JP4127461B2 - Backup system and method in disk shared file system - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a system and method for backing up data stored in a recording medium such as a disk of a computer (computer) system and restoring it when necessary.
[0002]
[Prior art]
In a conventional computer system, when a file system performs a backup, first, block information such as an address of a used block is checked for each file. Then, by reading the data of the corresponding block from the disk, the file is read and the read data is copied to the tape. By repeating such an operation for each file, the file is backed up.
[0003]
However, with this method, when a large number of files are backed up, access to the disk is close to random access, which is a cause of deteriorating system performance.
[0004]
Therefore, in order to improve the efficiency of backup, image backup that directly copies a plurality of blocks occupied by a file has been used. In this method, the computer system copies the block area occupied by the file in the disk at once instead of selectively copying the file. For this reason, the backup is performed by one disk access, and the processing becomes efficient.
[0005]
[Problems to be solved by the invention]
However, the conventional backup method described above has the following problems.
[0006]
In the conventional image backup, data can be copied in units of disks, but files and directories cannot be copied in units. For this reason, there was a problem that even unnecessary data had to be copied. In addition, in order to restore the backed up data, it was necessary to copy and expand all the data on the disk.
[0007]
Further, when a backup is performed in a cluster system in which a plurality of computers share a disk and perform processing, the following problem occurs.
The cluster system includes a file system (disk shared file system) for a plurality of computers to simultaneously access a shared disk, and each computer includes an area for caching write data. For this reason, simply copying the shared disk does not reflect the contents of the cached write data (write cache) in the copy, so it is impossible to perform normal image backup.
[0008]
In addition, in a conventional computer system, in order to perform image backup while the business is running (online), the file system saves the original data before the change to another area when the data is changed, and the disk is saved. After copying, the saved original data was overwritten on the copy. As a result, the contents at the start of the backup can be determined on the backup data.
[0009]
However, in a cluster system, changes by multiple computers may occur almost simultaneously in the same file area, and the method of determining the contents at the start of backup using the original data cannot be applied directly. There is.
[0010]
Thus, the conventional backup method cannot efficiently back up a large amount of data in the cluster system. For this reason, an effective backup method in the cluster system has not been developed, and there is no method for efficiently browsing the backed up data.
[0011]
An object of the present invention is to provide a system and method for efficiently backing up data in a computer system having a disk shared file system.
[0012]
[Means for Solving the Problems]
FIG. 1 is a principle diagram of a backup system according to the present invention.
In the first aspect of the present invention, the backup system includes a copy unit 1 and a control unit 2 and backs up a shared medium 4 shared by a plurality of computers 3.
[0013]
The copy unit 1 copies a plurality of unit areas of the shared medium 4 to the backup medium 5 at once. The control means 2 manages the write data that each computer 3 writes to the shared medium 4 and reflects the write data of each computer 3 to the shared medium 4 at the time of backup.
[0014]
Each computer 3 holds the write data as a write cache when the data in the shared medium 4 is changed, and writes the contents in the shared medium 4 when the shared medium 4 becomes accessible. The control means 2 manages the presence or absence of write data held by each computer 3 and performs control to write the write data held by the computer 3 into the shared medium 4 at the time of backup.
[0015]
The storage area of the shared medium 4 is divided into unit areas such as blocks, for example. After all the write data is written, the copy unit 1 copies a plurality of unit areas of the shared medium 4 to the backup medium 5 by a method such as an image backup, for example.
[0016]
According to such a backup system, it is possible to efficiently back up the shared medium 4 including the write data held by each computer 3 in the disk shared file system.
[0017]
In the second aspect of the present invention, the backup system includes a log management unit 6 and a generation unit 7, and includes a shared medium 4 shared by a plurality of computers 3. image Make a backup.
[0018]
The log management means 6 is one of a plurality of computers 3 Calculator 3 Shared media 4 When writing to a certain area , Image of the area before writing Log data as On the computer 3 that wrote Manage, Managed by multiple computers 3 Collect the logs and generate the whole log. When there are a plurality of logs for the same area, the oldest log is used after the image backup start time. The generation means 7 When restoring, Data at the start of image backup is generated using the entire log.
[0019]
When the computer 3 changes the data of the shared medium 4, the original data before the change is saved as a log. The log management means 6 manages the logs of each computer 3 and collects the logs of two or more computers 3 to generate the entire system log. And the production | generation means 7 determines the content at the time of a backup start by overwriting the whole log on the backup data of the shared medium 4, for example.
[0020]
According to such a backup system, the original data stored in response to the data change by the plurality of computers 3 is edited, and the entire log is generated. Therefore, even in the disk shared file system, backup can be performed efficiently while the system is operating.
[0021]
In the third aspect of the present invention, the backup system includes a copy unit 1 and a group management unit 8 and backs up a shared medium 4 shared by a plurality of computers 3.
[0022]
The group management means 8 sets a group of files stored in the shared medium 4 and lists unit areas occupied by files included in the group. The copy unit 1 copies a plurality of listed unit areas to the backup medium 5 at once.
[0023]
The group management means 8 sets a group including one or more files, and lists unit areas occupied by each file included in the group. Then, the copy unit 1 copies a plurality of listed unit areas to the backup medium 5 by a method such as image backup without distinguishing each file.
[0024]
According to such a backup system, it is possible to specify a file to be backed up in the disk shared file system, and there is no need to copy an unnecessary file. Therefore, the backup is made efficient.
[0025]
In the fourth aspect of the present invention, the backup system includes a copy unit 1 and an area management unit 9 and backs up a storage medium 4 that stores a file accessed from the computer 3.
[0026]
The area management unit 9 determines whether each unit area of the storage medium 4 is used and lists the used unit areas. The copy unit 1 copies a plurality of listed unit areas to the backup medium 5 at once.
[0027]
The area management unit 9 manages each unit area of the storage medium 4, determines whether each unit area is used as a file, and lists the unit areas occupied by the file. Then, the copy unit 1 copies a plurality of listed unit areas to the backup medium 5 by a method such as image backup without distinguishing each file.
[0028]
According to such a backup system, it is not necessary to copy a unit area that is not used as a file. Therefore, the backup in the file system is made efficient.
[0029]
In the fifth aspect of the present invention, the backup system includes a copy unit 1 and an area management unit 9 and backs up a storage medium 4 that stores a file accessed from the computer 3.
