JP3251182B2 - Mirror disk control method and mirror disk control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a mirror disk control system, and more particularly to a process for matching the contents of disks when a system goes down.

[0002]

2. Description of the Related Art Conventionally, there is a technique of duplicating or multiplexing a disk device in order to improve the reliability of data at the time of occurrence of a failure or to pursue high-speed data search processing. In this method, since a plurality of physical disks are virtually treated as one, all the contents of each disk must be matched. For this purpose, generally, a disk control device is installed between each disk and the CPU, and a so-called mirror disk function for making all the contents of each disk identical is exhibited.

The disk control unit matches the contents of each disk by writing the same data to each disk device during system operation. However, the problem occurs when an unexpected forced system down occurs due to a failure or the like while the same data is being written to each disk device. Conventionally, this problem has been recovered by the following method.

FIG. 11 is a configuration diagram of a conventional mirror disk control device described in Japanese Patent Application Laid-Open No. 7-319637. A disk device 103 is connected to each of two disk control devices 120 connected to a host computer having a CPU 101. Each disk control device 120 includes a disk input / output device 104 for inputting / outputting data to / from the disk device 103, a non-volatile backup memory 105, and a communication unit 107 for performing communication with another disk control device 120. Have. Further, the backup memory 105 includes a disk information block information group 106, which includes information on an area (logical block) which is managed by dividing the disk device 103 satisfying a predetermined condition into blocks. Is written.

The operation of the first conventional configuration when a write request is made to the disk device 103 will be described.

In a normal operation, when there is a write request from the CPU 101 to the disk device 103, the CPU 101
Reference numeral 101 denotes a logical block address to which data to be written is stored, and a block information group 106 during disk writing.
a and 106b are sequentially written. Each disk controller 12
0a and 120b are disk information block information group 1
The data is written in a predetermined logical block based on the contents of 06a and 106b. After mutually confirming the end of the writing process of each logical block by the communication means 107, the block information groups 106a, 106
The logical block address written in b is erased. By repeating this process constantly, the contents of the disk devices 103a and 103b always match.

If an unexpected system down occurs during this write processing, each disk drive 10
The contents of 3a and 103b do not always match.
For example, even if data writing of the same logical block is performed on both disk devices 103a and 103b, the writing process on the disk device 103a ends immediately before down, and the writing process on the disk device 103b has not been completed yet. There may not be. That is,
Since the writing process of the same data to the same logical block does not always end at the same time, each disk device 1
There is no guarantee that the contents of 03a and 103b always match.

In this first conventional example, the contents of the disk writing block information group 106 are checked at the time of restart after a system failure, and if a logical block address still remains, the data of the relevant logical block is read. From one to the other disk drive. By this processing, the contents of the disk devices 103a and 103b are matched. It should be noted that the disk writing block information group 106
Is stored in the non-volatile backup memory 105, so that its contents are not lost even when the system goes down.

FIG. 12 is a configuration diagram of a conventional duplicate volume equivalence discrimination method described in Japanese Patent Application Laid-Open No. 4-369710. FIG. 12 shows a user program 2
Duplex volume 205 based on an I / O request from
Volume update determination means 202 provided for performing input / output control on the
3. Duplicated volume management table 20 in which duplicated input / output ending means 204 and equivalence determination flag 209 are set
8 is shown. In the duplicate volume 205, the primary volume 206 and the secondary volume 207 are duplicated. Further, there is shown a duplicate volume non-equivalence notifying unit 210 that notifies the operator by displaying on the system console 211 that the equivalence of the contents of the duplicate volume 205 cannot be guaranteed.

The operation when a write request to the duplex volume 205 is made in the configuration of the second conventional example will be described.

During normal operation, the dual volume 2
When a write request is made to the duplicate volume 05, the equivalence determination flag 209 corresponding to the duplicated volume 205 is set to "invalid" before starting the write processing. And duplicated volume 2
When the writing of both the primary volume 206 and the secondary volume 207 included in the disk drive 05 has been completed, the equivalence determination flag 209 is set to “valid”.

If an unexpected system down occurs during this write processing, the primary volume 206 immediately after the down
And the contents of the secondary volume 207 do not always match, as described above.

Therefore, in the second conventional example, when the system is shut down and restarted, the equivalence determination flag 20
9, if the contents are “invalid”, it is determined that the contents of the primary volume 206 and the secondary volume 207 are unequal, and the volume names are displayed on the system console 211 to urge the recovery. Like that.

[0014]

However, in the above conventional example, information is set every time a write request is made to the disk device, that is, writing to the block information group being written to the disk or setting of validity / invalidity to the equivalence determination flag. Therefore, there is a problem that the overhead for the writing process to the disk device becomes very large. In particular, even when writing to the same area occurs consecutively, the same information must be set each time.

In the first conventional example, special hardware, that is, a non-volatile backup memory needs to be used, so that there is a problem that the apparatus cost is increased.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide a mirror disk control method and method for reducing overhead in mirror disk control.

[0017]

In order to achieve the above object, a mirror disk control method according to a first aspect of the present invention provides a memory, a plurality of disk devices constituting a mirror disk, and a disk device. in mirror disk control method and a mirror disk controller for chromatic mirror disk control function to match constantly all the content the disk device by writing the same data at the same time, a predetermined process of writing a series of data to the disk device Is divided into a plurality of write processing groups based on the conditions of
During normal system operation, if a write request is issued,
Write the data resulting from the write processing to each of the disk devices, and write the storage position information of the data to both the memory and each of the disk devices only when the storage position information of the data has not yet been written to the memory; At least during the processing of the write processing group is held, and when all of the write processing included in the write processing group is completed, the storage position information held in the memory and the disk devices is initialized, At the time of system recovery, only the data corresponding to the storage position information held in each of the disk devices is copied from one of the disk devices to another of the disk devices so that the contents of the respective disk devices are matched. Things.

