What is the backup for?

Backup: In order to cope with possible accidents such as loss or damage of files and data, the data in computer storage devices are copied to large-capacity storage devices such as tapes. So as to separate a copy of a separately stored program or file from the original text. (Excerpted from Computer Science)

If the system's hardware or storage medium fails, backup tools can help you protect your data from unexpected losses. For example, you can use backup to create a copy of data on your hard disk and then store the data on other storage devices. The backup storage medium can be a logical drive (such as a hard disk), an independent storage device (such as a removable disk) or an entire disk library or tape library organized and controlled by an autochanger. If the original data on the hard disk is accidentally deleted or overwritten, or the data cannot be accessed due to a hard disk failure, you can easily recover the data from the archived copy.

Backup classification

Backup can be divided into system backup and data backup.

1. System backup: refers to the loss of the system files of the user's operating system due to disk damage or damage, computer virus or human error deletion, and the computer operating system cannot be started normally. Therefore, system backup is used to store the operating system in advance for backup support after failure.

2. Data backup: refers to data packets stored by users, including files, databases and applications, for data recovery.

Backup function

The function of backup is for backup support and use.

Backup is the basis of disaster tolerance, which refers to the process of copying all or part of the data set from the hard disk or array of the application host to other storage media in order to prevent data loss caused by system operation errors or system failures. Traditional data backup mainly uses built-in or external tape drives for cold backup. However, this method can only prevent human errors such as operational errors, and its recovery time is also very long. With the continuous development of technology and the massive increase of data, many enterprises begin to adopt network backup. Network backup is generally realized by professional data storage management software combined with corresponding hardware and storage devices.

Backup mode

Common backup methods are:

Back up data with tape regularly.

Remote tape library and CD library backup. That is, send the data to a remote backup center to make a complete backup tape or CD. ? [ 1]?

Remote critical data+tape backup. The production machine uses tape to back up data and sends key data to the backup machine in real time.

Remote database backup. Create a copy of the primary database on a backup computer that is independent of the production computer where the primary database resides.

Network data mirroring. In this way, the database data of the production system and the update of important target files to be tracked are monitored and tracked, and the update log is transmitted to the backup system in real time through the network, and the backup system updates the disk according to the log.

Remote mirror disk. Mirror disks are extended to places far away from production machines through high-speed Fibre Channel lines and disk control technology. The data of the mirror disk is exactly the same as that of the main disk, and the update mode is synchronous or asynchronous.

Data backup must consider the problem of data recovery, including the adoption of? [2] A variety of disaster prevention measures, such as dual-machine hot standby, disk mirroring or fault tolerance, remote storage of backup tapes, redundancy of key components, etc. These measures can restore the system after a system failure. However, these measures can only deal with a single point of computer failure, and they are helpless to regional and devastating disasters and do not have the ability to recover from disasters.

Backup technology

System disaster recovery

(IDR = Intelligent Disaster Recovery)?

System disaster tolerance means that when the system crashes, it only takes a few steps to rebuild the system, including the above system patches, application software and data. This can improve the accuracy of recovery, shorten the recovery time and shorten the business interruption time.

Mainly? [3] Disaster recovery technology, there are several recovery methods:

Use a universal CD, which is used by all systems. This CD-ROM can start the system and network at the same time, and then the backup server will restore the contents of the whole hard disk or the first main partition backed up in the tape library (or virtual tape library) to the disaster machine. In this way, no matter how much data the disaster recovery machine has, it can be stored in the backup equipment of the whole backup system for unified management, and it is not necessary to engrave the CD frequently, which means that the CD does not need to be updated frequently. The typical representative is bakbone netvault disaster recovery, and the data flow of Bakbone Netvault disaster backup and recovery is shown on the right. All systems use a unified CD.

It is necessary to carve a separate CD for each computer, and use a targeted CD for recovery. Each machine needs its own CD, and it needs to be updated regularly. The biggest disadvantage of this method is that it often needs to be engraved. Otherwise, if there is no CD or the CD is too long, it will affect the speed of recovery and the state after recovery. The typical representative of this technology is the EMC Network Recovery Manager module.

When restoring, you need to start with the help of the network, that is, you need to have a host with the same operating system as the boot machine, and then use the backup content to restore. The typical representative of this technology is the bare metal recovery module of Symantec NetBackup.