[0030]
The area management unit 9 lists unit areas of the storage medium 4 that have been changed after the previous backup as differences. The copying unit 1 copies a plurality of listed unit areas collectively to the backup medium 5 as differential backup data.
[0031]
The backup system backs up the storage medium 4 in time series at an appropriate timing. The area management unit 9 manages each unit area of the storage medium 4 and lists a unit area changed after the previous backup or a unit area newly used as a file. Then, the copy unit 1 copies a plurality of listed unit areas to the backup medium 5 by a method such as image backup without distinguishing each file. Thereby, only the changed unit area is stored as a difference.
[0032]
According to such a backup system, it is not necessary to copy a unit area whose data has not changed since the previous backup. Therefore, the backup in the file system is made efficient.
[0033]
For example, the shared medium 4 in FIG. 1 corresponds to a shared disk 13 in FIG. 2 described later, and the backup medium 5 in FIG. 1 corresponds to the backup medium 15 or the tape 16 in FIG. 1 corresponds to the copy management unit 25 in FIG. 2, the control unit 2 in FIG. 1 corresponds to the cache control unit 21 in FIG. 2, and the log management unit 6 in FIG. 2 corresponds to the log management unit 26 in FIG. 2, the generation unit 7 and the area management unit 9 in FIG. 1 correspond to the block management unit 22 in FIG. 2, and the group management unit 8 in FIG. This corresponds to unit 23.
[0034]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
The computer system of this embodiment includes a plurality of computers, a shared disk shared by these computers, and a file system for allowing the plurality of computers to simultaneously access the shared disk.
[0035]
This computer system directly copies the entire contents of a disk to a backup medium by means of an improved image backup at the time of data backup. Further, the computer access to the disk is detected, and the original data before the access occurs is stored in the log medium. Then, using the data stored in the log medium, an image (data content) at the time of starting backup is determined. Hereinafter, this original data will be referred to as a Before Image Log (BI log) or simply a log.
[0036]
In the image backup of the present embodiment, the main features regarding the backup operation are as follows.
(A.1) The contents of the write cache on the memory of each computer are managed, and the contents of the write cache of each computer are reflected on the disk at the time of backup. This prevents data inconsistencies in the cluster from occurring.
(A.2) Each computer that attempts to write to the disk leaves a BI log, and merges the BI logs of multiple computers at the time of backup. As a result, the log of the entire system in which all the logs are collected after the backup is completed is determined, and the data is determined using this log, so that the destruction of the data due to the write during the backup (copying) is prevented.
(A.3) Each computer that intends to write to the disk notifies the specific computer responsible for the BI log of the write, and the computer manages the BI log. As a result, logs of a plurality of computers are sent to a specific computer, merged, and stored in a log medium. By confirming data using this log, destruction of data due to writing during backup is prevented.
(A.4) The same medium as the backup data is selected as the medium for storing the BI log. Thereby, a log can be preserve | saved simultaneously with backup.
(A.5) A medium different from backup data is selected as a medium for storing the BI log. As a result, even when the medium for storing backup data cannot be overwritten, a log can be left.
(A.6) The BI log is overwritten on the backup data and then stored in the backup medium. This eliminates the need to refer to a plurality of media during restoration.
(A.7) The address information of the backup data to be overwritten is written in the BI log. As a result, it is possible to overwrite the backup data by simply reading the log without accessing the log management information.
(A.8) List used blocks among the blocks on the disk, and copy only necessary parts. Thereby, since the amount of data to be copied is reduced, it is possible to shorten the copy time and the required medium capacity.
(A.9) Of the used blocks on the disk, those that have changed since the previous backup (difference) are listed, and only the changed part is copied. Since such differential backup reduces the amount of data to be copied, it is possible to shorten the copy time and the required medium capacity.
(A.10) After backup is completed and before restoration, the differential backup data or the contents of the differential backup data and the entire backup data are merged in units of blocks. Restoration is made more efficient by collecting differential backup data.
(A.11) The recording start time of differential backup data can be selected. By using only differential backup data after the selected point in time of restoration, changes made before that point can be ignored, and flexible restoration becomes possible.
(A.12) The disk copy process is distributed to a plurality of computers in the cluster. This distributes the load and shortens the copy time.
[0037]
In the image backup according to the present embodiment, main characteristics regarding file grouping are as follows.
(B.1) Files are grouped, the blocks occupied by the files included in the group are managed, and only the blocks used by those files are copied during backup. This makes it possible to set file groups and back up files in groups.
(B.2) Files are grouped in units of directories, and all files included in the directory are set as a group. This makes it possible to set file groups and back up files in groups.
(B.3) A specific file or directory under a directory set as a group is excluded from the group. Thereby, a specific file included in a directory set as a certain group can be excluded from the group, and a flexible group setting becomes possible.
(B.4) Set up a plurality of groups and perform backups according to different schedules. This enables flexible group setting and backup.
(B.5) Recognize that one file belongs to a plurality of groups. This enables flexible group setting and backup.
[0038]
In the image backup according to the present embodiment, main features regarding the restore operation are as follows.
(C.1) The file system mounts the medium for storing the backup data as it is instead of the disk. As a result, a medium for storing backup data can be accessed instead of a disk, and a special operation for restoration is not required.
(C.2) When the above-described differential backup is performed, the backup data of each generation is searched and traced back to the entire backup data as necessary. Thereby, when the block of the file is not included in the latest differential backup data, the block stored by the previous backup can be referred to and used. Therefore, at the time of restoration, it becomes possible for the user to appear as if all the data exists.
(C.3) Load only the necessary blocks from the backup tape into the disk used as a buffer, and use those blocks as a cache. As a result, only necessary blocks can be arranged on the buffer, and the access efficiency to frequently accessed blocks is improved.
(C.4) Load only the necessary blocks from the backup tape onto the disk and show the data to a computer not connected to the tape. This makes it possible for a computer that does not have a tape in the cluster to read backup data stored on the tape.
(C.5) When the BI log is not overwritten on the backup data, the BI log is referred to first, and the backup data is referred to later if necessary. If the log and backup data are stored separately on the medium, the existence and contents of the log are confirmed, and if there is a log, the log is referenced, and if there is no log, the backup data is referenced. As a result, no contradiction occurs in the restored data.