The mirror disk control method according to the second aspect of the present invention includes a memory, a mirror disk, a plurality of disk devices for writing data in one or a plurality of data blocks, and a plurality of disk devices. and <br/> mirror disk controller for chromatic mirror disk control function to match constantly all the content the disk device by writing the same data at the same time, is composed of the same contents to each of the memory and the disk devices, the That is set in the access history information table including access history information indicating whether data is written to the data block included in the disk device and the access history information in the memory corresponding to the data block to be written. Only when there is no memory corresponding to the data block A write presence / absence determining means for setting the access history information of each disk device to that effect, and a write process for dividing a series of data write processes to the disk device into a plurality of write process groups based on predetermined conditions Group forming means, and table initializing means for simultaneously initializing the contents of the memory and the access history information table on the disk device at the time of completion of all write processing constituting each write processing group, This guarantees the equivalence of the disk devices for each write processing group.

[0019] The third invention is the second invention, the
Mirror disk control means, at the time of the system down recovery
A of said access history information table before the SL disk device
Referring to access history information, when the setting indicating that a data write target is only the data block corresponding to the access history information, the other from the disc device
It is intended to match the contents of the respective disk devices by copying square to the disk device.

According to a fourth aspect of the present invention, in the mirror disk control method according to the second aspect, the write processing group forming means includes an access count definition area for presetting a maximum access history information setting number for the access history information table. And an access counter that counts the number of times a write process has been performed on a data block to which data has not been written since the access history information table was initialized, and the access history information table was initialized. After that, one write processing group is formed by the write processing executed until the count number of the access counter reaches the maximum access history information set number, and the table initialization unit performs the counting of the access counter. Access when the number reaches the maximum access history information set number The contents of the gravel information table is intended to be initialized.

In a fifth aspect based on the second aspect, the write processing group forming means has a timer interrupt generation section for generating a timer interrupt , and
Data write processing is divided by timer interrupt
And the table initialization means initializes the contents of the access history information table each time a timer interrupt occurs .

In a mirror disk control method according to a sixth aspect of the present invention, a memory and a mirror disk are configured, a plurality of disk devices for writing data in one or a plurality of data blocks, and the same data are stored in each of the disk devices. a mirror disk control unit that have a mirror disk control function to match constantly all the content the disk device by writing simultaneously made of the same contents to each of the memory and the disk devices, the data contained in the disk device An access history information table including access history information indicating a point in time at which data is written to a block; and a process of writing a series of data to each of the disk devices is divided into a plurality of write process groups based on a predetermined condition, and a write process is performed. Processing group forming means for holding the identification information of the writing processing group And write identification means for setting the identification information of the write processing group in the access history information corresponding to the data block to which data is to be written. Guarantee.

According to a seventh invention, in the mirror disk control system according to the sixth invention, the mirror disk control means is provided.
, The latest of the writing from the access history information table before the Symbol all disk device at the time of the system down recovery
Find the identity of treatment groups, the data block only corresponding to the access history information identification information of the latest of the writing process group that has been the search is set to copy from the disk device to another side of said disk device In this way, the contents of the respective disk devices are matched .

According to an eighth aspect of the present invention, in the second or sixth aspect, when the memory and each of the disk devices have the same contents and the continuous data block to be written is one block, A mapping table that defines storage location identification information and a size of the block, and sets access history information in the access history information table on the memory and the disk device so as to correspond to each of the blocks; The block and the access history information correspond to n: 1 (n is a natural number) .

In a ninth aspect based on the eighth aspect, the block division is set based on a logical division understood by a file system .

In a tenth aspect based on the eighth aspect, the block division is set based on a physical division of the disk device .

In an eleventh aspect based on the eighth aspect, the block division is set based on a logical access unit unique to a file system .

[0028] The twelfth aspect of the present invention, in the eighth, collected
If Ri types of data handled is beforehand recognizable, the delimiter of the block shall be set based on the type of the data
It is.

According to a thirteenth aspect, in the eighth aspect, there is provided a mapping table automatic setting means for automatically setting the contents of the mapping table in accordance with the status of the writing process to the disk device .

According to a fourteenth aspect, in the thirteenth aspect,
The mapping table automatic setting unit includes: an access history holding unit that holds, as history information, a storage position information and a size of data in each writing process to the disk device; And a table creation unit for creating the mapping table .

Furthermore, the invention of the first 5, in the thirteenth invention, the mapping table automatic setting means, said access history information table is to operate at a timing that is initialized.

[0032]

Preferred embodiments of the present invention will be described below with reference to the drawings. The same elements as those in the conventional example are denoted by the same reference numerals, and description thereof will be omitted.

Embodiment 1 FIG. 1 is a configuration diagram showing a first embodiment of a mirror disk control system according to the present invention. FIG. 1 shows disk devices 2a and 2b constituting a mirror disk, a mirror disk control device 4 having a mirror disk function and controlling and managing the disk device 2, and a higher-level computer on which a user process for accessing a disk operates. 6 and a memory 8 for holding an access history information table and the like.