Disaster recovery tools provided by the operating system itself. For most Unix minicomputers, system backup tools are provided. With the help of the tape drive that comes with the system and a simple command, HP-UX can use make_recovery to back up the entire root volume to the 4mm tape that comes with the server. When restoring, this tape can start the system by itself, and the entire root volume can be restored to the hard disk with one command. The advantage of this method is simple and economical, especially for Unix systems, which is far more convenient, safe and economical than the BMR module provided by backup software, because the BMR module provided by backup software often needs other servers on the same platform to start.

Remote data replication

Make a usable copy of the data in another place in real time. This copy can be used immediately without data recovery. The biggest advantage of data replication is that the copied data can be used immediately, and there is no data recovery time, so RTO is very good. At the same time, because it is real-time replication, RPO is also very good, and there will be almost no data loss. The disadvantage is that the cost is much higher than that of data backup. Not only the data replication system is expensive, but also other hard disk storage space and host system are needed, and even a remote computer room is set up. Considering network wiring, it will bring great cost increase, so the construction and maintenance cost of data replication is far greater than that of data backup.

At present, data replication can be achieved in the following ways:

Host based. Host-based data replication technology can ignore the isomorphism of storage systems, as long as the hosts are the same operating system. There are softwares that support data replication between heterogeneous hosts, such as BakBone NetVault Replicator, which can support data replication between heterogeneous servers and remote real-time replication across WAN. The disadvantage is that it takes up a little host resources.

Based on storage system. Using data replication software provided by storage systems, replicated data streams can be passed between storage systems, regardless of the host. The advantage of this method is that data replication does not occupy host resources, but the disadvantage is that the storage system of disaster recovery center and the storage system of production center have strict compatibility requirements, and generally need the storage system of the same manufacturer, which limits the choice of storage system of disaster recovery center for users.

Based on optical fiber switch. This technology is developing, using the new function of optical fiber switch, or using management software to control the optical fiber switch to virtualize the storage system, and then the management software performs volume management, volume replication, volume mirroring and other technologies on the managed virtual storage pool to realize remote data replication. Typical ones are Storag, Falcon, etc.

Data replication based on application. This technology has some limitations, and it is aimed at specific applications. Mainly use the replication module provided by the database itself, such as OracleDataGuard, Sybase replication and so on.

Data protection technology

Continuous data protection

CDP technology is the most popular data protection technology at present. It can capture all data write changes at file level or block level, and can restore backup objects in more detailed granularity, which can be restored to any point in time.

CDP technology is a new technology, which is gradually integrated into many traditional backup software. For example, BakBone NetVault Backup 8.0 adds TrueCDP module, Symantec Backup Exec 12.5 and so on. Other companies, including EMC and Symantec, have acquired some CDP software and are integrating with traditional backup software. They are all integrating.

CDP technology includes two types:

Near CDP is what we call quasi CDP. Its biggest feature is that it can only recover some data at a specified point in time (FPIT), which is somewhat similar to a logical snapshot of a storage system and cannot recover a fixed point in time. CDP backed up by Symantec, CommVault and Kay all belong to this type.

TrueCDP, which we call TrueCDP, can recover any point in time (APIT) within a specified period of time. BakBone TrueCDP belongs to true CDP type.

Application status

If the disaster recovery plan is distributed in various industries, everyone will use it according to a certain gradient.

Firstly, system disaster recovery (BMR bare metal recovery) is adopted, because this scheme has the lowest cost, and it can be completed by adding some modules when building a data backup system. No extra storage space, no extra storage space? [4] Disaster tolerance room, so it can be realized by almost all qualified users. Only some users use the backup tools provided by the operating system to assist, and some users use the disaster module provided by the backup software to complete. This kind of users have the largest number and are distributed in various industries. BakBone's VaultDR is widely used by users who use BakBone NetVault's backup software. Many users like BakBone VaultDR very much. Regardless of the type of operating system, a standard CD can be applied to all Intel x86 servers for disaster backup and recovery.

Secondly, users who build disaster recovery systems and high-end enterprises with large amounts of data generally adopt storage-based data replication technology, such as telecommunications and financial industries. Low-end users generally use host-based data replication software, which has low cost and does not need strict storage system procurement. In particular, the NetVault Replicator of BakBone can not only replicate data remotely, but also support heterogeneous platforms, so it has a wide range of users at home and abroad, some government departments, power companies, securities departments, website companies and so on. This method is adopted, and it is mostly used on Windows and Linux platforms. Although the virtual storage technology based on storage exchange layer also has some users, due to the limited start time of the technology, the maturity of the technology needs further verification, and many users just take it as a pilot and have not popularized it on a large scale. For application-based data replication, some high-end customers are adopting it, and there are a large number of users.