[0039]
FIG. 2 is a configuration diagram of a cluster system that performs image backup as described above. The cluster system in FIG. 2 includes a plurality of computers 11 and 12, a shared disk 13, a log medium 14, a backup medium 15, and a tape 16.
[0040]
The plurality of computers 11 share the shared disk 13 and perform data processing while accessing files stored on the disk 13. The computers 11 and 12 and the shared disk 13 constitute a cluster, and generally one or more shared disks 13 are provided in the cluster. The log medium 14 stores the BI log of the computer 11, and the backup medium 15 and the tape 16 store backup data of files in the disk 13.
[0041]
The computer 12 is a computer that manages a cluster, and includes a cache control unit 21, a block management unit 22, a group management unit 23, a medium control unit 24, a copy management unit 25, a log management unit 26, and a tape control unit 27. These management unit and control unit correspond to, for example, software described by a program, and the block management unit 22 corresponds to a main part of the file system.
[0042]
The cache control unit 21 controls a cache 28 provided on the memory of each computer 11. In the cache 28, data to be written by the computer 11 with respect to the cluster file stored on the disk 13 is temporarily stored.
[0043]
In addition, the block management unit 22 performs block allocation to the file, and manages to which address of which disk 13 each block of the file is allocated. The group management unit 23 manages a user-defined group and files included in the group.
[0044]
The medium control unit 24 controls access to the backup medium 15, and the copy management unit 25 manages the data copy operation from the disk 13 to the backup medium 15. The log management unit 26 manages the BI log and the log medium 14 of each computer 11, and the tape control unit 27 controls access to the tape 16.
[0045]
According to such a cluster system, the block management unit 22 manages the blocks of the backup target file on the disk 13, and the copy management unit 25 copies these blocks to the backup medium 15, thereby performing backup. The BI log based on the writing of each computer 11 in the cluster generated during copying is copied to the log medium by the log management unit 26. The BI log stored in the log medium is later reflected in the backup medium 15 or is retained in the log format as it is.
[0046]
The log medium 14 and the backup medium 15 are non-volatile media such as disks and tapes. When the backup medium 15 is a disk, it is also used as a buffer for the tape 16, and the tape control unit 27 copies backup data from the backup medium 15 to the tape 16.
[0047]
First, the operations relating to the above-described features (a.1) to (a.7) will be described in detail with reference to FIGS.
The cache control unit 21 performs cache management of (a.1), and reflects the write data (write cache) in the cache 28 on the disk 13 at the time of backup. The cache control unit 21 manages a cache table in which the following information is registered for all the caches 28 in the cluster.
・ Computer name
·file name
-Area in the file (offset, size)
Whether the cache is dirty (whether write data remains without being reflected on the disk 13)
When a certain computer 11 generates a dirty cache, the cache control unit 21 instructs each computer 11 to discard the write cache owned by the other computer 11 in the corresponding area of the corresponding file. At the time of backup, each computer 11 is instructed to write all dirty caches to the disk 13. The computer 11 instructed to write writes the write cache in the cache 28 to the disk 13.
[0048]
In this way, when all the write caches in the cluster are reflected on the disk 13, an image backup is executed by the copy management unit 25, and the data in the disk 13 is copied to the backup medium 15. As a result, backup is performed without inconsistency of data in the cluster.
[0049]
FIG. 3 is a flowchart of processing in which the cache control unit 21 reflects the write cache on the disk 13. When backup is started, the cache control unit 21 first searches the cache table and checks whether or not there is a dirty cache remaining in the cluster (step S1).
[0050]
If the dirty cache remains, the computer name, the write destination file name, and the area information are acquired (step S2). Next, the computer is instructed to reflect the dirty cache in the corresponding area of the corresponding file on the disk 13 (step S3), and the processing after step S1 is repeated. When there is no dirty cache in step S1, the process is terminated.
[0051]
Further, the log management unit 26 performs log management of (a.2) or (a.3) regarding the BI log.
FIG. 4 shows the log management of (a.2). In FIG. 4, each computer 11 has its own temporary log medium 31 and log management file 32, and the log management unit 26 edits the log of the entire system by collecting the logs of all the computers 11 after the backup is completed. Store in the log medium 14. The log management file 32 includes a log list in which the following information is recorded for each log.
-Device name of disk 13
-Area (offset, size)
·Times of Day
Among these, the time represents the time when the log is generated and is used to determine the context with other logs. Here, instead of the actual time, the logical time generated by the clock unit 33 provided in the computer 12 is used. For example, the clock unit 33 generates a logical time “1” when the first log is generated, and thereafter increments the logical time by one each time a log is generated.
[0052]
Each computer 11 sends the log of the temporary log medium 31 and the log list of the log management file 32 to the log management unit 26. The log management unit 26 is the oldest if there are multiple logs in the same area in the received logs. Edit the log, leaving things preferentially. By overwriting the edited log on the backup medium 15, the image at the time of starting backup is confirmed, and the change caused by writing to the disk 13 being backed up can be canceled.
[0053]
FIG. 5 is a flowchart of log editing processing by the log management unit 26. First, the log management unit 26 receives logs and a log list from all the computers 11 in the cluster (step S11), sorts the received logs in chronological order, and prepares a work log list (step S12). .
[0054]
Next, the oldest log is selected (step S13), and it is checked whether or not the log in the same area is in the work log list (step S14). If there is no log in the same area, the selected log is added to the work log list (step S15), and if there is a log in the same area, the selected log is discarded (step S16).
[0055]
Next, it is checked whether or not there is an unselected log (step S17). If such a log remains, the processes in and after step S13 are repeated. When all the logs are selected, the logs included in the work log list are recorded on the log medium 14 (step S18), and the process ends.
[0056]
FIG. 6 shows the log management of (a.3). In FIG. 6, each computer 11 sends a log to the log management unit 26, and the log management unit 26 records the received log in the log medium 14, and stores the corresponding device name and area (offset, size) in the log management file. 34. As described above, in the log management of FIG. 6, the log is placed on the log medium 14 from the beginning. By overwriting the log thus recorded on the backup medium 15, the image at the time of starting the backup is determined.