Access history information tables 81 and 2 set in the memory 8 and the disk devices 2a and 2b, respectively.
1a and 21b are all composed of information having the same contents, and one is set for each. The access history information tables 81 and 21 include access history information as flag information indicating the presence or absence of data writing corresponding to each data block of the disk device 2. As the access history information, “1” (set) is set by the mirror disk controller 4 when the corresponding data block has been accessed for a predetermined time, and “0” (clear) when the corresponding data block has not been accessed. Is done. Therefore, in the present embodiment, each piece of access history information can be composed of one bit. The access counter 83 on the memory 8 is a counter that counts the number of times a write process has been issued to a data block to which data has not been written since the access history information table 81 was initialized. The access count definition areas 82, 22a, and 22b set in the memory 8 and the disk devices 2a and 2b are all composed of the same information, and are set one for each. In the access count definition areas 82, 22a, 22b, condition values for initializing the contents of the access history information tables 81, 21a, 21b, that is, for setting all the access history information to "0" are defined in advance. This value can be arbitrarily set by the system designer. In the present embodiment, the initialization is performed when a predetermined condition, that is, the count value of the access counter 83 reaches the value of the access count definition areas 82 and 22. The number of times of access (maximum number of access history information settings) at which the initialization is performed is set. The access history information table 81 and the like may be stored in the memory 8 of either the mirror disk control device 4 or the computer 6,
The holding means can be constituted by a general main storage device. In the present embodiment, the memory 8 has the access history information table 81 and the access count definition area 82 having the same contents as those of the disk device 2. As a result, it is possible to prevent a reduction in processing speed due to access to the access history information table 81 and the like. On the other hand, since the disk device 2 is provided with the access history information table 21 and the access count definition area 22 having the same contents as the memory 8, even if an unexpected system failure occurs, data loss can be prevented.

The mirror disk control device 4 includes a conventional disk access control unit 41 having a function of controlling access to all the disk devices 2 and a conventional mirror disk control function, and the access history information tables 81 and 81.
1, a dual control unit 42 that performs a mirror disk control function peculiar to the present embodiment by controlling and managing a characteristic configuration such as the access count definition areas 82 and 22 and the access counter 83 in the present embodiment.

In the present embodiment, the presence / absence of access to the disk device is managed using flag information, as in the conventional example. A characteristic of this embodiment is that a write request is accepted and the flag information is temporarily stored. After setting,
Even if data is written in a predetermined data block and the contents of all the disk devices 2 become equivalent, the flag information is not cleared at that time. Then, when a predetermined condition is satisfied, specifically, when the value of the access counter 83 based on the write request to the disk device 2 reaches the set value of the access count definition areas 82 and 22, the flag information is deleted. It is to initialize all at once. That is, by minimizing the number of times of setting / clearing the access history information indicating the presence / absence of the write process, it is possible to suppress the overhead in the process of writing data to the mirror disk.

Next, the operation of the mirror disk control device 4 in this embodiment will be described with reference to the flowchart shown in FIG. The access history information table 81 at the time of system startup and normal termination of the system,
The setting of 21a and 21b will be described later.
The operation when the system is operating normally will be described. It is assumed that the access count t is set in the access count definition areas 82 and 22 in advance.

When the mirror disk controller 4 first receives an access request from a user process, it checks whether it is a write request or a read request (step 10).
1). If it is a read request, the disk access control unit 4
1, the corresponding data is read from the disk device 2 and passed to the user process (step 111). In the case of a write request, information relating to the data block to be written to by the duplication control unit 42, for example, the disk device 2
Get the logical address of Then, the content of the access history information corresponding to the data block is obtained by referring to the access history information table 81 in the memory 8 (step 102).

If the access history information is "0", that is, if the access history information is not set (step 103), the writing has not been performed on the data block before the current write request after the initialization of the access history information table 81. Can be determined. At this time, the access counter 83 is referred to and compared with the access count t set in the access count definition area 82 (step 104).
As a result, the value of the access counter 83 becomes the access number t
If not, the value of the access counter 83 is incremented by 1 (step 107). Then, the memory 8 and the access history information table 8 of each disk device 2a, 2b
After setting “1” to the access history information corresponding to the data blocks of the data blocks 1, 21a and 21b (steps 108 and 109), the write data is stored in the disk device 2
(Step 110).

On the other hand, if the value of the access counter 83 has reached the access count t in step 104,
The memory 8 and the access history information tables 81, 21a, 21b of the disk devices 2a, 2b are initialized, that is, all the access history information is set to "0" (step 10).
5). Then, the access counter 83 is reset (step 106). If there is only a write request such that writing is performed one data block at a time, when the value of the access counter 83 reaches the access count t, t access history information tables 81, 21a, and 21b include t This means that “1” has existed. Thereafter, as described above, based on a write request for a data block to which data has not been written, the access counter 8
3 and the memory 8 and each of the disk devices 2a, 2b
Is set to predetermined access history information in the access history information tables 81, 21a, and 21b, and write data is written to the disk device 2 (steps 107 to 11).
0).

If the access history information obtained in step 102 by receiving the write request is set in the above-described processing (step 108) (step 103), the corresponding data block is already based on the write request. Since it can be determined that an access has been made, processing for only writing the write data to the disk device 2 is performed (step 110).

As described above, according to the present embodiment,
When writing data to a predetermined data block of the disk device 2, if “1” has already been set in the access history information corresponding to the data block, that is, if the data block has already been written, Since the access history information is not set to "1" again, it is possible to reduce the overhead required for the writing process. Then, when a predetermined condition is satisfied without clearing the corresponding access history information every time data is written, that is, when the access counter 83
Is collectively cleared when the number of accesses reaches the number of accesses t, it is possible to reduce the overhead for the mirror disk function at the time of write processing.

At the end of the system, all the contents of the access history information table 21 stored in the disk device 2 may be cleared and the operation may be ended. During normal operation,
This is because the data is normally written to the disk device 2 regardless of the value of the access history information. Note that the access history information table 81, the access count definition area 82, and the access counter 83 on the memory 8 may be lost when the system goes down. Access history information table 81 and access count definition area 82
Is stored in the disk device 2, and the access counter 83 is always "0" at the next use since the contents of the access history information table 21 are cleared. As described above, when the system is normally terminated, it is only necessary to initialize the contents of the access history information table 21, so that the overhead at the time of system termination for exerting the mirror disk function can be reduced.

