Selection Guide
Last updated: 2025-11-24 20:39:07
Before purchasing a TencentDB for MySQL instance, it is important to learn about the features of different instances so that you can select one that meets your business requirements.
Instance Information
Prior to purchasing a TencentDB for MySQL instance, you must consider factors such as cost, performance, workload, and usage scenarios, so that you can select a suitable instance with optimal costeffectiveness. Since the database storage engine, instance architecture, storage type, and resource isolation policy are intricately intertwined and are mutually impacted, you may be confused when selecting an instance. This document briefly introduces these components to assist you in selecting an appropriate instance.
1. Database Storage Engine
A storage engine refers to the type of tables. The storage engine of database determines the manner in which tables are stored in a computer.
InnoDB: The most frequently used OLTP storage engine, uses multi-version concurrency control (MVCC) and row-level locking technologies, offering high performance and reliable processing capabilities. In comparison to other MySQL storage engines, InnoDB offers functions including data foreign key and rollback, ensuring better data integrity. It also provides higher-level query functionalities. InnoDB kernel has been optimized a lot by Tencent Cloud and therefore has great performance advantages and is extensively applied in scenarios that involve high concurrency and require high performance.
RocksDB: A popular high-performance persistent key-value (KV) storage based on which Tencent's TXSQL team develops TXRocks, a transactional storage engine. TXRocks, benefitting from the LSM Tree storage structure of RocksDB, has significantly reduced the waste caused by InnoDB's half-full page and fragmentation mechanisms, and utilized compact format storage. In this way, compared with InnoDB, TXRocks provides similar performance with half the storage space. It's more suitable for business scenarios that involve a large volume of data and require high read and write performance.
LibraDB: A self-developed OLAP storage engine. With capabilities such as columnar storage, large-scale concurrency execution, and the vectorized execution engine, LibraDB accelerates various complex and time-consuming SQL queries such as complex queries, slow SQL queries, and fuzzy matching in the business, effectively improving the overall SQL execution efficiency. LibraDB is suitable for use cases such as real-time reporting, online analysis, and HTAP. Currently, only read-only instances support the LibraDB engine.
2. Instance Architecture
TencentDB for MySQL supports four types of instance architectures: single-node, two-node, three-node, and Cluster Edition.
Architecture | Description | Applicable Scenarios |
Single-Node | Supported versions: MySQL 5.7 and 8.0. Node: Single Node. | Personal learning, micro-websites, non-core small-scale enterprise systems, and the development and testing environments of large and medium-sized corporations. |
Two-Node | Supported versions: MySQL 5.6, 5.7, and 8.0. Nodes: One Primary and One Standby. Primary-Standby Replication Mode: Asynchronous (default) and Semi-Synchronous | Gaming, internet, IoT, retailing e-commerce, logistics, insurance, and securities, etc. |
Three-Node | Supported versions: MySQL 5.6, 5.7, and 8.0. Nodes: One Primary and Two Standbys. Primary-Standby Replication Mode: Asynchronous (default), Strong-synchronous, and Semi-Synchronous | Gaming, internet, IoT, retailing e-commerce, logistics, insurance, and securities, etc. |
Cluster Edition | Supported Versions: MySQL 5.7, 8.0. Nodes: One read-write node and up to five read-only nodes. Primary-Replica Replication Method: Asynchronous, and semi-synchronous (default). | Gaming, internet, IoT, retailing and e-commerce, logistics, insurance, and securities, etc. |
3. Storage Classification
The underlying storage of TencentDB for MySQL supports local SSD disks, SSD cloud disks, Enhanced SSD cloud disks, and Tremendous SSD cloud disks.
Performance metrics | Tremendoust SSD Cloud Disk | Enhanced SSD Cloud Disk | Premium Disk | SSD | Local SSD |
Maximum single disk capacity (GB) | 32000 | 32000 | 32000 | 32000 | 12000GB |
IOPS per Disk | After adding extra capacity, the performance reaches up to 1,000,000. | Additional performance overlay reaches 100,000 | 6000 | 26000 | 150000 |
Random IOPS Calculation Formula | Benchmark performance
:Random IOPS = min{4000+Capacity(GiB)×100, 50000}
Additional Performance
:Maximum IOPS =
min{Additional Performance Value×128, 950000} | Benchmark performance :Random IOPS = min{1800+capacity(GiB)×50, 50000} Additional performance :Maximum IOPS = min{Additional performance value×128, 50000} For details, see Enhanced SSD Cloud Block Storage Performance Description | Random IOPS = min{1800 + Capacity (GiB) x 8, 6000} | Random IOPS = min{1800+capacity(GiB)×30, 26000} | |
Maximum single disk Throughput (MB/s) | After adding extra capacity, the performance reaches up to 4000 MB/s. | Additional performance overlay achieves 1000MB/s | 150MB/s | 260MB/s | - |
Throughput performance calculation formula (MB/s) | Benchmark performance :Throughput = min{120+Capacity(GiB)×0.5, 350} Additional Performance
:Throughput = min{Additional Performance Value×1, 3650} | Benchmark performance :Throughput = min{120+Capacity(GiB)×0.5, 350} Additional performance: Throughput = min{Additional performance value×1, 650} For details, see Enhanced SSD Cloud Block Storage Performance Description | Throughput = min{100 + Capacity (GiB) x 0.15, 150} | Throughput = min{120+Capacity(GiB)×0.2, 260} | - |
Single-lane random read and write latency (ms) | 0.1ms - 0.5ms | 0.2ms - 1ms | 0.8ms - 5ms | 0.5ms - 3ms | Microsecond level |
4. Resource Isolation Policy
The isolation policies of TencentDB for MySQL include basic, economical, general, exclusive, standard, enhanced, and flagship types.