[0057]
FIG. 7 is a flowchart of log recording processing by the log management unit 26. The log management unit 26 first receives a log from the computers 11 in the cluster (step S21), and checks whether a log in the same area is already stored in the log medium 14 (step S22). If the log of the same area is not saved, the received log is recorded in the log medium 14 (step S23), and if the log of the same area is saved, the received log is not recorded.
[0058]
Next, it is checked whether or not logs have been received from all the computers 11 in the cluster (step S24). If there is a computer 11 that has not sent logs, the processing from step S21 is repeated. Then, when logs are received from all the computers 11, the processing is terminated.
[0059]
The log management of FIG. 4 has an advantage that the communication cost is low because the logs of the computers 11 are collectively sent to the log management unit 26 after the backup is completed. However, since the temporary log medium 31 is required, the hardware cost increases, and the log must be edited after the backup is completed, so that post-processing is required.
[0060]
On the other hand, the log management of FIG. 6 has an advantage that the temporary log medium 31 and post-processing are unnecessary. However, each time a log is generated in each computer 11, the log is sent to the log management unit 26, so the communication cost increases compared to the case of FIG.
[0061]
The log management in FIG. 6 can be further classified into (a.4) or (a.5) log management. In the log management (a.4), a log is stored in the backup medium 15 instead of the log medium 14, and in the log management (a.5), the log is stored in the log medium 14.
[0062]
FIG. 8 shows the log management of (a.4). When the backup medium 15 that is the copy destination of the disk 13 is not a tape but a disk used as a buffer for the tape 16, partial overwrite on the backup medium 15 is possible. Therefore, the copy management unit 25 and the log management unit 26 cooperate to record a log on the backup medium 15 at the same time as copying when the disk 13 to be backed up is being copied.
[0063]
At this time, first, after the log management unit 26 records the log on the backup medium 15, the copy management unit 25 copies the data on the disk 13 to the backup medium 15 for an area where no log exists.
[0064]
FIG. 9 is a flowchart of log recording processing by the log management unit 26. First, the log management unit 26 records the management information of the log to be stored in the log management file 34 as shown in FIG. 10 (step S31), and copies the log to the backup medium 15 (step S32). Next, it is checked whether there is another log that has not been copied (step S33). If there is such a log, the processing from step S31 onward is repeated. Then, when all the logs have been copied, the process is terminated.
[0065]
In the log management file of FIG. 10, the device name, the original address, and the length are recorded for each log. The device name represents identification information of the corresponding disk 13, and the original address and length represent the offset and size of the corresponding area, respectively.
[0066]
FIG. 11 is a flowchart of copy processing by the copy management unit 25. First, the copy management unit 25 starts copying using the start address of the disk 13 to be backed up as the current address (step S41), and checks whether the current address is the end address (step S42).
[0067]
If the current address is not the end address, it is next checked whether or not the address exists in the log management file 34 (step S43). If the current address does not exist in the log management file 34, the block at that address is copied to the backup medium 15 (step S44), and if it exists in the log management file 34, the block at that address is not copied.
[0068]
Next, the next address is set as the current address (step S45), and the processes after step S42 are repeated. Then, when the current address matches the end address in step S42, the process ends.
[0069]
FIG. 12 shows the log management of (a.5). When the backup medium 15 is a medium that cannot be partially overwritten, such as a tape, a log medium 14 different from the backup medium 15 is prepared, and only the log is placed on the medium. Thereby, a log can be left on a medium different from the backup medium 15. At this time, the log management unit 26 and the copy management unit 25 respectively perform log recording and disk copy independently.
[0070]
FIG. 13 is a flowchart of log recording processing by the log management unit 26. The processes in steps S51 and S53 in FIG. 13 are the same as the processes in steps S31 and S33 in FIG. 9, respectively. After the process of step S51, the log management unit 26 copies the log to the log medium 14 (step S52), and performs the process of step S53.
[0071]
FIG. 14 is a flowchart of copy processing by the copy management unit 25. The processes in steps S61, S62, S63, and S64 in FIG. 14 are the same as the processes in steps S41, S42, S44, and S45 in FIG. 11, respectively. In this case, all blocks of the disk 13 to be backed up are copied to the backup medium 15.
[0072]
In the log management of (a.6), the log management unit 26 overwrites the BI log on the backup medium 15 and saves it at the time of backup. If the log and the backup data are merged and stored in advance as described above, only the backup medium 15 needs to be referred to at the time of restoration, and there is no need to refer to a plurality of media. Therefore, restoration is made efficient.
[0073]
FIG. 15 shows log management of (a.6). In FIG. 15, the log management unit 26 merges the log and the backup data by overwriting the plurality of logs of the log medium 14 with the corresponding areas of the backup medium 15 while referring to the log management file 34. Thereafter, the tape control unit 27 stores the data of the backup medium 15 on the tape 16.
[0074]
In the log management of (a.7), the log management unit 26 records the address information of the log overwrite destination backup data together with the BI log data at the time of backup. Originally, the log cannot be accessed without referring to the log management file that is the management information of the log, but if the management information is written in the log, the log can be overwritten with the backup data only by reading the log. Therefore, the log can be eliminated without referring to the log management information, and the log management is made efficient.
[0075]
FIG. 16 shows the data format of such a log medium. In FIG. 16, the original address and length represent the offset and size of the corresponding overwrite destination area, respectively, and these correspond to log management information.
[0076]
Next, the operations relating to the features (a.8) to (a.12) and (b.1) to (b.5) described above will be described in detail with reference to FIGS.
FIG. 17 illustrates block management (a.8) and (a.9) and group management (b.1) to (b.5). The used block list 41 is used in block management (a.8), and the changed block list 42 is used in block management (a.9). The group block list 43 and the group change block list 44 are used in group management (b.1) to (b.5).
[0077]
In the block management (a.8), the block management unit 22 records and manages the blocks allocated to the files on the disk 13 in the used block list 41. The block management unit 22 notifies the copy management unit 25 of the blocks recorded in the used block list 41, and the copy management unit 25 copies only the notified block. Thus, by copying only the necessary blocks as backup data, the copy time is shortened and the required medium capacity is reduced.
[0078]
The used block list 41 is generated from, for example, a free space management table as shown in FIG. The free space management table in FIG. 18 is managed by the block management unit 22 and includes block identification information (block numbers) of all blocks on the disk 13 and flag information indicating whether each block is in use. Here, the flag “◯” represents an empty block, and the flag “×” represents a block in use.