As for CDP technology, it has just started. This technology meets the needs of many users who are concerned about disaster recovery, and can obtain high standards in RTO and RPO. At the same time, it can reproduce any version of history, which is attracting more and more users' attention. I believe there will be a very wide range of users in the future. BakBone's TrueCDP can be restored at any time, and it is a real CDP, which will have greater advantages than other quasi-CDP products and will be selected by more users who care about disaster recovery.

Application misunderstanding

Disaster tolerance system has been paid more and more attention because it can bring business continuity, but some misunderstandings should be remembered in the process of use.

There is no choice that suits you. Without careful analysis and definition of business continuity requirements, RTO and RPO have not carefully studied or done it, but want to pursue the most comprehensive, and may ignore the most needed construction and fail to meet their own needs.

It is considered that disaster recovery system is omnipotent, which exaggerates the function of disaster recovery system and ignores the construction of backup system and high availability system. In fact, each system has its own functions and needs to be built according to actual needs.

There is no one-step construction of disaster tolerance, and we always want to wait until the actual maturity before building it in an all-round way, or even start building a disaster tolerance system without building a data backup system first. Without the construction of backup system, the bottom line of data security will not be reached, and a complete data will be lacking at the critical moment, and spending a lot of financial resources on the construction of disaster recovery system will not be worth the candle.

The selection of various disaster recovery technologies is unreasonable. For example, the host-based replication software can run the resources of the host system, but the design based on the storage system is selected, which requires the selection of the same type of storage devices, which increases the overall cost.

For disaster backup of BMR system, we always want to choose the BMR module that comes with the backup software. In fact, the BMR module of backup software is convenient for Window and Linux, but it is not suitable for all Unix systems. In terms of Unix system backup, Unix system backup tools and tape drives are very convenient. One command can complete system backup without relying on other servers, which is far superior to BMR modules of some backup software, which is not only convenient to operate, but also reduces the cost.

Future development direction

The future technological development should be the coexistence of various technologies, and the way to meet customers' needs can attract customers' attention.

One of the future development directions should be that heterogeneous replication technology based on host will have a broader market. Because most customers have heterogeneous host environments and support data replication technology in heterogeneous environments, they can use the existing environment to replicate each other. For replication software that does not support heterogeneous environment, it is necessary to purchase the same storage or the same operating system host for data replication, which increases the overall cost of disaster recovery.

The second development in the future is CDP technology. CDP combines the advantages of data backup and data replication, which can not only protect real-time data, but also restore historical data at any time, and will have stronger vitality. With the seamless connection between TrueCDP and traditional backup software, more and more users will use TrueCDP to build disaster recovery systems. The picture on the right shows the popular architecture of BakBone NetVault TureCDP backup system.

backup system

The role of backup system

Many system administrators think that it is a waste to invest in managing a backup task. Compared with the original complex backup system, it only works when it is restored, and it usually adds a lot of management tasks, which is a great waste for enterprises.

This problem is a big one, which is directly related to the return on investment (ROI) of enterprise information systems. When the importance of IT system is very high, and the life-related data of enterprises are all in the computer system, then the protection of data is very important, which is one of them. In addition, the implementation of the backup system will not bring more management tasks, on the contrary, it will bring a lot of maintenance convenience to the system administrator, mainly as follows:

The automation of backup reduces the risk brought by the operation of maintenance personnel;

Online database backup to ensure 24×7-hour business operation;

Point-in-time recovery of file system and database data, version history management;

Tape cloning can reduce the probability of tape error and realize disaster tolerance and remote storage;

Network backup, LAN-free and server-free backup methods;

Recover quickly from system disaster.

Therefore, if the enterprise must realize the above functions, but the enterprise does not implement the backup system, and if the investment in manpower and equipment is greater than the investment in implementing the backup system, then the investment in the backup system is acceptable.

Functions of backup system

Users only need simple backup, which is not so complicated. Users have bought many useless advanced technologies in the backup system, so it is unnecessary to realize too many complicated functions in the system.

