JP3254766B2 - Multiple writing method of the same data, data reading method and data recovery method, and control device therefor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001] The present invention relates to a 2
The present invention relates to a multiplex writing method of the same data in a secondary storage device, a data reading method, a data recovery method, and a control device therefor. In particular, the present invention relates to a method and apparatus for performing multiplex writing control of the same data on a predetermined number of storage units of a plurality of storage units, such as disk drives, of a secondary storage device.

[0002]

2. Description of the Related Art Up to now, performance improvement of secondary storage devices has been mainly aimed at increasing storage device capacity and data transfer speed. However, with the improvement of the processing speed of the computer, the improvement of the input / output throughput for the secondary storage device has become an issue. One solution is to write the same data to all n (n is an integer of 2 or more) disk drives, issue a data output command to all disk drives when reading data, and A method of reading the data from a disk drive that can process the data output command has been proposed.

In this regard, Dina Bi of the University of Illinois
tton et al .: Disk Shadowing (Dina Bitt)
on, et al, “Disk Shadowing”, Proceedings of the
14th International Conference on Very Large Data B
ase, 1988, pp. 331-338) and JP-A-1-17101.
No. 5 (JP-A-1-17015).

As another method for improving the input / output throughput, there is known a technique in which a plurality of disk drives are independently driven to divide one piece of data and write the data to the plurality of disk drives. For example, RAID (Redundant Ar
rays of Inexpensive Disks).

A case of redundant array of inexpensive disks such as D. Patterson of the school (David A. Patterson, et al,
“A Case of Redundant Array of Inexpensive Disks
(RAID) ", ACMSIGMOD Conference, 1988).

[0006] Here, a technique for improving reliability using parity, ECC, and the like is shown for the problem of a decrease in reliability due to data being divided and stored in a plurality of disk drives. That is, correction data is created for each data using parity, ECC, and the like.
A technique is disclosed in which, when a failure occurs in a certain disk drive, data in the failed disk drive is recovered from correction data and data remaining in a normal disk drive. This technique is disclosed in, for example, Japanese Patent Application Laid-Open No. 2-236714 (JP-A-2-236714).

[0007]

In the former technique typified by the paper "Disk Shadowing", the waiting state of an input / output command in each disk drive is considered when reading / writing the same data to a plurality of disk drives. Not. Therefore, when writing the same data to a plurality of disk drives, if the plurality of disk drives includes at least one disk drive that already has a large number of input / output instructions, the operation of writing the same data as a whole is I will be late.

When the same data is read from a plurality of disk drives, a read command is issued to all the disk drives storing the data. Unnecessary input / output instructions are added to a disk drive that already has many input / output instructions.

On the other hand, even in the latter technique typified by the article "RAID", the waiting state of an input / output command in each disk drive during data read / write operation to a plurality of disk drives is not considered. Also,
If a disk drive fails, its data recovery requires special correction data.

A main object of the present invention is to provide a method for multiplexing and writing the same data in which the input / output throughput of a secondary storage device having a plurality of storage units is improved, and a control device therefor.

Another object of the present invention is to provide a method for multiplex writing of the same data in a secondary storage device capable of automatically generating history data, and a control device therefor.

Another object of the present invention is to provide a secondary storage device having a plurality of storage units, a method of multiplex writing the same data and a method of reading the data, which have little effect on other input / output instructions, and their control. It is to provide a device.

Still another object of the present invention is to provide a data recovery method and a control device therefor which can easily recover data when a failure occurs in a certain storage unit in a secondary storage device having a plurality of storage units. Is to provide.

Still another object of the present invention is to provide a secondary storage device having a plurality of storage units, a method for writing the same data and a method for reading the data, which are flexible with respect to the addition of unused storage units. It is to provide a control device for that.

[0015]

According to a preferred embodiment of the present invention, in a secondary storage device having a plurality of storage units, a storage unit having a low degree of waiting for processing of an input / output instruction is designated by the same storage unit. This is achieved by selecting data as a group of storage units to be multiply written.

Further, according to a preferred embodiment of the present invention, in the secondary storage device having a plurality of storage units, the object of the present invention is to consider the data free space of each storage unit and the degree of waiting for processing of an input / output instruction. This is achieved by selecting a storage unit to which the same data is to be written multiple times.

[0017]

In the multiplex storage device of the present invention, as shown in FIG. 15, one data is stored in a plurality of storage units (disk devices # 1, # 2).
# 3) When processing output instructions to store them,
Since data is called from the storage unit that can be used first among them and the data #n is used, output processing performance can be improved. Further, the controller, when processing a plurality of input / output commands, if a storage unit for storing input data cannot be used for processing another input / output command, the other storage unit ( Disk devices # 2 and #
Since the data #n is written in the empty area of 4), it is possible to achieve an improvement in input processing performance. By associating the address specified by the input / output instruction with the number and address of the storage unit where the data is actually stored, and automatically managing them in the data address management table,
The user can perform optimal input / output processing without concern for the position of the data in the storage unit. Also, when performing a recovery from a failure in a storage unit, data to be read is dispersed in each storage unit, so that only one storage unit is not occupied in a concentrated manner. However, the writing is not concentrated on only one storage unit, and the processing performance of the input / output request can be prevented from deteriorating. Also, if a failure of the storage unit occurs and data cannot be read or written,
To compensate for this by reading out the other multiplexed data on the storage unit, and if all the multiplexed data becomes unusable, increase the reliability by substituting the stored history data. Can be done.

