Windows Azure failover mode and high availability mode!

Windows Azure failover mode and high availability mode!
At present, many users in China have misunderstandings about the availability of cloud services. What is the misunderstanding? For example, a cloud service provider has an IDC in a certain region in South China and an IDC in East China.

This customer mentioned a requirement. What does the 99% availability concept you provide mean? Is there a problem with my machine in the southern data center? Will my machine be automatically transferred to the East China Data Center?

From the current communication and communication with customers, it seems that most users have this idea. They think that cloud services should be highly available across regions and sites, it is hard to achieve this goal.

In the financial industry, the concept of three centers in two locations often exists. In the concept of three centers in two locations, we often see the following definitions:

Primary data center

In order to ensure reliable data access and high availability, the data center is generally within one kilometer from our office location. All data, database, and access layer applications are hosted in the data center.

Auxiliary Data Center

The business of the secondary data center is equivalent to all copies of the primary data center. To ensure the overall data security, our data center is about 40 kilometers away from the primary data center, preventing small-scale disasters in areas such as fire and water flooding in the primary data center.

The secondary data center switches data after a disaster similar to a small scale occurs in the primary data center. Most enterprises use the secondary data center as a backup site, many enterprises also use their local secondary data centers as their standby sites.

When the standby site is enabled, most users in the enterprise's business data center can use the office facilities of the Auxiliary Data Center for office work.

Third-party Auxiliary Data Center

Generally, the third data center is a local data center that cannot recover services within a short period of time due to large-scale natural disasters. Therefore, the third data center must be enabled to start services to ensure the continuous operation of services. Generally, it takes about four hours to switch between networks, services, and data in a remote data center to resume operation. After the switchover, key business personnel can work remotely by plane to ensure continuous operation of key business data.


Therefore, more than 99% of the SLA we mentioned in most cases refer to local high availability.

However, based on the three centers in two locations model, we have invested a lot, because we must take into account the required hardware and software costs to reach the SLA level. For the two locations, to reach an equivalent application level, we must ensure that at least our hardware investment is more than twice. After the third center is enabled, our cost is also the corresponding hardware cost at least three or dozens of times.

At present, the high availability provided by most clouds does not mean that cross-region transfer can be achieved by raising the high availability SLA level. In most cases, it is also the high availability of local applications, remote High Availability means that it is not implemented using a single instance architecture.

At present, Microsoft cloud has two different data centers in China East and North China, providing 99.95% high availability, of course, we must also design a high-availability architecture to ensure that our real data O & M is normal. There is a problem here. When there is a problem with the data and application, how can we ensure that the data is normal? This relies on a key feature provided by Azure:

TrafficManager

Traffic Manager is translated into Traffic control, which many people think is to control the Traffic size. In fact, a very important feature of Traffic Manger is to control our Traffic trend.

Currently, Traffic supports three Traffic direction modes:

Failover

His work mode is to use aliases to implement failover for two or more cloud services, the system's built-in automatic detection method can detect the status of the corresponding website and directory. Once a problem occurs, it uses the automatic detection mechanism to direct traffic to a normal website.

Performance-based

To balance the load of endpoints distributed across different data centers around the world, you can direct incoming traffic to the nearest endpoint, because the client sending the request has the lowest latency between the end point and the end point. Generally, the "closest" endpoint corresponds to the endpoint with the shortest geographic distance. The "performance" load balancing method can be used for distribution based on location and latency, but network configurations or real-time changes in load cannot be considered.

The traffic manager periodically generates the "Internet latency table ". The traffic manager infrastructure runs tests to determine the round-trip time between global differences and Azure data centers hosting endpoints.

DNSRR

In this way, user access is allocated based on the weight of the accessed application, or random access is implemented under the same weight, that is, the access load based on DNS Round Robin.

Our next goal is to implement failover of application access. Next, we will achieve load balancing of applications and access based on the simple complexity of applications.

The following example uses a simple static page based on SQL Azure and SQL virtual machines. There is also the multi-site synchronization of Mysql and Mysql virtual machines to complete data multi-site access and replication technology.

In the next article, we will take the simple configuration after subscription as an operation mode and configure the failover model according to different requirements.