Tengxun cloud database selection guide (from 0 to architecture design, a clear article)
As a cloud computing agent for many years, I found that when many architects do technology selection, the most headache is to choose too many, do not know how to choose
Tencent Cloud offers an extremely comprehensive database product portfolio (just glancing at the console can leave you overwhelmed by choice), but the underlying logic is actually very clear. Today, I am not going to read the parameters. I will directly start from the architecture selection and business pain points to help you solve the problem of Tengxun cloud database selection at one time.
1. relational databases:
If your data is structured (users, orders, inventory) and requires
Transaction consistency
That relationship is the only starting point.
1. Managed CVM Database (Basic Edition)
Representative Products: TencentDB for MySQL / PostgreSQL / SQL Server
Old blogger comments: this belongs to "out of the box". You don't need to be in charge of master-slave replication, automatic backup and patching. Tencent has helped you to complete the operation and maintenance.
Selection logic: MySQL: default option for 80% Internet projects (applet, SaaS, website). PostgreSQL: Ideal for geographic information (GIS), complex analytical queries, or scenarios with strong JSONB requirements. SQL Server: . NET ecological or traditional enterprise office system "imperial" solution.
Applicable: start-up projects, smooth business, do not pursue the ultimate flexibility of the scene.
2. Cloud native database: born for "outbreak"
Representative products: TDSQL-C (original CynosDB)
Old blogger comments: this is the current mainstream trend. It separates computing and storage, and automatically expands storage space.
Selection logic: If your business has obvious traffic fluctuations (such as e-commerce promotion, sudden hot spots), do not want to manually expand the disk in the middle of the night, TDSQL-C is the most preferred.
Advantage: "The future does not need to be reconstructed". When the business goes up, it can automatically get bigger; when the business goes down, it can charge by volume.
2. NoSQL: The "high-performance plug-in" of the system"
When your MySQL starts to report slow queries and QPS reaches the bottleneck, don't rush to divide databases and tables, look at these three types of NoSQL first.
1. Caching layer: a watershed of performance
Representative Products: Redis / Tendis
Old blogger comments: As long as your system has visits, Redis is almost a must. It is not to store data, but to "block the knife" for relational databases ".
Best partners: cache hotspots, Session sharing, leaderboards, seconds kill counter.
2. Document storage: solve the "structural uncertainty"
Representative products: MongoDB
Comments by old bloggers: fields often change? Schema is not fixed? Add a field with MySQL to delaminate the skin, and using MongoDB is as smooth as writing JSON.
Best partners: content management (comments, articles), embedded log, user information extension table.
3.
Timing and Vector: A Domain-Specific "Big Kill"
Time Series (CTSDB): Process IoT device data and system monitoring metrics. Write by time, check by time.
Vector database (VectorDB): now a must for AI large models, RAG (retrieval enhanced generation). Specially store Embedding vectors.
3. Distributed TDSQL: Financial and the End of Billion Scale
Representative Products: TDSQL (Distributed Edition)
Old blogger comments: Many banks' core systems run on TDSQL. It solves the problem of single-machine performance upper limit through sharding (Sharding) and supports strong consistency transactions.
Selection logic: only consider it when you are faced with 100 million users, TB data volume, and abnormal requirements for accounting security. Ordinary business should not be easily carried out, and the complexity of the architecture will increase exponentially.
4. actual combat selection decision (quick look-up table)
Business Type
Core Portfolio Recommendations
Reason
Start-ups/small projects
MySQL Redis
Low cost, ecological maturity, operation and maintenance of the most worry-free
Rapid growth/e-commerce
TDSQL-C Redis
Separation of computing and storage, automatic expansion, and response to traffic burst
Social/Content Platform
MySQL MongoDB
MySQL stores core account number, Mongo stores dynamic/comment
Finance/Payments Core
TDSQL (Distributed Edition)
Strong consistency to solve the bottleneck of stand-alone expansion
Intelligent customer service/AI assistant
VectorDB (vector library)
Implementation of Private Knowledge Base Retrieval with LLM
5. easily overlooked "invisible life-saving grass"
DTS (data transmission): the "lifeline" of database migration and remote multi-active data synchronization ". Without it, downtime and relocation will make you stay up all night to doubt your life.
DBbrain (database governance): It is your "DBA on the cloud". Where it is slow and where the index is not added well, it will tell you directly, not let you guess.
4. Selection Formula (Direct Number Seating)
In order to make you more worry-free, I have summed up a set of "life-saving formula":
Regular start: MySQL Redis (stable as an old dog, the most cost-effective).
The pursuit of flexibility: TDSQL-C (suitable for Internet bosses who want to be dispensers).
Variable structure: MongoDB (suitable for product managers whose needs change from day to day).
Large families: TDSQL is distributed (there is too much data to be installed on a single machine to consider).
