We can divide the protection level of the disaster data recovery system into data-level disaster recovery and application-level disaster recovery.
Data-level disaster recovery refers to periodically saving the business data of the production machine on a different machine. If the business data of the production machine is lost due to a disaster. After the production machine is repaired, the data saved on the different machines can be used to restore the business. This is data-level disaster recovery.
Application-level disaster recovery focuses on business integrity. As far as the business is not interrupted to a huge extent it can be quickly restored. Cloud computing makes it easier to recover data.
Data-level disaster recovery is the prerequisite of application-level disaster recovery. That is, application-level disaster recovery also requires a backup of the production machine’s business data on the different machines, and when the production machine has a disaster, the different machines can use the backup data to quickly restore the business. Generally speaking, application-level disaster recovery is more complicated and costly.
According to the different implementation levels of data replication adopted by the disaster recovery system, it can be further divided into replication based on the host layer, replication based on the array layer, and replication based on the network layer.
① Application based on the host layer
Install data replication software on the host operating system, or data replication and disaster recovery tools provided by the application program (such as database-related tools), and use TCP/IP network to connect to the server of the remote disaster recovery site to realize remote data replication.
Data replication is built on the application host, users do not need to replace too many existing system architectures, and do not have to worry about the compatibility of back-end storage systems. And it can meet the different data protection requirements of users, provide a variety of different data protection modes, and achieve the most demanding application-level disaster recovery.
Since the replication is based on its corresponding database or volume management, the application has certain limitations. Each application server must have it installed. The system operation needs to occupy host resources. Which will affect the host’s system performance; the operating system version and storage software version compatibility need to be confirmed one by one.
② Replication based on the storage layer
Based on the built-in value-added functions of the storage system (disk array, NAS), it is connected through a transmission interface. Such as an IP network or fiber channel. And the data is copied to the remote side in a synchronous or asynchronous manner. Currently, various mainstream storage vendors can provide this kind of disaster recovery technology.
Separate data from the operation, which has a relatively small impact on the operating resources of the host system. In addition, because the operating mechanism mostly uses mirroring to copy data and speeds up I/O access with the help of high-speed buffer memory, the data difference time point between the two ends is relatively small, and the storage system itself has a certain fault recovery so that it has certain performance and reliability.
High cost. Since users must configure two identical storage systems on the local side and the disaster recovery side, not only the procurement cost is high, but also subject to a single equipment manufacturer, the future scalability is bound to be inflexible. In addition, if a Fibre Channel storage system is to construct remote disaster recovery, one FC-to-IP adapter must be installed at the local end and one at the disaster recovery end. In addition to the cost of network bandwidth, the overall cost is very large.
③Mirror or copy based on virtualized gateway
In the storage area network (SAN) between the front-end application server and the back-end storage system, a storage gateway is added, the front-end is connected to the server host, and the back-end is connected to the storage device. Its role is like the traffic police in the storage network. It manages and controls all I/O. It can not only realize the application and mirroring of the local storage system but also realize the replication of data in different places.
Here, data mirroring or copying is performed through the storage gateway. So, the impact on the performance of the host is very low. Through the virtualization technology of the storage gateway, it is possible to integrate front-end heterogeneous platform servers and back-end storage devices of different brands.
High cost. Since users must add virtualized gateway devices, the cost is higher than storage replication.
In summary, these three disaster recovery replication methods have their own merits. Users need to choose the disaster recovery replication method that suits them according to the characteristics of their own IT systems.