[0079]
The block management unit 22 creates a used block list 41 by listing the block numbers of blocks in use from the free space management table at the time of backup. For example, a used block list as shown in FIG. 19 is generated from the free space management table of FIG.
[0080]
In block management (a.9), the block management unit 22 manages the blocks on the disk 13 that have been changed after the previous backup by recording them in the changed block list 42. The block management unit 22 notifies the copy management unit 25 of the blocks recorded in the changed block list 42 as differences, and the copy management unit 25 copies only the notified block. Thereby, differential backup (incremental backup) is performed.
[0081]
For example, when the file f uses blocks x, y, and z, after the previous backup, the block x is overwritten and a new block u is added. In this case, the blocks x and u are recorded in the changed block list 42, and the contents of these blocks are copied at the time of backup.
[0082]
By performing such differential backup instead of performing backup (entire backup) of all blocks on the disk 13, the copy time is shortened and the required medium capacity is reduced.
[0083]
FIG. 20 is a flowchart of changed block list update processing by the block management unit 22. In this process, the changed block is determined from the write request to the file, and the block is added to the changed block list 42.
[0084]
First, the block management unit 22 receives a file write request from the computer 11 (step S71). The write request includes the file name, the write area offset a, and the size S. Next, the block number of the block assigned to the range of a to a + S of the corresponding file is obtained (step S72), and the block number is added to the changed block list 42 (step S73). Then, the requested block is accessed, a process for writing is performed (step S74), and the process ends.
[0085]
Further, in the group management of (b.1), the group management unit 23 manages files for each group in cooperation with the block management unit 22, and blocks used by files belonging to each group are displayed in the group block list. It records in 43 and manages. The group management unit 23 notifies the copy management unit 25 of the blocks recorded in the group block list 43 of the specific group, and the copy management unit 25 copies only the notified block. Thereby, the backup regarding a specific group is performed.
[0086]
In the group management of (b.2), the group management unit 23 groups files in units of directories, and sets all files included in the directories as groups.
[0087]
In the group management (b.3), the group management unit 23 excludes a specific file or directory under the directory set as a group from the group. Thereby, a specific file included in a directory set as a certain group can be excluded from the group.
[0088]
By the group management of (b.1) to (b.3) described above, it becomes possible for the user to group arbitrary files and perform file backup in units of groups.
[0089]
As an example of such group management, consider a case where a directory tree as shown in FIG. 21 exists on the file system. In FIG. 21, A, B, C, and D represent directory names, and a, b, c, d, e, and f represent file names. The user can set a group of files using an arbitrary directory name and file name.
[0090]
Here, it is assumed that the user inputs a group list as shown in FIG. 22 and instructs group setting. In FIG. 2, “dir_A / *” and “dir_C / *” indicate that all files in directories A and C are included in the group, and “X dir_D / file_d” excludes file d in directory D from the group. Represents.
[0091]
At this time, the group management unit 23 selects the remaining files a, b, c, e, and f excluding the file d among all the files belonging to the directories A and C as a group. Then, the block number of the block allocated to each file is acquired from the block management unit 22 and recorded in the group block list 43.
[0092]
FIG. 23 shows an example of a group list relating to another directory tree. This group list indicates that the file a in the directory X, the file b in the directory Y, and all the files in the directory Z are included in the group, and the file c in the directory Z is excluded from the group.
[0093]
From this block list, for example, a group block list as shown in FIG. 24 is generated. In FIG. 24, “blockno” represents a block number, and a plurality of consecutive block numbers are recorded as one unit.
[0094]
At the time of backup, the meta information about the file, the block number recorded in the group block list 43, and the data of the corresponding block are copied to the backup medium. As the meta information, the file names and attributes of all files included in the directory tree or the file names and attributes of files belonging to the group are used.
[0095]
When a file is referred to at the time of restoration, the block management unit 22 obtains a corresponding block number from the file name and accesses the backup data of the block.
[0096]
At this time, if the information of all the files is recorded as meta information, the computer 11 can also see the file names of files that do not belong to the group, such as the file d in FIG. However, since there is no backup data for the block of file d, an error is returned when this file is referenced. On the other hand, if only the information of the files belonging to the group is recorded as meta information, the computer 11 cannot see the file names of the files that do not belong to the group, so that all the visible files can be referred to. become.
[0097]
Further, the group management unit 23 records and manages the blocks changed after the previous backup in the group change block list 44 in units of groups. The group management unit 23 notifies the copy management unit 25 of the blocks recorded in the group change block list 44 as differences, and the copy management unit 25 copies only the notified blocks. Thereby, differential backup is performed in units of groups.
[0098]
In the block lists of FIGS. 19 and 24, block numbers are explicitly recorded, but instead, a set of a plurality of consecutive blocks may be recorded using the original address and length. The same applies to other block lists.
[0099]
In (b.4) group management, the group management unit 23 sets a plurality of groups, and performs backup according to different schedules. In the group management (b.5), the group management unit 23 performs grouping that recognizes that one file belongs to a plurality of groups. This enables flexible group setting and backup.
[0100]
The hatched portion of the disk 13 in FIG. 17 represents a set of blocks recorded in one of the various block lists described above, and the data of these blocks is stored in the backup medium 15 by the copy management unit 25. Copied to the shaded area.
[0101]
According to such a method, since backup is performed based on a block list generated in advance, a plurality of blocks including blocks of different files can be copied together. Therefore, the number of accesses to the disk 13 is greatly reduced as compared with the case of copying in units of files, and a situation close to random access hardly occurs.
[0102]
In the block management of (a.10), after performing differential backup and before restoration, the differential backup data or the contents of the differential backup data and the entire backup data are merged in units of blocks. Further, two or more backup data including the whole backup data and the differential backup data may be merged. Restoration is made more efficient by collecting differential backup data in advance.
[0103]
FIG. 25 shows an example of such a merge process. In FIG. 25, the entire backup data 51 represents the first generation G3 backup data, and the differential backup data 52 and 53 represent the differences in the generations G2 and G1, respectively. In this case, the generation G3 is the oldest and the generation G1 is the newest. The hatched portion represents an area where backup data exists.