The function of backup system is the product of the development of backup technology to a certain stage, and the emergence of advanced technology is also due to the needs of user systems. When the IT system develops to a certain extent, users' dependence on the IT system increases, and the data volume of the IT system becomes larger and larger, so the requirements for system backup become higher and higher. But a good backup system should have the following characteristics:

The backup system can be easily expanded according to the needs of the application system;

Meet the backup system pressure brought by future data volume and application system upgrade;

In the aspect of backup system, especially the upgrade ability of backup software.

Generally speaking, backup technology has gone through several stages of development, from traditional tape backup to network backup, from SCSI? No LAN backup to dynamic sharing of SAN structure without LAN backup, until the emergence of no server backup. It is foreseeable that the future backup products will have the following trends:

Backup CD

situation

With the further decline of SATA disk price, the backup advantage of disk is gradually reflected. EMC has introduced a product that uses disk as virtual tape library, which can replace the function of tape library in function. The author believes that the batch listing of this product will trigger the upgrade of backup hardware equipment to a certain extent.

ISCSI technology

This technology can use the existing TCP/IP network for data transmission, and users can conveniently realize remote protection of data. Many such products have appeared in the market. I believe that with the maturity of the products, many remote backup and disaster recovery schemes will choose this technology.

NDMP protocol

As a standard, NDMP (Network Data Management Protocol) has been developed to the fourth edition, and products supporting this protocol can easily realize fast backup and recovery of NAS server data. For a large number of small files, this technology has unique advantages.

Backup technology

Many disk arrays provide snapshot function. For snapshots on disk arrays, we can make full use of snapshot technology to backup data without server, which is of great significance to large enterprise systems, greatly reducing the resource occupation of production systems during backup, and can be recovered very quickly.

Realize the life cycle management of backup media

Usually, each tape has a certain number of uses. Therefore, for the tape backup system, after a certain number of uses, the tape should be discarded and cannot be used to back up key data.

maintain

People invest in fully automatic backup system to reduce maintenance workload. As long as the system is implemented, the maintenance work can be relaxed. Just check the backup status once a week or once a month.

After the user realizes the backup system, the maintenance of the backup system is still very important. Mainly determined by the following factors.

Because many IT systems in many units have many hosts, many application systems, and each application system has corresponding management and maintenance personnel, backup is the centralized management of various application data backup tasks. Therefore, for users with complex application systems, backup system administrators or storage backup engineers can be set up to maintain the overall backup system.

With the increase of data and application systems, the backup strategy should be optimized with the migration of time.

Backup system involves many technologies, such as operating system, database, storage, tape library and so on. Therefore, it is suggested that the backup system administrator should have a certain understanding of all kinds of knowledge, besides the backup system knowledge, he should also participate in professional training of operating system, database and other products to ensure that the application system can recover data as soon as possible in the event of a disaster.

If the daily backup of the key database fails, it may cause the database to be suspended. For example, for? [5] For [5]Oracle database, if the "archive log" of the database is not backed up in time, the whole database will stop.

If there is a "cloning" function in the tape library backup system, it is necessary to take out the "cloned" media every day and store them in different places to facilitate disaster recovery.

product

When choosing products, the backup tape library should be as large as possible, the tape drive should be as fast as possible, and the most advanced technology should be used as much as possible.

In the process of product selection, cost performance is one of the most important indicators, but it is by no means the whole. Users are advised to consider the following factors when selecting products:

Type, quantity, backup strategy (full backup, incremental backup, etc. ) and the retention time of critical data determine the total capacity of the tape library.

The minimum number of tape drives is determined by the backup time window and the peak data volume of backup data volume;

The type of application system and data type determine the technology used for backup.

If you use NAS devices for backup, try to choose NDMP backup; SAN architecture backup should be dynamically shared by tape drives as much as possible; Disk array provides mirroring or snapshot function, and snapshot backup technology can be used.

In the choice of tape drive technology, it is suggested to choose more general equipment. For this tape drive, the operating system and backup software are compatible.

Because the tape drive is a mechanical device, the failure rate is higher than that of disks, optical disks and other devices, so try to keep redundancy when configuring the tape drive. SATA disk technology has developed by leaps and bounds, and disk backup technology has become the development direction.

For hardware equipment, considering the provision of spare parts and the service level of suppliers, and for backup software, providing professional services and technical support is also an important factor to be considered.

Backup software should be widely compatible.