Next, FIG.
This will be described with reference to the flowchart shown in FIG.

First, when the system is started, the contents of the access history information table 81, the access count definition area 82, and the access counter 83 of the newly formed memory 8 are initialized (step 121) and stored in the disk device 2. The contents of the access history information table 21 and the access count definition area 22 are read into the corresponding tables 81 and 82 on the memory 8 (step 12).
2). At this time, if the contents of the access history information tables 21a and 21b seem to be different due to an unexpected system down or the like, the access history information is read as "1". After the reading is completed, it is checked whether the access history information table 81 has access history information of "1" (set) (step 12).
3). As described above, if the system has been normally terminated, all the contents of the access history information table 21 should have been cleared, but "1" remains due to an unexpected system down. If so, the contents of the corresponding data block may not be equivalent to the disk device 2. For this reason, the data block is copied from the positive disk, for example, the disk device 2a to the secondary disk device 2b (step 1).
24). As a result, the contents of the data blocks whose match was not guaranteed match. Thereafter, the access history information tables 81 and 21 of the memory 8 and the disk device 2 are stored.
Are initialized (step 125), and the routine shifts to the normal processing shown in FIG.

As described above, according to the present embodiment, instead of copying the entire contents of the disk device 2 at the time of starting the system, only the data blocks whose contents are not guaranteed to match are copied. It is also possible to reduce the overhead at the time of starting the system for exerting the mirror disk function.

As described above, according to the present embodiment, the access history information provided for exhibiting the mirror disk function does not have to be set every time a write process is performed.
Since clearing (initialization) is performed at once, overhead in data write processing can be reduced.

Note that the initialization is performed collectively,
The data block in which the access history information is set at the beginning of the counting of the number of accesses after the initialization is identical until the next initialization, even though the contents are identical with other disk units. Information is not retained. In this state, when an unexpected system failure occurs, in order to make the contents of all the disk devices 2 coincide, the data block in which the access history information is set is transferred from one disk device 2a to the other disk device 2b. This copy process (step 124 in FIG. 3) requires more processing time than the first conventional example in which access history information (flag information) is updated every time a disk is accessed. However, the processing time is shorter than that of the second conventional example in which copying is performed in disk units.

However, while the computer 8 is operating normally, there should be no problem even if the initialization process is performed collectively. Considering the frequency of unexpected system downtime, The fact that the number of times of setting / clearing the flag information can be reduced as in the present embodiment and the overhead required for the mirror disk function processing can be greatly reduced is considered to be very effective.

In addition, the fact that the equivalence of the data subjected to the writing process during the time when the value of the access counter 83 has reached the number of accesses t after the initialization is not considered to have been guaranteed does not mean that the writing has been performed t times. When the processing is considered as one write processing group, it is conceivable that data equivalence is guaranteed for each group. That is, a series of write processing is divided into t pieces, and the write processing group being processed is t
When the contents of the access history information tables 81 and 21 are initialized, the equivalence of the data processed by the group is guaranteed. Then, when the processing in the next write processing group starts and the system goes down at that time, the data of the write processing group to be processed at the time when the system down occurs is guaranteed regardless of the actual completion of the write processing. Only the data blocks to be processed in all the write processes constituting the write process group are determined from the disk device 2a to the disk device 2
b. As a result, according to the present embodiment, the overhead required for the mirror disk function processing can be significantly reduced.

In the above embodiment, the access counter 8
3, the access history information tables 81, 21a, and 21b are used when the number of accesses based on a write request for a new data block exceeds a predetermined number of accesses t.
Is a predetermined condition for initializing the timer. For example, a timer interrupt generating unit for generating a timer interrupt is provided in place of the access count definition area 22 and the access counter 83, and the write executed during the timer interrupt is generated. One write processing group is formed by the processing, and the duplication control unit 42
The same effect as described above can be achieved by initializing the contents of the access history information tables 81 and 21 each time a timer interrupt occurs. However, the number of accesses t in this case is not a fixed number but a variable.

In addition to the above, initialization is performed based on a mere number of write requests regardless of the contents of the access history information, or when a write request is received from an important user process or an instruction is received from a management process. Other predetermined conditions such as time may be provided, and this is also within the scope of the present invention.

Embodiment 2 In the first embodiment, the access history information is provided corresponding to each data block. However, in the present embodiment,
It is characterized in that data blocks are grouped in arbitrary units to form a block group, and access history information is provided for each block group. In order to set this data block group, mapping tables 84, 24a, and 24b having the same contents are set in the memory 8 and each of the disk devices 2a and 2b, as shown in FIG. In the mapping tables 84 and 24, the leading logical address and the block size (the number of data blocks) of the data blocks forming the block group are set, respectively. That is, in the first embodiment, the data block and the access history information are in one-to-one correspondence, whereas in the present embodiment, it is n-to-one.
(N is a positive integer of 2 or more). Thereby, the access history information tables 81, 21
Can be reduced in size. That is, the number of records in the mapping tables 84 and 24 is the number of access history information set in the access history information tables 81 and 21. The sequential numbers “No.” provided in the mapping tables 84 and 24 correspond to the arrangement order of the access history information in the access history information tables 81 and 21. However, the operation is possible without this data.

The operation of this embodiment is almost the same as that of the first embodiment. However, when the access history information corresponding to the data block to be written is obtained, in this embodiment, the mapping on the memory 8 is performed. Although it is necessary to always refer to the table 84, the number of accesses to the access history information table 81 having a larger capacity than the mapping table 84 is reduced, so that the overhead can be further reduced. For example, 3 consecutive from logical address 0
In the case where there is a write request to a block, the number of accesses to the access history information table 81 (the number of times of execution of step 102 in FIG. 2) is required three times in the first embodiment, whereas in the first embodiment Then, it is only necessary to refer to the mapping table 84 once. Also, the number of times of setting access history information for three consecutive blocks starting from logical address 0 (step 108 in FIG. 2,
109), the overhead can be further reduced.