Resource Isolation Policy | Description |
Basic Type | Only single-node instances support the basic isolation policy (formerly basic edition), where there is a separation between computing and storage, with the underlying layer using cloud disk storage. |
Economical | Exclusive memory and disk allocation for instances. CPU resources are shared among general specification instances located on the same physical server.* Fixed instance specifications and disk capacity. Suitable for lightweight and low-load use cases such as small websites, Web applications, blogs, forums, and cloud development/test/learning environments, offering excellent cost performance. |
General Type | An instance exclusively utilizes allocated memory and disk resources while sharing CPU resources with other general instances on the same physical machine. * Benefit from resource sharing, bringing higher cost-effectiveness and minor CPU resource reutilization. |
Dedicated Type | An instance has dedicated CPU (with core binding), memory, and disk resources. It promises long-term performance stability and remains unaffected by the behavior of other instances on the physical machine. The peak configuration of the dedicated type is to occupy a physical machine alone, taking full control over all its resources. |
Standard Type | Exclusive allocation of CPU and memory ensures long-term stable performance. A decoupled architecture of computing and storage offers flexible configuration options. |
Enhanced Type | Exclusive allocation of CPU and memory ensures long-term stable performance. A decoupled architecture of computing and storage offers flexible configuration options. Supports Tremendous SSD cloud disk, providing stable and reliable performance. |
Flagship | CPU core with a higher frequency, offering outstanding performance. Exclusive CPU and memory allocation for instances, with long-term stable performance. Storage-computing separation architecture with flexible configurations. Tremendous SSD is supported, providing stable and reliable performance. |
*In extreme cases, resource contention may occur (extremely low probability) for the general isolation policies.
Product Selection
You can follow the following steps to select an instance:
1. Selecting Database Storage Engine
If you need full transaction support and strong read-write concurrency, InnoDB is recommended. To reduce storage costs, RocksDB is recommended, which can save up to half or more of the storage space compared to InnoDB while offering similar performance. For use cases such as real-time reporting, online analysis, and HTAP, it is recommended to add the LibraDB read-only analysis engine after you create two-node or three-node architecture instances, effectively improving the overall SQL execution efficiency.
2. Selecting Instance Architecture
Generally, a two-node architecture employing a classic primary-standby high-availability architecture is recommended. It is appropriate for industries like Internet, IoT, retailing e-commerce, logistics, gaming, or medium to large corporations.
If your operations demand a financial-grade reliability, security, high availability, and disaster recovery abilities typical for industries such as finance, securities, insurance, or core database of large corporations, we recommend a three-node architecture.
If your business is complex, involves frequent changes, handles large data volumes, requires high read performance, and needs frequent scaling or addition/deletion of read-only instances, while also requiring high reliability, security, availability, and disaster recovery capabilities, Cluster Edition is recommended.
For personal learning, miniature websites, non-core small-scale enterprise systems, or the development and testing environments of large and medium-sized corporations, a single-node architecture would be a better choice.
3. Selecting Storage Type
In terms of storage types, the instances with two-node or three-node architecture currently support local SSD; the single-node architecture instances support Cloud SSD, Premium Disk, and Enhanced SSD; the Cluster Edition architecture instances support Tremendous SSD, Enhanced SSD, Premium Disk, and Cloud SSD.
The cloud disk architecture single-node instances, based on a cloud-native architecture, are suitable for testing, development, personal learning and other scenarios. They support up to 30 TB storage space. The storage size has an impact on IOPS.
4. Selecting Resource Isolation Policy and Instance Specifications
The single-node architecture supports the basic isolation policy; the two-node architecture supports economical, general, and exclusive isolation policies; the three-node architecture supports general and exclusive isolation policies; the Cluster Edition architecture supports standard, enhanced, and flagship isolation policies. Parameters of instance specifications include vCPU, memory, maximum IOPS, and storage capacity, allowing you to choose based on your business needs.
Note：
Related Documents
Was this page helpful?
You can also Contact Sales or Submit a Ticket for help.
Yes
No
Feedback