgreSQL. There is no persistent storage in the computing layer. The computing layer currently supports one Primary (Read and write), and 1-15 Read nodes. From some channels to communicate with basic confirmation in 2021Q1 also realize a layer of Proxy layer unified access. The storage tier uses Tencent’s block storage COS, which is similar to Aurora’s minimalist IO model. The diagram below:
As you can see from the graph, data writes are only written to the Commit log. Logs are merged into cells after being written to the queue and merged into data blocks through log steps. In this way, the IO operation of the database is simplified, and the log is the database. At present, TDSQL-C can also support the generation of Binlog, and support the fast backfile based on Binlog.
Tdsql-c introduces continuous backup, snapshot backup, write redirection technology, and high concurrency COS upload function of independent distributed storage.
In terms of control platform, there is a corresponding concept of data autonomous driving. It also integrates DBBrain from Tencent Cloud to provide intelligent diagnosis and processing.
On the whole, TDSQL-C is positioned as a product implementation like Aurora. In terms of Tencent’s style, it must be cheap and sufficient in the early stage. In terms of performance, tdSQL-C is not necessarily the best architecture, but it should not be bad. Common high concurrency business can be held under this architecture.
Because tdSQL-C hand data is not much, can not further analysis, welcome to understand or used TDSQL-C friends to provide experience to share.
Over the last three weeks, Aurora V2, PolarDB, and TDSQL-C are all examples of cloud native databases. It’s also possible to see a lot of opportunities on the cloud native database track. It can be said that this new star TiDB is also a cloud native database start-up upstart. As technology continues to improve, we can also use the dividends of new technology to increase the efficiency and reduce the cost of our business. If you are particularly interested in cloud native database, you are also welcome to join my wechat group communication.