When the system is started, the mapping table 24 of the disk device 2 is read into the memory 8 in the same manner as the access count definition area 22 is read. And the mapping table 8
4, the access history information table 81 is created, and the access history information table 81 is copied to the access history information table 21. Since the mapping table 24 having contents different from those at the time of the previous system operation may be set, in other words, the above-mentioned value of n to n may be changed. This is because it is necessary to set the access history information tables 81 and 21 in the.

The mapping tables 84 and 24 are
It can be set arbitrarily by the system designer. For example, the setting may be based on a logical partition (for example, partition or volume) understood by the file system, or may be set based on a physical partition (for example, for each disk in a multi-volume case). Further, in the case of a Unix file system, for example, the block size may be set to 256 blocks if sfs, or to 4096 blocks if vxfs, depending on the type of file system. Further, when dealing with a relatively small file such as a text file, the block size is set to 16 blocks, and when dealing with a relatively large file such as bitmap data, the block size is set to 4096.
It may be set according to the type of data to be handled, such as a block. As a result, an access history information table can be set according to system characteristics, so that more efficient mirror disk control can be performed and overhead can be reduced. Of course, the system designer may set it freely.

Embodiment 3 In the second embodiment, the mapping tables 84 and 24
Are set in advance, but in the present embodiment, the mapping tables 84 and 24 are set by the learning function.
Is automatically set. In the present embodiment, as shown in FIG. 5, an access data size history table 43 is provided in the mirror disk control device 4, and every time a write process to the disk device 2 is performed,
As the data storage position information, the head logical address of the block group subjected to the writing process and the block size (the number of data blocks) are held as history information. This process and the process of creating the mapping table 84 based on the history information held in the access data size history table 43 are performed by the duplexing control unit 42.

Next, the operation of the mirror disk control device 4 in this embodiment when the system is operating normally will be described with reference to the flowchart shown in FIG. Note that the processing in the present embodiment is the same as the processing shown in FIG. 2 except that step 301 is added.
Details will be omitted in the following description.

In the processing up to step 301, "1" is set in the access history information corresponding to a predetermined data block in response to a write request. However, in the present embodiment, the data block and the access history information are n-to-one.
Therefore, the number of times of setting is smaller than that in the first embodiment. Then, in step 110, the data is actually written to the disk device 2, but in the present embodiment, the data is
The head logical address and the block size of the area to be written into the access data size history table 43 are recorded in the access data size history table 43 (step 301). The example of the first row of the access data size history table 43 shown in FIG. 5 means that the write processing of the block length “3” from the start address “0” has been executed. In this way, every time the writing process is performed, the access history is recorded in the access data size history table 43.

In the present embodiment, a fixed number of records can be recorded in the access data size history table 43, and when the number of histories becomes full, old records are overwritten. Of course, this can be set appropriately according to the disk capacity of the mirror disk control device 4 that stores the access data size history table 43.

At the end of the system, all the contents of the access history information table 21 stored in the disk device 2 need to be cleared as in the above-described embodiments, and the system ends. No special processing is required.
Therefore, the contents of the mapping table 84 on the memory 8 are lost.

When the system is started, the same processing (steps 121 to 125) as the processing shown in FIG. 3 is performed, but only a processing for creating the mapping table 84 is added at the end. Therefore, the processing at the time of starting the system is not particularly illustrated. FIG. 7 is a flowchart showing a process for creating the mapping table 84. Next, this process will be described.

First, the contents of the access data size history table 43 are read (step 311). Next, “0” is set as a start address and a division size (S) when the block size is large is set as an initial value of the process (step 312). The division size is set according to the logical or physical delimiter, the type of data, or the like according to the method of determining the block size of the mapping table described in the second embodiment, or the data size or access data size history immediately before. It can be set arbitrarily based on the average value of the block sizes recorded in the table 43 and the like.

Next, the set start address (at first,
It is searched whether the information of 0) is recorded in the access data size history table 43 (step 313). If it has been recorded, it is set in the mapping table 84 with its start address and the corresponding block size (steps 314, 315). If a plurality of pieces of information of the start address are recorded, the information is set with the largest block size. In addition, it may be the latest instead of the maximum.

On the other hand, if the information of the start address is not recorded in the access data size history table 43 (step 313), the minimum start address information is searched from the access data size history table 43 (step 316). ). Then, a comparison reference value obtained by adding the division size (S) to the searched start address is compared with the next smallest start address value (step 317). Is set as the start address, and the difference between the next smallest start address and the smallest start address is set as a block size in the mapping table 84 (steps 318 and 315). On the other hand, if the comparison reference value is smaller than the next smallest start address value, the smallest start address is set as the start address, and the division size (S) is set as the block size in the mapping table 84 (steps 319 and 315).

Next, the set block size is added to the set start address to calculate a start address of information to be read next (step 320). If the calculated start address is smaller than the capacity of the disk device 2, the process returns to step 313 and the above-described processing is repeated (step 321). When the calculated start address becomes equal to or larger than the capacity of the disk device 2, the setting of the mapping table 84 for the disk device 2 has been completed. The contents of the mapping table 84 created by the above processing are arranged in ascending order of the start address.

When the mapping table 84 is created,
Based on this, the access history information table 81 is created, and the access history information table 81 is copied to the access history information table 21 (step 32).
2). Further, the mapping table 84 is copied to the mapping table 24 of the disk device 2 (Step 323).