[0104]
Here, when the whole backup data 51 and the differential backup data 52 are merged, backup data 54 is generated. However, newer data is preferentially stored in the same area. In this case, at the time of restoration, data is referred to using only the backup data 54 and the differential backup data 53.
[0105]
Further, when the differential backup data 52 and the differential backup data 53 are merged, backup data 55 is generated. In this case, at the time of restoration, data is referred to using only the backup data 55 and the entire backup data 51.
[0106]
In the block management of (a.11), when performing differential backup, the user selects a time point that is a reference for the difference, and the block management unit 22 selects only blocks that have changed since the specified time point. The change block list 42 is recorded. Then, the copy management unit 25 copies only those blocks.
[0107]
As a result, the starting point of the differential backup can be changed as necessary, and changes that occur between the time when the previous backup was performed and the selected time are not stored in the backup medium 15. Therefore, it is possible to select changes to be reflected in the restoration.
[0108]
FIG. 26 shows an example of changing the starting point of differential backup. Here, the previous differential backup is performed at time t0, blocks x and y are added to the changed block list 42 between times t0 and t1, and block z is added to the changed block list 42 between times t1 and t2. Shall be. If the user does not change the starting point of the differential backup, the blocks x, y, and z are recorded as differences from the time t0 in the changed block list 42 at time t2.
[0109]
However, when the user designates the time t1 as the starting point of the differential backup, the block management unit 22 once clears the changed block list 42 and erases the block numbers of the blocks x and y at the time t1. Thereafter, the block z is added to the changed block list 42, and at time t2, only the block z is recorded as a difference from the time t1. Then, at the time of the next backup, differential backup is performed based on the difference from time t1.
[0110]
Further, in the copy management of (a.12), when a plurality of disks 13 exist in the cluster, the copy management unit 25 instructs each computer 11 to copy any one of the disks 13 and performs the copy within the cluster. The computers 11 to be distributed are distributed. Then, each computer 11 performs copying of the disk 13 instructed from the copy management unit 25. As described above, when a plurality of computers 11 perform copying, the backup load is distributed and the copying time is shortened.
[0111]
FIG. 27 shows an example of such copy management. In the cluster of FIG. 27, a plurality of backup media 15 are provided. The copy management unit 25 notifies each computer 11 of the device names of the copy target disk 13 and the copy destination backup medium 15 and requests a copy operation. The computer 11 requested to copy the data copies the data of the notified disk 13 having the device name to the backup medium 15 having the notified device name. At this time, the copy work is performed in parallel by the plurality of computers 11.
[0112]
Next, with reference to FIG. 28 to FIG. 35, the operation at the time of restoration related to the above-described features (c.1) to (c.5) will be described in detail.
In the operation of (c.1), the file system restores data by mounting the backup medium 15 as it is instead of the disk 13. As a result, each computer 11 can directly access the backup data stored in the backup medium 15, and a special operation for restoration is not required.
[0113]
FIG. 28 shows a process in which the block management unit 22 mounts the backup medium 15 on the file system. In FIG. 28, after data on the disk 13 (shaded portion) is copied to the backup medium 15 and a reference request for the disk 13 is received from the computer 11, the block management unit 22 transfers the corresponding data on the backup medium 15 to the computer. Return to 11.
[0114]
FIG. 29 is a flowchart of such a reference process. The block management unit 22 first receives a file read request from the computer 11 (step S81). The read request includes the file name, the read area offset a, and the size S. Next, the meta information of the backup medium 15 is referred to (step S82), and the block number #x assigned to the range a to a + S of the corresponding file is obtained (step S83).
[0115]
Next, the block number #y on the backup medium 15 in which the data of the block number #x is stored is obtained (step S84), the block data is read (step S85), and the read request is responded ( Step S86), the process is terminated.
[0116]
In the operation of (c.2), when restoring the differential backup, the plurality of differential backup data are searched in order from the latest one as necessary. In the case of differential backup, since necessary data is stored in any of a plurality of generations of backup data, all data can be presented to the computer 11 by searching for the backup data.
[0117]
FIG. 30 shows such generation management. In FIG. 30, backup data for each generation is stored in different backup media 15. The block management information 61 maps the file name to the device name and block number of the disk 13. The block management information 62 is provided for each generation of backup data, and maps the device name and block number of the disk 13 to the identification information and block number of the backup medium 15.
[0118]
When receiving a file access request from the computer 11, the block management unit 22 refers to the block management information 61 to acquire a device name and a block number corresponding to the file name, and passes them to the medium control unit 24.
[0119]
The medium control unit 24 manages the backup data of each generation in the order of generation, refers to the block management information 62 of the newest generation G1, and checks whether there is a given device name and block number (block information) To do. If there is the given block information, the block number of the backup medium 15 corresponding to the block information is obtained, and the backup medium 15 of the generation G1 is referred to. If there is no given block information, the block management information 62 of the previous generation G2 is referred to and it is checked whether or not the block information exists.
[0120]
If the generations are traced back one by one while repeating such processing, the data corresponding to the given block information can be referred to in the backup medium 15 of any generation. Thus, even when differential backup is performed, it is possible to show all data to the user using past backup data.
[0121]
FIG. 31 shows an example of differential backup restoration. In FIG. 31, overall backup data 71 represents the first generation G3 backup data, and differential backup data 72 and 73 represent the differences in generations G2 and G1, respectively. The hatched portion represents a block in which backup data exists. For example, in the differential backup data 72, the blocks 81 and 82 correspond to the changed data, and the block 83 corresponds to the newly added data.
[0122]
When data of blocks 93 and 94 is requested at the time of restoration, the corresponding block of the differential backup data 73 of generation G1 is referred to, and when data of blocks 92 and 97 is requested, the differential backup is traced back to generation G2. A corresponding block of data 72 is referenced. Further, when the data of the blocks 91, 95, 96 are requested, the corresponding block of the whole backup data 71 is referred back to the generation G3.
[0123]
Further, when the merge process as shown in FIG. 25 is performed, a similar restore operation is performed using the backup data generated by the merge instead of the merged two backup data.
[0124]
In the operations (c.3) and (c.4), when backup data is stored on the tape 16, the backup medium 15 is used as a buffer accessible from all the computers 11 in the cluster. When the computer 11 refers to the backup data, only necessary blocks are loaded onto the backup medium 15 and those blocks are used as a cache. As a result, only necessary data can be arranged on the backup medium 15, and access efficiency to frequently accessed data is improved.
