Selecting ECS Specifications

Scenarios

Huawei Cloud provides various ECS specifications to meet different service needs. Different types of ECSs provide different computing and storage capabilities. For the same ECS type, you can select different ECS specifications based on the number of vCPUs and memory size.

This section describes how to select ECSs suitable for your service requirements.

Specifications

1. To learn details about the available ECS specifications, see A Summary List of x86 ECS Specifications.
2. To learn how ECS flavors are named, see ECS Flavor Naming Rules.
3. To learn more about common metrics, see vCPUs and network QoS.

Specification Selection

• By type

  View the ECS specification details to learn about the major ECS types, computing and network performance, and supported disk types. You can select ECS specifications based on different properties. ECS supports the following CPU architectures: x86 and Kunpeng. For details, see x86 and Kunpeng Architectures.

  This section is intended for users who are familiar with the CPU architecture, vCPUs, memory, and instance family and generation of ECSs and want to select specific specifications.

  x86 ECSs

  

  Kunpeng ECSs

  

  On the ECS purchase page, you can select an ECS by specifying its vCPUs, memory, or flavor name.

  

• By scenario

  This section is intended for users who have specific service requirements.

  Category	Sub-category	Characteristics	Recommended ECS (Example)
  Web applications	Traditional office	High security and reliability, suitable for traditional office scenarios like OA, ERP, and CRM with less than 200 concurrent access requests	C7 and C6
  	Enterprise websites	A balance of compute, memory, and network resources with a baseline level of vCPU performance and high cost-effectiveness	C7 and C6 S7 and S6
  	Personal application setup	A balance of compute, memory, and network resources with a baseline level of vCPU performance and high cost-effectiveness	S7 and S6 T6
  	Development and testing	A balance of compute, memory, and network resources with a baseline level of vCPU performance and the ability to provide burst CPU power at any time for as long as required	S7 and S6
  	Front-end servers	A balance of compute, memory, and network resources with a baseline level of vCPU performance. These ECSs can be used as front-end servers like Apache, Nginx, and IIS	S7 and S6
  	Back-end servers	High vCPUs-to-memory ratios, high performance, and low latency These ECSs are cost-effective options for back-end servers like Tomcat and JBoss	C7, C6, and C6s
  Website applications/E-commerce	100,000 pageviews/1,000 active users	Cost-effective, flexible, elastic resources available anytime	S7 and S6 T6
  	200,000 pageviews/2,000 active users	Suitable for e-commerce websites, which require high-performance cloud servers with fast elasticity and high stability to handle traffic bursts typical of special promotions, flash sales, and live commerce	C7, C6, and C6s S7 and S6
  	500,000 pageviews/5,000 active users	Suitable for e-commerce websites, which require high-performance cloud servers with fast elasticity and high stability to handle traffic bursts typical of special promotions, flash sales, and live commerce	C7, C6, and C6s
  Gaming	Gaming	Suitable for gaming services, which require high performance, high stability, high cost-effectiveness, and low latency	C7, C6, and C6s
  Databases	Compute	Stable, high-performance compute	C7, C6, and C6s
  	Storage	Servers that use local disks with high storage bandwidth and IOPS to provide cost-effective mass data storage	M7 and M6 D7 and D6
  	Network	High PPS performance, high TPS throughput, and low network latency for rapid data exchange and processing	E7 and E6 M7 and M6 C7, C6, and C6s
  Data analytics	Management nodes	A large volume of compute resources scheduled to accelerate data processing	C7, C6, and C6s
  	Compute nodes	Balanced compute with high performance and stability	M7 and M6 C7, C6, and C6s I7, Ir7, I3, and Ir3 S7 and S6
  	Storage nodes	Cost-effective, high-bandwidth storage for processing large amounts of reads and writes	I3 and Ir3 D7, D6, and D3
  Image rendering	Animation rendering	CPU-accelerated rendering with high precision and stability	aC7, C6, and C3
  	Video rendering	GPU-accelerated rendering with high processing speed	G5 and G6
  AI/Machine learning	AI training	Compatible with NVIDIA smart NICs for deep learning training, scientific computing, computational fluid dynamics, computational finance, seismic analysis, molecular modeling, and genomics	P2vs, P2v, and P2s
  	AI inference	Compatible with NVIDIA smart NICs for image classification and recognition, speech recognition, natural language processing, video encoding and decoding, machine learning, and lightweight training	Pi2

  On the ECS purchase page, you can select specifications based on service scenarios.

  

• By pre-installed software

  Select an ECS based on the pre-installed software you need.

  Scenario	Application	Requirement	Recommended ECS (Example)
  Cache	Redis Memcached	Low CPU computing capability Very high memory	M7 and M6
  Big data	Spark Hive	High CPU computing capability High memory bandwidth I/O: high storage bandwidth	D7i, I7, and D7
  Load balancing	Nginx	Very high CPU computing capability	C7, C6, and X1
  Database	MySQL NoSQL	Lightweight database	aC8 and C7
  Real-time computing	Flink	You can select general-purpose ECSs or cloud disks based on the storage capacity. You can also select disk-intensive ECSs.	I7 and D7
  Offline computing	Hadoop HDFS	vCPU to memory ratio: 1:4 (preferred)	D7i and D7
  Message queue	Kafka RabbitMQ	vCPU to memory ratio: 1:1	aC8, C7, and X1
  Text search	ElasticSearch	Large vCPU-to-memory ratio	I7

• Verification and adjustment

  After selecting an ECS and using it, you can check whether the selected ECS specifications are proper based on performance monitoring of the ECS within a period of time.

  If the CPU usage of the ECS is too low, check whether the memory usage is very high. For details, see Monitoring Using Cloud Eye.

  If the memory usage of the ECS is very high, you can adjust the vCPU-to-memory ratio to an appropriate ratio. For details, see Modifying Specifications of Individual ECSs.

• Related Documents
  • For details about Cloud Eye, see Cloud Eye Function Overview.