When the processing at the time of starting the system is completed, the above-described processing of FIG. 6 is repeatedly performed.

According to the present embodiment, the access history of the actually executed write processing is recorded in the access data size history table 43, and the mapping tables 24 and 84 are generated based on the access history. That is, at the time of the next system operation, a data block group is formed according to the previous disk access pattern actually performed, so that more efficient write processing to the mirror disk with less overhead can be performed.

The mapping table 8 shown in FIG.
Although the process of creating 4 is performed at the time of system startup in consideration of unexpected system down, this process may be performed at normal termination of the system. In this case, the contents of the created mapping table 84 are copied and stored in the mapping table 24 of the disk device 2. In this case, the processing is the same as that of the second embodiment.
Will be read.

Embodiment 4 In the third embodiment, the process of creating the mapping table 84 is performed at the time of system startup or normal termination of the system. However, in the third embodiment, the process of resetting the access counter 83 during normal operation of the system is performed. It is characterized in that it is performed. That is, during the normal operation of the system, the processing shown in FIG. 7 is executed immediately before or immediately after step 106 in FIG. As described above, the mapping table 8 based on the immediately preceding write process between the time when the access history information tables 81 and 21 are initialized and the time when the next
4, 24 are created, so that when there is a local access due to an access time, a user process, or the like, a more efficient write process to the mirror disk with less overhead can be performed. In a system that operates continuously, efficiency can be improved during use.

Embodiment 5 Access history information table 8 in each of the above embodiments
The access history information set to 1, 21 is the flag information indicating the presence or absence of the access based on the write request as described above. An access counter 83 and the like are provided so that the access history information set when the number of accesses to the entire disk device 2 reaches a predetermined number is collectively cleared. The present embodiment is characterized in that access history information is provided not as flag information (bit information) but as update information that can take a continuous value. FIG. 8 is a configuration diagram showing a fifth embodiment of the mirror disk control system according to the present invention. As shown in the access history information tables 81 and 21, the access history information includes integer values, respectively. A writable data storage area is used. In this embodiment, a loop counter 85 is further added to the memory 8. The access counter 83 is cleared when its value reaches the number of accesses t set in the access count definition area 82. However, the loop counter 85 is incremented by one at the timing when the access counter 83 is cleared.

Next, the operation of the mirror disk control device 4 in this embodiment when the system is operating normally will be described with reference to the flowchart shown in FIG.

When the mirror disk controller 4 receives an access request from a user process, it checks whether it is a write request or a read request (step 50).
1). If it is a read request, the disk access control unit 4
1, the corresponding data is read from the disk device 2 and passed to the user process (step 508). If it is a write request, the duplexing control unit 42
Of the memory 8 and the access history information tables 81 and 21 of the disk devices 2a and 2b.
The access history information corresponding to the data block to which a and 21b are to be written is set (step 502). This means that the value of the loop counter 85 is written whatever the current value of the access history information is. That is, the value of the loop counter 85 can be regarded as identification information assigned to each write processing group.

If the value of the access counter 83 has reached the number of accesses t set in the access count definition area 82 (step 503), 1 is added to the value of the loop counter 85 (step 504). Then, after resetting the access counter 83 (step 505), 1 is added to the value of the access counter 83 (step 506). On the other hand, if the value of the access counter 83 has not reached the access count t, only the addition of the access counter 83 is performed (steps 503 and 506).
Then, the write data is written to the disk device 2 (step 507).

As described above, the present embodiment operates, but differs from the first embodiment in the following points.

In the first embodiment, the access history information corresponding to a data block for which a write request has been made for the first time is set to “1”. Actually, in step 110, data is written in the data block of each of the disk devices 2a and 2b. However, in order to reduce the overhead, the access counter 83 reaches the predetermined number of accesses t and the access history information table 81, Until the initialization is performed, information that the data may not be written in consideration of an unexpected system down is held. Then, after the system down, only the data block corresponding to the access history information in which "1" is set is copied from the primary to the secondary disk device 2.

In the present embodiment, in order to obtain the same effect, the access history information uses the update information instead of the flag information. For example, assuming that the number of accesses t is “100”, the access history information “71” in FIG. 8 indicates that the writing process has been performed between the 7001th and 7100th times. If the value of the loop counter 85 is “71”, it means that the data was written in the latest write processing (the value of the access counter 83 is between “1” and “100”). Also, the access history information “60” means that the writing process has been performed between the 5901th and 6000th times. If the value of the loop counter 85 is “71”, it means that the data block has not been subjected to the write processing for about 1000 times simply calculated. That is, it can be said that the value set in the access history information is identification information indicating which write processing group belongs.

Then, the value of the loop counter 85 becomes "7".
The data block corresponding to the access history information having a value of 70 or less when the value is 1 "is normally written to all the disk devices 2 and the equivalence is guaranteed. Therefore, according to the present embodiment. For example, after an unexpected system failure, only the access history information having the same value as the value of the loop counter 85 needs to be copied from the primary disk to the secondary disk device 2. However, as is clear from FIG. The disk device 2 has a loop counter 8
The value of 5 is not stored. Of course, the value may be stored, but in the present embodiment, instead of storing the value, both access history information tables 21a, 21a,
21b is searched for the maximum value, and only the data block corresponding to the access history information for which the maximum value has been set is copied from the primary disk to the secondary disk unit 2, thereby obtaining all disk units. 2 are matched.

As is clear from the above, when the system is terminated in this embodiment, all the contents of the access history information table 21 stored in the disk device 2 are cleared as in the first embodiment. And end.

The operation at system startup is basically the same as in the first embodiment, as shown in FIG. That is, the contents of the access history information table 81, the access count definition area 82, the access counter 83, and the loop counter 85 of the memory 8 are initialized (step 52).
1) The contents of the access history information table 21 and the access count definition area 22 stored in the disk device 2 are read into the corresponding tables 81 and area 82 on the memory 8 (step 522). However, the loop counter 85
Is set to "1" as an initial value. Then, the maximum value is searched from the access history information included in the access history information table 81 (step 523), and only the data block corresponding to the access history information having the maximum value is changed from the primary to the secondary disk device 2. (Step 524). Then, the contents of the memory 8 and the access history information tables 81 and 21 of the disk device 2 are initialized (step 525), and the processing shifts to the normal processing shown in FIG.

In the process at the time of starting the system, the copying process to the disk device 2 is always executed. However, the value of the loop counter 85 is also stored in the disk device 2 and the value is compared. In this case, it is possible to perform the copy process only when necessary.

As described above, according to the present embodiment, the same effect as in the first embodiment can be obtained even when the update information in which an integer value can be written is used instead of the flag information as the access history information. In addition, when the access counter 83 reaches a predetermined number of accesses t during normal operation, it is not necessary to perform only the addition processing of the loop counter 85 and initialize the contents of the access history information tables 81 and 21. Further, overhead in mirror disk control can be reduced.

In this embodiment, since the flag information is not used for the access history information, there is a possibility that the update information and the loop counter 85 may overflow. It can be prevented by the design and operation of the access history information tables 81 and 21 such as clearing.

Although the present embodiment has been described in comparison with the first embodiment, the present embodiment can be applied to the case where the data block and the access history information have an n-to-1 relationship as in the second embodiment. is there.

[0086]

According to the present invention, a series of data write processing to a disk device is divided into a plurality of groups, and the equivalence of the disk device is guaranteed for each write processing group. That is, by copying only the data to be processed in all the write processes constituting the write process group including the write process that was being processed when the system failure occurred from one disk device to another disk device. Since the contents of each disk device can be matched, the time required for recovery when the system goes down can be greatly reduced, and the overhead when mirror disk control is performed can be reduced. In addition, since information indicating whether or not data has been written need only be held for the latest write processing group, it is possible to reduce overhead when mirror disk control is performed during normal operation.

After the write request is received and the access history information corresponding to the data block is set to that effect, the data is written to the data block and even if the contents of all the disk devices become equivalent. At this point, the setting of the access history information is not canceled, and the access history information is collectively initialized when a predetermined condition is satisfied. That is, the access history information indicating the presence / absence of data writing is retained as long as a predetermined condition is satisfied, so that it is not necessary to set the access history information at the time of writing processing to the same data block thereafter. Therefore, overhead can be reduced. In addition, since the access history information is collectively initialized when a predetermined condition is satisfied, it is possible to reduce the overhead for the process of canceling the setting of the access history information. When a system failure occurs, only data blocks corresponding to the set access history information need be copied from one disk device to another disk device. Not only can be assured of equality, but also it can be recovered quickly.

The same effect as described above can be obtained even if the access history information is provided not as flag information indicating presence / absence of access based on a write request but as update information that can take a continuous value. In particular, according to the present invention, it is necessary to perform an initialization process of the access history information table performed when a predetermined condition is satisfied during normal operation, as compared with the case where an access history information table holding flag information is used. Therefore, overhead in mirror disk control can be further reduced.

Further, if a mapping table is provided so that data blocks and access history information correspond to n: 1 (n is a positive integer of 2 or more), the size of the access history information table can be reduced. it can. The mapping table defines the storage location identification information and the size of each block to be written, and separates each block with a logical partition understood by the file system, a physical partition of the disk device, and a file. By setting based on the logical access unit unique to the system or the type of data to be handled, a more effective access history information table can be created, and as a result, overhead in mirror disk control can be reduced.

Further, since it is possible to obtain the storage position information and the size in accordance with the previous disk access pattern actually performed, the mapping table can be created more effectively.

[Brief description of the drawings]

FIG. 1 shows a first embodiment of a mirror disk control method according to the present invention.
FIG. 2 is a configuration diagram showing an embodiment.

FIG. 2 is a flowchart showing an operation of the mirror disk control device during a normal operation of the system in the first embodiment.

FIG. 3 is a flowchart showing an operation of the mirror disk control device at the time of starting the system in the first embodiment.

FIG. 4 shows a second embodiment of the mirror disk control method according to the present invention.
FIG. 2 is a configuration diagram showing an embodiment.

FIG. 5 shows a third embodiment of the mirror disk control method according to the present invention.
FIG. 2 is a configuration diagram showing an embodiment.

FIG. 6 is a flowchart showing an operation of the mirror disk control device during a normal operation of the system in the third embodiment.

FIG. 7 is a flowchart illustrating a process of creating a mapping table in the third embodiment.

FIG. 8 shows a fifth embodiment of the mirror disk control method according to the present invention.
FIG. 2 is a configuration diagram showing an embodiment.

FIG. 9 is a flowchart showing an operation of the mirror disk control device during a normal operation of the system according to the fifth embodiment.

FIG. 10 is a flowchart showing the operation of the mirror disk control device when the system is started up in the fifth embodiment.

FIG. 11 is a configuration diagram of a conventional mirror disk control device.

FIG. 12 is a configuration diagram of a conventional duplicate volume equivalence determination method.