[0125]
FIG. 32 shows processing for using the backup medium 15 as a buffer. Here, since only the computer 12 is connected to the tape 16 in the configuration of FIG. 2, the tape control unit 27 of the computer 12 reads out necessary block data from the tape 16 and arranges it on the backup medium 15. . For example, a disk is used as the backup medium 15. By referring to the data loaded on the backup medium 15 in this way, the computer 11 not connected to the tape 16 can read the backup data stored on the tape 16.
[0126]
FIG. 33 is a flowchart of backup data reference processing. The processes in steps S91 to S93 in FIG. 32 are the same as the processes in steps S81 to S83 in FIG. Next, the block management unit 22 obtains the block number #y on the tape 16 where the data of the block number #x obtained in step S93 is stored (step S94), and the cache of the block is stored in the backup medium 15. It is checked whether it is above (step S95).
[0127]
If the cache is not on the backup medium 15, the tape control unit 27 reads the data of the block from the tape 16 and writes it in an empty block #z on the backup medium 15 (step S96). Then, the block management unit 22 responds to the read request using the written data (step S97), and ends the process. If the cache is on the backup medium 15, the block management unit 22 responds to the read request using the data (step S97) and ends the process.
[0128]
In the operation of (c.5), when the BI log of the log medium 14 is not overwritten with the backup data of the backup medium 15 and the log and the backup data are stored separately, the log is referred to first. Refer back to the backup data later if necessary. By referring to the log at the time of restoration, the image at the start of the backup is reproduced, and data inconsistency does not occur.
[0129]
FIG. 34 shows processing for referring to the log of the log medium 14. When receiving an access request from the computer 11, the block management unit 22 confirms the existence and contents of the log in the log management unit 26. For example, the backup medium 15 is referred to.
[0130]
FIG. 35 is a flowchart of such a reference process. The processing in steps S101 to S103 in FIG. 35 is the same as the processing in steps S81 to S83 in FIG. Next, the block management unit 22 inquires of the log management unit 26 whether there is a log of the block number #x obtained in step S103 (step S104), and checks the answer (step S105).
[0131]
If there is a log of the inquired block, the log on the log medium 14 is read. If there is no such log, the block number #y on the corresponding backup medium 15 is obtained and the backup data of the block is read. (Step S107). Then, in response to the read request (step S108), the process ends.
[0132]
In the cluster system of FIG. 2, the cache control unit 21, the block management unit 22, the group management unit 23, the medium control unit 24, the copy management unit 25, the log management unit 26, and the tape control unit 27 are managed by a management computer. However, some or all of these control units and management units may be distributed among a plurality of computers 11.
[0133]
FIG. 36 shows a backup operation in such a cluster system. In FIG. 36, the copy management unit 25, the log management unit 26, and the tape control unit 27 are provided in a distributed manner in the computer 11, and the computer 11 having the copy management unit 25 copies the contents of the disk 13 to the backup medium 15. To do.
[0134]
Each computer 11 transfers the BI log generated during the copying to the computer 11 having the log management unit 26, and the computer 11 edits the log and writes it in the log medium 14. Then, the backup data and the log are written on the tape 16 by the computer 11 having the tape control unit 27, respectively.
[0135]
FIG. 37 shows the restore operation in the cluster system of FIG. In FIG. 37, the computer 11 having the tape control unit 27 receives a read request from another computer 11 and separately reads the necessary backup data and log from the tape 16, and the backup medium 15 and the log medium 14 respectively. Expand on top.
[0136]
When the expansion is completed, the computer 11 that has requested reading reads the log from the log medium 14 if there is a log. If there is no log, the block number corresponding to the file name is received from the block management unit 22 and the backup data is read from the backup medium 15.
[0137]
However, when the computer 11 having the tape control unit 27 requests reading, it is not necessary to expand the backup data and log on the backup medium 15 and the log medium 14.
[0138]
FIG. 38 shows a case where the log is overwritten before the backup data is written on the tape 16 in the cluster system of FIG. In FIG. 38, the log management unit 26 overwrites the backup data of the backup medium 15 with the log of the log medium 14 when the backup data and the log are finalized. When overwriting is completed, the tape control unit 27 writes the overwritten backup data on the tape 16.
[0139]
FIG. 39 shows a restore operation in the cluster system of FIG. In FIG. 39, the tape control unit 27 reads the backup data with the log overwritten from the tape 16 and develops it on the backup medium 15. Then, the computer 11 that has requested reading reads the necessary data from the backup medium 15. However, when the computer 11 itself having the tape control unit 27 requests reading, it is not necessary to expand the data on the backup medium 15.
[0140]
The computers 11 and 12 in FIG. 2 can be configured using, for example, an information processing apparatus as shown in FIG. 40 includes a CPU (central processing unit) 111, a memory 112, an input device 113, an output device 114, an external storage device 115, a medium driving device 116, and a network connection device 117, which are connected via a bus 118. Are connected to each other.
[0141]
The memory 112 includes, for example, a ROM (read only memory), a RAM (random access memory), and the like, and stores programs and data used for processing. The CPU 111 performs necessary processing by executing a program using the memory 112.
[0142]
The cache control unit 21, block management unit 22, group management unit 23, medium control unit 24, copy management unit 25, log management unit 26, and tape control unit 27 in FIG. 2 are, for example, software components described by a program. Stored in the memory 112.
[0143]
The input device 113 is, for example, a keyboard, a pointing device, a touch panel, and the like, and is used for inputting instructions and information from the user. The output device 114 is, for example, a display, a printer, a speaker, or the like, and is used for outputting an inquiry to a user and a processing result.
[0144]
The external storage device 115 is, for example, a magnetic disk device, an optical disk device, a magneto-optical disk device, a tape device, or the like. The information processing apparatus stores the above-described program and data in the external storage device 115, and loads them into the memory 112 for use as necessary. The external storage device 115 is used as the shared disk 13, the log medium 14, the backup medium 15, the tape 16, and the like.
[0145]
The medium driving device 116 drives the portable recording medium 119 and accesses the recorded contents. As the portable recording medium 119, an arbitrary computer-readable recording medium such as a memory card, a floppy disk, a CD-ROM (compact disk read only memory), an optical disk, or a magneto-optical disk is used. The user stores the above-described program and data in the portable recording medium 119 and loads them into the memory 112 and uses them as necessary.
