Analysis of the Four Major Disaster Recovery Technologies in Data Centers

Disaster recovery technology means that when a data center fails or a disaster occurs, other data centers can operate normally and take over key or all businesses to achieve the effect of mutual backup. Good disaster recovery technology can achieve "no perception of failure" for users. Disaster recovery is a comprehensive system engineering, involving a variety of different technologies such as backup, replication, and mirroring, and the system construction is highly complex. Therefore, it is generally only used in large enterprises and the financial industry. In 2007, my country promulgated the national standard "Information System Disaster Recovery Specification" (GB/T 20988-2007) on disaster recovery, which is an important reference document for us in disaster recovery construction. The business carried by data centers today is becoming more and more important. The introduction of effective disaster recovery technology can reduce the losses caused by data center failures. The overall disaster recovery technology of data centers can be divided into four types: cold backup, warm backup, hot backup and active-active.

[[137173]]

Cold Standby

Cold standby technology is a disaster recovery technology often used in small and medium-sized data centers or sites that carry unimportant services. The sites used for cold standby technology are usually empty sites, generally used in emergency situations; or just equipment that has been wired and powered on. When the entire data center fails and cannot provide services, the data center will temporarily find idle equipment or rent an external company's data center for temporary recovery. When its own data center recovers, the business will be cut back. In this way, the time for data center business recovery is difficult to guarantee, and sometimes the temporarily built platform may be interrupted again due to instability. Of course, this method does not require a large number of idle equipment to be prepared, and the maintenance cost can be ignored. It takes a high cost and time for cold standby technology to be enabled and actually start working, usually taking several days or even a week or longer. Cold standby technology is not a real disaster recovery technology, because cold standby basically means that the data center has never considered the possibility of a failure in the data center. Once a failure occurs, it is a "girl gets married before the ears are pierced" approach, with no foresight and advance investment in the failure. Of course, the shortcomings of cold standby technology are obvious. It has become increasingly unable to adapt to the high-demand development of data centers and has gradually become an obsolete technology.

Warm standby

Warm standby technology is implemented on the basis of the primary and standby data centers, and the premise is to have two data centers, one primary and one standby. The standby data center is deployed as a warm standby, and the application business is responded to by the primary data center. When the primary data center fails and the business is unavailable, it is necessary to achieve the overall switching of the data center within the specified RTO (Recover Time Objective, that is, the time required for the information system to resume normal operation after a disaster occurs). In the specific implementation, the network configurations of the two business systems in the primary and standby data centers are exactly the same, and the routes of the standby data center are not usually released to the public. When the primary and standby data centers are switched, it is necessary to disconnect the routing link of the primary data center and connect the routing link of the standby data center to ensure that only one data center is online at the same time. Warm standby technology is still a manual method. From knowing the failure of the primary data center to the work of the standby data center, someone needs to be on duty 24 hours a day to complete, and the work effect is low.

Hot Standby

Compared with warm standby, the most important feature of hot standby is the realization of overall automatic switching. The other aspects are basically the same as warm standby. The data center that realizes hot standby only needs to deploy one more software than the data center that realizes warm standby. The software can automatically sense data center failures and ensure automatic switching of application services. The service is responded by the primary data center. When a data center failure causes the service to be unavailable, it is necessary to automatically switch the service to the backup data center within the specified RTO time. In specific implementation, GTM (Wide Area Traffic Manager) is deployed in both the primary and backup data centers. Information is synchronized between GTMs. GTMs detect the application services of their own centers and determine the availability of application services based on the server status of GTMs. When the GTM or data center link is DOWN, the service is automatically switched to the disaster recovery data center. The hot standby data center uses GTM technology to achieve automatic primary and backup data center switching.

Active-Active

Active-active technology can realize that both the primary and standby data centers provide external services. During normal operation, the services of the two data centers can be load-shared according to weights, without distinction between the primary and standby data centers. Each data center responds to a portion of users. The weight can be divided by region, data center service capacity or external bandwidth. When one of the data centers fails, the other data center will take on all services. In specific implementation, multi-active technology deploys many ways to detect faults, such as: ICMP Monitor, TCP Monitor, HTTP Monitor, FTP Monitor. It can also detect the running status of the server and the load balancing of the server in real time. Even when there is no fault, it can be adjusted between the multi-active data centers according to the application business volume. The most important feature of multi-active is that it will not cause waste of data center resources. All data centers carry application business operations. It will not cause the situation that the hot standby and warm standby primary data centers are almost fully loaded, while the standby data center is very idle. In addition to active-active technology, there is also multi-active technology. Multi-active means that the business runs on multiple data centers at the same time. When one or more data centers fail, other data centers will automatically take over all application businesses. Obviously, multi-active is more reliable than active-active, but the investment cost will also be high, and the implementation technology is more complicated. Data centers in the financial and Internet industries are now happy to adopt multi-active technology. Although the investment is large, stability is the priority for these data centers.

Table 1 below lists the characteristics of four disaster recovery technologies.

Table 1: Comparison of the characteristics of four disaster recovery technologies

No matter which disaster recovery technology is used, it should be considered comprehensively based on the importance of the data center's application business, construction funds, and personnel skill levels. It is not necessary to use dual-active/multi-active disaster recovery technology. Although this technology has the highest reliability, it is complex to implement. Expansion and business changes require a lot of professional technical knowledge. Therefore, in addition to the high construction cost, the maintenance cost of the later investment is also high, which will bring a heavy operational burden to the data center. You should study these four disaster recovery technologies in depth and make a comprehensive choice based on your actual situation. Disaster recovery technology is becoming more and more important in data centers and has gradually become one of the important technologies that must be used in data centers.