[Explanation of symbols]

2a, 2b disk device, 4 mirror disk control device, 6 computer, 8 memory, 21a, 21b,
81 access history information table, 22a, 22b, 8
2 Access count definition area, 24a, 24b, 84
Mapping table, 41 disk access control unit, 42 duplex control unit, 43 access data size history table, 83 access counter, 85 loop counter.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-2-166509 (JP, A) JP-A-58-175065 (JP, A) JP-A-2-136943 (JP, A) JP-A-6-166943 67810 (JP, A) JP-A-4-107725 (JP, A) JP-A-1-266640 (JP, A) JP-A-63-59644 (JP, A) JP-A-3-93942 (JP, U) 6-75041 (JP, U) Mike Loukides, UNIX System Tuning, ASCII, Japan, First Edition, 202-209 (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 3 / 06 G06F 12/00 531 G06F 12/16 310

Claims (15)

(57) [Claims] 1. A memory and the mirror disk to have a mirror disk control function to match constantly all the content the disk device and a plurality of disk drives constituting the mirror disk, by writing the same data simultaneously to the each disk device In the mirror disk control method having control means, a series of data write processing to the disk device is divided into a plurality of write processing groups based on predetermined conditions, and a write processing request is issued during normal system operation. If
Write the data resulting from the write processing to each of the disk devices, and write the storage position information of the data to both the memory and each of the disk devices only when the storage position information of the data has not yet been written to the memory; At least during the processing of the write processing group is held, and when all of the write processing included in the write processing group is completed, the storage position information held in the memory and the disk devices is initialized, At the time of system recovery, only the data corresponding to the storage position information held in each of the disk devices is copied from one of the disk devices to another of the disk devices so that the contents of the respective disk devices are matched. And a mirror disk control method. 2. A memory, a plurality of disk devices that constitute a mirror disk and write data in units of one or a plurality of data blocks, and write all the disk devices simultaneously by writing the same data to each of the disk devices. a mirror disk control unit that have a mirror disk control function to match the content at all times, is composed of the same contents to each of the memory and said disk devices, access indicating whether or not data write to the data block included in the disk device An access history information table including history information, and the memory and each disk corresponding to the data block to be written only when access history information in the memory corresponding to the data block to which data is written is not set to that effect. To that effect in the access history information of the device Writing presence / absence determining means for performing setting; writing processing group forming means for dividing a series of data writing processing to the disk device into a plurality of writing processing groups based on predetermined conditions; Table initializing means for simultaneously initializing the contents of the memory and the access history information table on the disk device at the time of completion of the write process, and the equivalence of the disk device for each write process group. A mirror disk control method characterized by guaranteeing. 3. The mirror disk control means refers to access history information in the access history information table of the disk device at the time of restoration from a system failure, and sets a data write target when the mirror disk control means has been set. 3. The mirror disk control method according to claim 2, wherein the content of each disk device is matched by copying only the data block corresponding to the access history information from the disk device to the other disk device. 4. The write processing group forming means includes: an access count definition area for setting a maximum set number of access history information to the access history information table in advance; and writing data after the access history information table is initialized. An access counter that counts the number of times a write process has been issued to a data block for which access has not been performed, and the access counter has been set to a maximum access history information set number since the access history information table was initialized. Forming a single write processing group by the write processing executed until the condition that the number of access counters reaches the maximum access history information set number. The contents of a history information table are initialized. Mirror disk control method described. 5. The write processing group forming means has a timer interrupt generation unit for generating a timer interrupt,
The process of writing a series of data to the disk device is divided into a plurality of write process groups by dividing time by a timer interrupt, and the table initializing unit initializes the contents of the access history information table each time a timer interrupt occurs. The mirror disk control method according to claim 2, wherein the mirror disk control method is used. 6. A memory, a plurality of disk devices forming a mirror disk, and writing data in units of one or a plurality of data blocks, and simultaneously writing the same data to each of the disk devices so that all of the disk devices a mirror disk control unit that have a mirror disk control function to match the content at all times, is composed of the same contents to each of the memory and said disk devices, access history indicating a data writing time to the data blocks included in said disk device An access history information table including information, and a process of writing a series of data to each of the disk devices is divided into a plurality of write processing groups based on a predetermined condition, and identification information of the write processing group during the write processing is held. A write processing group forming means, and a data Write-time identification means for setting the identification information of the write processing group in the access history information corresponding to the lock, and assuring equivalence of the disk devices in units of the write processing group. Mirror disk control method. 7. The mirror disk control means searches for the latest identification information of the write processing group from the access history information tables of all the disk devices at the time of restoration from a system failure, and searches the latest write processing group. The content of each disk device is matched by copying only the data block corresponding to the access history information in which the identification information of the processing group is set from the disk device to the other disk device. Item 6
The mirror disk control method described. 8. A mapping configured to have the same contents in each of the memory and each of the disk devices and defining storage position identification information and a size of each continuous data block to be written when the block is one block. By setting access history information corresponding to each of the blocks in the access history information table on the memory and the disk device, a data block and access history information can be stored in an n-to-one (n 7. The mirror disk control method according to claim 2, wherein the mirror disk control method corresponds to a natural number. 9. The mirror disk control method according to claim 8, wherein the block division is set based on a logical division understood by a file system. 10. The mirror disk control method according to claim 8, wherein the block division is set based on a physical division of the disk device. 11. The mirror disk control method according to claim 8, wherein the block delimiter is set based on a logical access unit unique to a file system. 12. The mirror disk control method according to claim 8, wherein, if the type of data to be handled can be recognized in advance, the block division is set based on the type of the data. 13. The mirror disk control method according to claim 8, further comprising mapping table automatic setting means for automatically setting the contents of said mapping table in accordance with the status of writing processing to said disk device. 14. An automatic mapping table setting unit, wherein: an access history holding unit for holding, as history information, a storage position information and a size of data in each writing process to the disk device; 14. The mirror disk control method according to claim 13, further comprising: a table creating unit that creates the mapping table based on the history information. 15. The mirror disk control system according to claim 13, wherein said mapping table automatic setting means operates at a timing when said access history information table is initialized.