[0146]
The network connection device 117 is used for connection to a communication network that connects computers, and performs data conversion accompanying communication. The information processing apparatus receives the above-described program and data from another apparatus via the network connection apparatus 117, and loads them into the memory 112 and uses them as necessary.
[0147]
FIG. 41 shows a computer-readable recording medium that can supply a program and data to the information processing apparatus of FIG. Programs and data stored in the portable recording medium 119 or the external database 120 are loaded into the memory 112. Then, the CPU 111 executes the program using the data and performs necessary processing.
[0148]
【The invention's effect】
According to the present invention, in a computer system having a disk shared file system such as a cluster system, data can be efficiently backed up while the system is operating. In addition, backup data can be referred to efficiently at the time of restoration.
[Brief description of the drawings]
FIG. 1 is a principle diagram of a backup system according to the present invention.
FIG. 2 is a configuration diagram of a cluster system.
FIG. 3 is a flowchart of cache control processing.
FIG. 4 is a diagram illustrating first log management;
FIG. 5 is a flowchart of log editing processing.
FIG. 6 is a diagram illustrating second log management.
FIG. 7 is a flowchart of first log recording processing;
FIG. 8 is a diagram illustrating third log management;
FIG. 9 is a flowchart of second log recording processing;
FIG. 10 is a diagram showing a log management file.
FIG. 11 is a flowchart of first copy processing.
FIG. 12 is a diagram illustrating fourth log management.
FIG. 13 is a flowchart of third log recording processing;
FIG. 14 is a flowchart of second copy processing;
FIG. 15 is a diagram illustrating fifth log management;
FIG. 16 is a diagram illustrating a data format of a log medium.
FIG. 17 is a diagram illustrating block management and group management.
FIG. 18 is a diagram showing a free space management table.
FIG. 19 is a diagram illustrating a used block list.
FIG. 20 is a flowchart of changed block list update processing;
FIG. 21 is a diagram showing a directory tree.
FIG. 22 is a diagram showing a first group list.
FIG. 23 is a diagram illustrating a second group list.
FIG. 24 is a diagram showing a group block list.
FIG. 25 is a diagram illustrating merging of differential backup data.
FIG. 26 is a diagram showing a change at the start point of differential backup.
FIG. 27 is a diagram showing copy management.
FIG. 28 is a diagram illustrating mounting of a backup medium.
FIG. 29 is a flowchart of first reference processing;
FIG. 30 is a diagram illustrating generation management.
FIG. 31 is a diagram showing differential backup restoration;
FIG. 32 is a diagram showing a backup medium as a buffer.
FIG. 33 is a flowchart of second reference processing.
FIG. 34 is a diagram illustrating log reference.
FIG. 35 is a flowchart of third reference processing.
FIG. 36 is a diagram showing a first backup.
FIG. 37 is a diagram illustrating first restoration.
FIG. 38 is a diagram showing a second backup.
FIG. 39 is a diagram showing a second restoration.
FIG. 40 is a configuration diagram of an information processing apparatus.
FIG. 41 is a diagram illustrating a recording medium.
[Explanation of symbols]
1 Copying means
2 Control means
3, 11, 12 Calculator
4 Shared media
5, 15 Backup media
6 Log management means
7 Generation means
8 Group management means
9 Area management means
13 Shared disk
14 Log media
16 tapes
21 Cache control unit
22 Block management department
23 Group Management Department
24 Medium control unit
25 Copy Management Department
26 Log management department
27 Tape controller
28 cash
31 Temporary log media
32, 34 Log management file
33 Clock part
41 Used Block List
42 Changed block list
43 Group Block List
44 Group change block list
51, 71 Whole backup data
52, 53, 72, 73 Differential backup data
54, 55 Backup data
61, 62 Block management information
81, 82, 83, 91, 92, 93, 94, 95, 96, 97 blocks
101 Access request
111 CPU
112 memory
113 Input device
114 output device
115 External storage device
116 Medium drive device
117 Network connection device
118 bus
119 Portable recording media
120 database

Claims (9)

A backup system for performing image backup of a shared medium shared by a plurality of computers,
When any one of the plurality of computers writes in a certain area of the shared medium , the image data before writing in the area is managed by the computer that performed the writing as a log,
Log management means for generating logs as a whole by collecting logs managed by the plurality of computers, and when there are a plurality of logs for the same area, log management means using the oldest log after the image backup start time ,
During the restore, the backup system comprising: a generating means for generating data of the image backup start time using the whole log.   Temporary log storage means for temporarily saving each log of the plurality of computers is further provided, and the log management means edits each log saved in the temporary log storage means to generate the entire log The backup system according to claim 1, wherein:   When one of the plurality of computers accesses the shared medium, the log management means receives an access notification from the accessed computer and stores the log of the accessed computer, thereby saving the entire log. The backup system according to claim 1, wherein the backup system is generated.   2. The backup system according to claim 1, further comprising backup storage means for storing backup data of the shared medium, wherein the log management means stores the entire log in the backup storage means.   The backup system according to claim 1, further comprising backup storage means for storing backup data of the shared medium, and log storage means for storing the entire log.   The backup system according to claim 1, further comprising backup storage means for storing backup data of the shared medium, wherein the log management means overwrites the entire log on the backup data.   Log storage means for storing a log managed by the log management means and address information of backup data overwritten by the log, and the generation means converts the log into corresponding backup data based on the address information; 2. The backup system according to claim 1, wherein overwriting is performed.   The generation unit refers to the entire log first when the entire log is not overwritten on the backup data of the shared medium, and refers to the backup data later if necessary. The backup system according to claim 1. A backup method for performing an image backup of a shared medium shared by a plurality of computers,
After the start of image backup,
When any one of the plurality of computers writes in a certain area of the shared medium , the image data before writing in the area is managed by the computer that performed the writing as a log,
When the computer that performs the image backup generates a whole log by collecting logs managed by the plurality of computers, and there are a plurality of logs for the same area when the whole log is generated, the image backup start time or later Use the oldest log,
During the restore, to generate data of the image backup start time using the whole log,
A backup method characterized by that .

Images