This document describes how to select a CVM model that suits you best from the aspects of features and scenarios, and provides detailed instructions and best practices.
A region is the geographical location of the cloud computing resource you purchase. It directly determines the network conditions for accessing the resource.
Note that there are regional differences on network quality, compliance policy and image use limits. For example, switching between Windows and Linux is only available in the Chinese mainland.
Each region has one or more availability zones. The available CVM instance types may vary with availability zone in the same region. The resource interconnection may suffer different network latency between availability zones.
For more information on regions and availability zones, see Regions and AZs.
Tencent Cloud provides various instance types. Each type supports multiple instance specifications. CVM types can be divided into x86, ARM, bare metal, heterogeneous computing (GPU/FPGA), and BatchCompute according to the architecture. They can also be classified into Standard, Computing, Memory Optimized, High I/O, Big Data according to features. This document takes the second classification method as detailed below.
Standard instances provide a balanced performance. They are suitable for most applications, such as websites and application integration. Standard instances mainly fall into the following families:
This family provides the highest single-core computing performance, making it suitable for compute-intensive applications such as batch processing, high performance computing, and dedicated game servers. It is also applicable to other compute-intensive services such as high-traffic Web frontend server and massively multiplayer online (MMO) game servers.
This family features a large memory, with a CPU/RAM ratio of 1:8 and the lowest price per GB of memory among instance types. It is suitable for applications that require memory-intensive operations, searches, and computations, such as high-performance databases (MySQL and Redis) and distributed in-memory caching.
This family uses local disk as the data disk. It is equipped with the latest NVME SSD storage featuring high random IOPS, high throughput and low latency. It provides an ultra-high IOPS at a low cost, making it suitable for I/O intensive applications that require fast disk I/O operations and low latency, such as high-performance relational databases, and Elasticsearch.
IT instances use local disk as the data disk, which may lose data (e.g., when the host crashes). If your application cannot guarantee data reliability, we recommend you choose an instance that can use cloud disks as the data disk.
This family is equipped with massive storage resources, features high throughput, and is suitable for throughput-intensive applications such as Hadoop distributed computing, massive log processing, distributed file systems, and large data warehouses.
D instances use local disk as the data disk, which may lose data (e.g., when the host crashes). If your application cannot guarantee data reliability, we recommend you choose an instance that can use cloud disks as the data disk.
This family is equipped with heterogeneous hardware such as GPU and FPGA to deliver real-time, fast parallel computing and floating-point computing capabilities. It is suitable for high-performance applications such as deep learning, scientific computing, video encoding/decoding, and graphics workstations.
NVIDIA GPU instances use NVIDIA Tesla GPUs, including the current mainstream choice T4/V100, and the latest generation V100, to provide excellent general-purpose computing capability, making this the top choice for applications such as deep learning training/inference and scientific computing.
Cloud Physical Machine (CPM) 2.0 is an elastic high-performing bare metal instance built on the latest virtualization technology of Tencent Cloud. It combines the elasticity of a virtual machine with the stability of a physical machine. It can be integrated seamlessly with all Tencent Cloud services such as networks and databases, and suitable for standard, high I/O, big data, and heterogeneous computing scenarios. CPM 2.0 supports third-party virtualization platforms and can help you build dedicated and isolated high-performance physical server clusters rapidly. With the nested virtualization technology, it also supports efficient and advanced hybrid cloud deployment with AnyStack.
High-performance computing cluster (HPC) is a cloud computing cluster that uses CPM 2.0 as compute nodes and provides high-speed RDMA network connection. It can be widely used in large-scale computing scenarios such as automotive simulation, fluid dynamics, and molecular dynamics. It also provides high-performance heterogeneous resources to support scenarios including large-scale machine learning and training.
|Use Case||Common Software||Description||Recommended Model|
|The Web service generally covers personal website, blog, and large-scale ecommerce website. This use case requires a balance of compute, storage, and memory resources. We recommend Standard instances.||Standard S and SA|
|The message queue service requires relatively balanced compute and memory resources. We recommend Standard instances using cloud disk as storage.||Standard S
|Database||MySQL||The database business requires extremely high I/O performance. We recommend instances using SSD cloud disks and local disks. When selecting an instance using local disk, remember to back up data to avoid data loss.||High I/O IT
Memory Optimized M
|The cache business has a high requirement for memory and moderate requirement for computing performance. We recommend Memory Optimized instances for the high CPU/RAM ratio.||Memory Optimized M|
|The big data business requires mass storage and moderate I/O throughput. We recommend Big Data instances. When selecting an instance using local disk, remember to back up data to avoid data loss.||Big Data D|
|High performance computing||StarCCM
|The high performance computing business requires both the ultimate single-machine computing power and the efficient multi-machine scaling. We recommend HPC with the high-speed RDMA network connection or Compute instances.||HPC
|The virtualization application requires the nested virtualization of a cloud server without incurring additional performance overhead while maintaining the virtualization capability like a physical machine. We recommend CPM 2.0 products.||HPC
|The video rendering business requires supporting graphic and image processing APIs, such as DirectX and OpenGL. We recommend GPU Rendering GN7vw instances.||GPU Rendering GN7vw|
|The AI-based computing business requires the parallel processing capability, high GPU computing power, and video memory.||GPU Computing
You can purchase other Tencent Cloud services to work with CVM instances as needed. This document takes building a website as an example to describe the associated Tencent Cloud services.
You can also select other Tencent Cloud services to meet your specific requirements. For example, after deploying applications, you can use the following Tencent Cloud services to implement disaster recovery to ensure the system robustness and provide data security: