Dual-AZ HA Design

The dual-AZ HA solution is the most commonly used solution on the public cloud. The four-layer architecture (access, application, middleware, and data layers) of applications are recommended for end-to-end dual-AZ deployment.

Figure 1 Dual-AZ HA design

Key design points:

• Service module: For services supporting cluster deployment, resources are deployed in two AZs, where loads are balanced by ELB. For single-node ECSs, SDRS is used for AZ-level DR.
• Cloud service HA: Primary and standby nodes are deployed in different AZs.
• Database synchronization: Primary/Standby RDS instances are deployed across AZs, where data can be synchronized.
• DR switchover: If an AZ fails, RDS database services automatically switch to the standby databases. Application services can be taken over by the DR servers automatically or in just a few clicks.
• DR drill: You can switch over applications or use the DR drill function of SDRS.
  

  For applications that are highly sensitive to latency, such as during the cloud migration of Telecom's NFV NEs, it is essential to conduct thorough testing, validation, and optimization. This ensures optimal performance and a seamless user experience of dual-AZ deployment on the public cloud.

  • Selecting appropriate AZs

    Cloud service providers provide the physical locations and network latency information of each AZ. To implement HA across AZs, select AZs that are close to each other to reduce the network latency and improve the application response speed.

  • Verifying latency

    Before implementing the dual-AZ HA solution, you need to fully test and verify the latency of applications across AZs. By conducting a simulated stress test, you can measure the latency of application calls across AZs to assess whether the service meets performance requirements and make informed decisions based on the results.