Azure High Availability Architecture Design (LB and Traffic Manager)

Azure is one of the representatives of more and more companies considering moving their business to the public cloud platform. However, even if Microsoft continues to emphasize the excellent high availability of azure, as an enterprise IT administrator, we must further understand and properly configure the high-availability features of the platform to ensure the security of the system on the cloud.

[正文]

I. High-availability architecture design principles
Designing a high-availability architecture on the Azure public cloud is based on the following principles: Separating the front-end from the database to avoid concentration of the front-end role Balanced, you can use the application's own technology, or network load balancing traction: LB (Load Balancing) load balancing: for the same cloud service, and limited to the same area (Shanghai or Beijing area) Azure Traffic Manager: failover or Load balancing; full cloud or local hybrid database uses its own high-availability technologies, such as active/standby replication, failover, and dual-active: MySQL master-slave replication Oracle RAC Note: The following example diagram only shows the idea, the application and technology used by the role. Can be replaced by actual

Second, Azure LoadBalance (for the same cloud service)
Azure LoadBalance is actually polling different address ports within the same cloud service. Taking a standard application structure as an example, including front-end and database roles, the high-availability design will be analyzed from two aspects. a. Front-end: In the same cloud service (publish the same domain name such as www.contoso.com), set up multiple front-end servers (such as SVR1 and SVR2), both of which can access the back-end database to enable LoadBalance for the cloud service (load Equilibrium), if 80-port polling is specified, the range will be assigned to different front-end servers when the range www.contoso.com:80. b. Database: Stored in the cloud or enterprise data center, and use the master/slave synchronization technology of the database itself. Copying

3. Azure Traffic Manager (between multiple cloud services)
Use Microsoft Azure Traffic Manager to control the distribution of user traffic to specified endpoints (which may include Azure cloud services, websites, and other endpoints) . Traffic Manager works by applying an intelligent policy engine to Domain Name System (DNS) queries that are performed on Internet resource domain names. Azure cloud services or websites can run in different data centers around the world.

3.1 Mode 1: Failover + Local Database
a. During normal access: With Traffic Manager failover mode, users access the Azure cloud service's front-end server by default on the Internet (and the cloud service has LB enabled) At this time, the local front-end server is in standby state, and does not provide actual service. Both of them allow access to the database of the local data center. The primary node database uses its own active/standby synchronization technology to replicate data, ensuring that the primary node can be switched at any time when the primary node fails. When accessing the standby node b. Fault access: The front-end server that the user cannot access the Azure cloud service on the Internet Azure Traffic Manager automatically switches the access to the locally released front-end server. The local front-end server immediately accesses the database of the local data center. The active/standby synchronization technology replicates data to ensure that the primary node fails to switch at any time. When the primary node fails, it can also switch to the standby node access.

3.2 Mode 2: Load Balancing (Full Cloud) + Local Database
a. During normal access: Using Traffic Ma Nager cycle or performance mode, users can take turns accessing (or proximity) different Azure cloud services on the Internet, such as a Shanghai area cloud service, a Beijing area cloud service, two cloud services are enabled with LB load balancing two cloud services The front-end server can access the database database of the data center to replicate data by using its own master-slave synchronization technology to ensure that the master node can switch at any time when the master node fails. When the master node fails, it can also switch to the standby node to access b. When failover: When a cloud service appears Failure, Traffic Manager will automatically direct all traffic to another cloud service. The front-end server of the cloud service continues to access the database database of the data center. It uses its own master-slave synchronization technology to replicate data, ensuring that the master node can be switched at any time when the master node fails. Switch to standby node access

3.3 Mode 3: Failover + Offsite Database
a. Normal access: With Traffic Manager failover mode, users access the Azure cloud service front-end server by default on the Internet (and the cloud) Service LB) The database master node is also deployed in Azure, providing direct access to the front-end server on the cloud. The local front-end server is in standby state. The front-end server that does not provide the actual service directly accesses the local database standby node without accessing the cloud. The master node master database uses its own master-slave synchronization technology to replicate data to the standby node. b. When the fault occurs: When the Azure cloud service fails, it means that the front-end and the database on the cloud may not work properly. At this time, the Traffic Manager failover and database are utilized. The active/standby switchover function, switching both to the local to continue to provide services

3.4 Mode 4: Load Balancing (Full Cloud) + Offsite Database
a. Normal Access: Using Traffic Manager Loop or Performance Mode, User On the Internet, you can take turns to access (or the nearest principle) different Azure cloud services, such as a Shanghai area cloud service, a Beijing area cloud service, two cloud services are enabled LB load balancing database, the primary node is deployed in one of the Azure areas (such as Shanghai) , providing the same in the same area The server directly accesses the database standby node deployed in another Azure area (such as Beijing), provides the front-end server in the same area to directly access the database to enable the main master replication mode, real-time synchronization and can provide services b. Fault access: When A cloud service failure in the Azure area (such as Shanghai) means that the cloud front end and database may not work properly. Traffic Manager automatically transfers all traffic to the normal area of ​​the cloud service (such as Beijing), and will continue to provide services.