lingducloud: Alibaba Cloud Database Nanny-level Purchase Guide (Engine, Architecture, and Storage)

2026-04-24 阅读 46
lingducloud: Alibaba Cloud Database Nanny-level Purchase Guide (Engine, Architecture, and Storage)
1

The choice of database is also very important in the purchase of cloud products.

Database selection is not only the choice of CPU and memory, but also the choice

Reliability

with

IO Throughput

This guide will take you from the database engine, architecture, storage three dimensions, dismantling the most suitable for your solution.

Step 1 of 1.: Select "Engine" (type according to business logic)

Database Engine

Core Label

The most suitable business

Expert advice

MySQL

Internet standard

E-commerce, CMS, applets, most Web applications

Select 8.0 version. Do not use 5.7 unless mandated by old code (community support has been discontinued).

PostgreSQL

Functional All-round

GIS geographic information, complex financial computing, scientific research

There are a large number of complex stored procedures and geospatial queries in business logic.

SQL Server

Microsoft Family Barrel

. NET development environment, old ERP/OA, medical/government system

Companies that rely heavily on Windows ecology choose it, but pay attention to its higher License costs.

MariaDB

Open Source Enhancements

Specific scenarios that pursue extreme open source and require advanced MySQL features

At present, the domestic ecology is weaker than MySQL, and it is recommended to look back at MySQL for general business.

2. the second step: select "series" (according to the stability of the framework)

Basic-"single-machine" architecture: only a single node, no redundancy. Scenario: personal learning, development and debugging, non-profit small test station. Warning: Do not use it for official business! As long as the underlying physical hardware fails, the business will be interrupted and there will be no high availability switching.

High Availability Edition (High-Availability)-"Enterprise Main Force" Architecture: One Master and One Standby, Automatic Switching in Seconds. Scenario: First choice for 90% of enterprise production environments. E-commerce, SaaS, API backend. Advantages: It has a complete backup and recovery mechanism and supports read-only instance expansion.

Cluster Edition (Cluster)-"performance ceiling" architecture: multi-master or multi-slave cluster. Scenario: large-scale Internet platform, double 11 level promotional activities. Advantages: Very high throughput capacity, performance increases linearly with the increase of nodes.

3. Step 3: Select "Store" (Resolve Database Caton)

90% of the database performance bottleneck is not in the CPU, but in the disk I/O (read and write speed).

Storage type

Performance Levels (IOPS)

Applicable Scenarios

cost performance evaluation

Premium ESSD

Beginner level

Low load official website, lightweight application

The most economical, suitable for concurrent insensitive business.

ESSD PL1

mainstream standard

Medium-sized e-commerce, standard enterprise systems, core business start

Production environment cost-effective choice, can handle most of the sudden traffic.

E

SSD PL2

High performance

High-frequency trading, spike scenarios, large ERP

Very fast read and write for I/O intensive business.

ESSD PL3

Extreme Performance

Financial-grade accounting, very large and highly concurrent systems

Businesses with adequate budgets and extremely demanding delays.

4. Gold Matching Scheme (Quick Comparison Table)

Business scenarios

Recommendation Engine

Recommended Architecture

Storage recommendations

Company official website/WordPress

MySQL 8.0

High Availability Edition (1 Core 2G/2 Core 4G)

Premium ESSD

Standard e-commerce/SaaS backend

MySQL 8.0

High availability version (4-core 8G/8-core 16G)

ESSD PL1

Complex Reports/Internal ERP

PostgreSQL

High availability version (8 cores 32G)

ESSD PL1 / PL2

Core Finance/Seckill System

MySQL / SQL Server

Cluster Edition

ESSD PL2 / PL3

Development/pre-release environment

Casual

Basic Edition

Premium ESSD

5. Key Supplement: How to Choose Region and Availability Zone?

Many people choose regions (Region) casually and feel the same everywhere. In fact, choosing the wrong region will not only slow down the user's access, but also may lead to extremely high latency between the server and the database, or even the inability to communicate.

1. Core principle: Nearby access

Choose where the target users are: * Core customers in the south (Guangdong, Fujian, etc.) choose South China 1 (Shenzhen) or East China 2 (Shanghai). Core customers in the north (Beijing, Hebei, etc.) choose North China 2 (Beijing). Core customers in China to choose East China 1 (Hangzhou) (Ali base camp, network hub, the most stable infrastructure).

Overseas business: Southeast Asian users choose Singapore. European and American users choose the United States (Silicon Valley/Virginia) or Germany (Frankfurt). Need to be exempt from filing and take into account the domestic visit preferred Hong Kong, China.

2. Core Pit Avoidance: ECS and Database Must Be "Same City"

This is the easiest mistake to make!

Principle: Your ECS cloud server and RDS database must be in the same region (for example, both in Hangzhou).

Reason: Only in the same region can the two be connected through the intranet. Intranet access is free and has very low latency (typically $<1ms$).

Consequences: If one chooses Beijing and the other chooses Shanghai, you can only connect through the public network, not only will you have to pay extra traffic charges, but the database response will become extremely slow and the business will be basically unavailable.

3. Advanced Strategy: Choice of Availability Zone (AZ)

Each region has multiple zones (such as Hangzhou zones A, B, C...), which are physically isolated from each other.

Pursue fast response: put ECS and database in the same availability zone. Closest physical distance, network extension

Delayed to a minimum.

High reliability (multi-data center disaster recovery): When purchasing a database, select multi-zone deployment. In this way, the main database is in computer room A and the standby database is in computer room B. Even if the street where room A is located is powered off or an accident occurs, the database can be switched to room B in seconds without stopping the business.

Logical Summary

Dimension

Recommended choice

Geographic location

The place physically closest to your core customer base.

Intranet interworking

The ECS instance and the database must be in the same region (for example, Beijing).

Disaster Recovery Requirements

For production environments, we recommend that you choose Multi-AZ deployment.

Filing compliance

Nodes in mainland China must be filed; nodes in Hong Kong and overseas do not need to be filed.

6. pit avoidance and value-added advice

Don't be cheap in architecture selection: the production environment must choose "high availability version". Don't take the risk of downtime to save that tens of dollars.

ARM architecture is a new trend: Aliyun's ARM version (relying on heaven) database is usually more than 20% cheaper than x86, and its performance is very stable in MySQL and other scenarios. It is recommended to try new projects.

Intranet connection is the most important: the database and ECS server must be in the same region and VPC (private network), so that they can be accessed through the intranet with the lowest delay and free of traffic charges.

Pay attention to monitoring alarms: Be sure to configure alarms for CPU usage and disk space after purchase. The database is most afraid of "slow disk" and "full space".

2
← 返回新闻中心