Data disaster tolerance

The loss of enterprise's key data will interrupt the normal business operation of the enterprise and cause huge economic losses. In order to protect data, enterprises need to back up disaster recovery systems. However, many enterprises think that they can sit back and relax after establishing a backup system. In fact, they need to set up a disaster recovery system. The connection between data disaster tolerance and data backup is mainly reflected in the following aspects:

Fundamentals of data backup

Data backup is the last line of defense for high availability of data, and its purpose is to recover data quickly when system data crashes. Although it is also a disaster recovery scheme, its disaster recovery ability is very limited, because the traditional backup is mainly carried out by tape drives with built-in or external data? [6] Cold backup and backup tapes are also managed in the computer room. Once there is a disaster in the whole computer room, such as fire, theft and earthquake, these backup tapes will be destroyed, and the stored tape backups will have no disaster tolerance function.

Disaster tolerance and backup

The real data disaster tolerance is to avoid the inherent shortcomings of traditional cold backup and restore the whole system in a comprehensive and timely manner when a disaster occurs. Disaster tolerance can be divided into several levels according to its disaster tolerance. For example, the disaster recovery system defined by the international standard SHARE 78 has seven levels: from the simplest local tape backup, to storing the backed-up tapes in different places, and then to establishing a remote backup system with real-time switching of application systems, the recovery time can also range from days to hours, to minutes, seconds or zero data loss.

No matter which disaster recovery scheme is adopted, data backup is still the most basic. Without backup data, any disaster recovery scheme is meaningless. But backup alone is not enough, and disaster tolerance is also essential. Disaster tolerance for IT is to provide a computer information system that can prevent all kinds of disasters. Technically, there are two main indicators to measure the disaster recovery system: RPO (recovery point object) and RTO (recovery time object), where RPO represents the amount of data allowed to be lost when a disaster occurs; RTO indicates the time of system recovery.

Disaster tolerance is not just technology.

Disaster tolerance is a project, not just technology. Many customers are still concerned about disaster recovery technology, but they are not clear about the process, specifications and specific measures of disaster recovery. I have never evaluated the feasibility of the disaster recovery plan, thinking that as long as the disaster recovery plan is established, I can rest easy. In fact, this is very risky. Some small and medium-sized enterprises, in particular, think that their enterprises have spent a lot of human and financial resources on data backup and disaster recovery all year round, and as a result, no major disasters have occurred for several years, so they have relaxed their vigilance. But once the disaster happens, it's too late to regret it! In this regard, foreign multinational companies have done very well. Although there have been no major disasters in recent years, it is almost useless to back up so many tapes, but as always, they have done every step very seriously, and basically evaluate the feasibility of the current disaster recovery plan every month and conduct field exercises. ? [7]?

Data disaster tolerance level

Design one? [4] Disaster recovery backup system needs to consider many factors, such as the size of backup/recovery data, the distance between application data center and backup data center and data transmission mode, the recovery speed required in case of disaster, the management and investment of backup center, etc. According to these factors and different applications, the common disaster recovery backup levels are as follows:

Local cold backup

This level of disaster recovery backup is actually the data backup mentioned above. Its disaster tolerance is the weakest, and it only backs up data locally, and the backed-up data tapes are only kept locally and not sent to other places.

What is the most commonly used equipment in this disaster recovery scheme? [8] Of course, the tape drive can be loaded manually or automatically according to actual needs. The former is mainly suitable for small and medium-sized enterprises with small storage data capacity.

reference data

1. Research on optimal expansion scheme of tape library based on NBU backup system. China hownet [citation date: 20 17-04-02]

2. Technical research and concrete implementation of dual-machine hot standby system. China hownet [reference date 20 17-04-02]

3. The application research of cloud computing in power system data disaster recovery service. China hownet [reference date 20 17-04-02]

4. Research and efficiency analysis of synchronization strategy in disaster recovery backup system. China hownet [citation date: 20 17-04-02]

5.Oracle database optimization. China hownet [reference date 20 17-04-02]

6. Realization of cold backup and recovery method of precursor management system database. China hownet [reference date 20 17-04-02]

7. Disaster tolerance theory and key technology analysis. China hownet [reference date 20 17-04-02]

8. Research and implementation of tape drive control program. China hownet [reference date 20 17-04-02]

This article is quoted from Professor Yao Yuan, Peking University, and comes from Baidu Encyclopedia.

Additional links: Web links