[0018]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows the configuration of an embodiment of a secondary storage device according to the present invention. In the present embodiment, a secondary storage device having a magnetic disk drive as a plurality of storage units will be described as an example. In FIG. 1, the secondary storage device includes a control unit 2 for controlling a plurality of storage units as a whole, magnetic disk drives 16-1 to 16-n, and a disk control device 15 for controlling the same.

In FIG. 1, reference numeral 1 denotes a CPU which is a host device.
In response to the user's instruction, the CPU 1 issues a command to the secondary storage device 101 and receives the processing result. The control unit 2 manages input / output commands from the CPU 1, manages stored files (data), and the like.
A command issued from the CPU 1 passes through the external interface path 3 and passes through an interface port (hereinafter referred to as I
F prot). Although FIG. 1 shows the configuration of two systems, only the operation of one system will be described below for the sake of simplicity. The other system performs the same operation. The microprocessor (MP) 11-1 includes:
Check whether there is a path available in the external interface path 3 and use the available external interface path 3
If there is, the MP 11-1 switches the channel path switch (hereinafter, CPS) 5 to perform command reception processing.
If the command is not accepted, MP11-1
Sends a non-acceptance response to the CPU 1.

The accepted command is sent to the channel port circuit (CHL prot) 6-1 to convert a channel protocol into a protocol in the control unit 2. The converted command passes through the internal command path 10-1 to the command buffer (CBUF) 7-1, and the data passes through the internal data path 9-1 to the data buffer (DBU).
F) Temporarily placed in 8-1, waiting for the next process.

The management table (MT) 12 includes a data address management table 50 for managing addresses of data stored in each disk drive,
An empty area management table 55 for managing an empty area of each disk drive and a drive management table 60 for managing the state of each disk drive are provided.

The data processed appropriately by the MP 11-1 is converted by the drive port 13-1 so as to conform to the protocol of the drive data path 19. The command is sent to the drive processors 17-1 to 17-n through the drive command path 18. MP11
"-1" checks whether there is an available path in the drive data path 19 in response to the command. If there is an available drive data path 19, the MP 11-1 switches the drive path switch (DPS) 14 to send commands and data to the drive processors 17-1 to 17-n.
If it is not accepted, the fact that it is not accepted is registered in the drive management table 60.

The arbitration processor 15-1 controls the internal command path 10. Each of the drive processors 17-1 to 17-n is connected to each disk drive 16.
-1 to 16-n, head movement, rotation wait, failure detection report, each disk drive 16-1 to 16
−n and controls data transfer and the like of the control unit 2. The number of drive processors 17-1 to 17-n and the number of disk drives 16-1 to 16-n are generally arbitrary.

In this embodiment, the same data sent from the CPU 1 is stored in a plurality of disk drives by multiplex writing.

FIG. 2 shows MP11-1 or MP11-
2 shows the processing executed and the contents of the MT 12. As is clear from FIG. 1, this embodiment includes MP11-1 and M
Although there are two systems P11-2, only ten systems will be described for convenience of explanation. The MP 11-1 includes an input / output management unit 70 that performs queuing processing of commands and data, a command processing management unit 75 that performs command address conversion, management of empty areas, and the like, a multiplex writing / reading of data, and a failure recovery processing. A multiple write / read failure management unit 80 for performing the operation, and a disk drive management unit 85 for managing the degree of use and state of the disk drive are prepared. The MT 12 is provided with a data address management table 50, an empty area management table 55, and a drive management table 60 as locations for storing information necessary for these processes. CBUF7-1 and DBUF8-1 are prepared as buffers used when executing the input / output management unit 70. Also, the drive processors 17-1 to 17-1
7-n has a disk drive control unit 90 that performs command processing, seek, search, data reading, and writing in the disk drives 16-1 to 16-n.

FIG. 3 shows the data address management table 5
0 shows the structure.

As shown in FIG. 3, the data address management table 50 includes a plurality of areas described below. 100
Reference numerals -1 and 100-2 are areas in which a user number for identifying a user and an assigned storage capacity available to the user are registered, and are defined for each user. In FIG. 3, the user number 1 is assigned a storage capacity of M1 bytes, and the user number 2 is assigned a storage capacity of M2 bytes.

105-1 and 105-2 are logical addresses designated by the CPU 1 as data storage locations,
This is an area for registering the number of history records (the number of pieces of data before updating the data has). In the example of FIG. 3, the logical address 1 has a history holding number of 3 and the logical address 2 has a history holding number of 4
have.

120-1, 120-2 and 120-3
Is an area for registering a history number indicating the order in which data has been updated and the multiplicity of data (the number of disk drives storing the same data). In this embodiment, the newer the data, the lower the history number. In FIG. 3, the multiplicity 3 is registered in the history number 1, the multiplicity 3 is registered in the history number 2, and the multiplicity 4 is registered in the history number 3.

130-1, 130-2 and 130-3
Is an area for registering a disk drive number for identifying a plurality of disk drives storing the data, and a physical address and a data capacity of an area where data is actually stored in the disk drive. In the example of FIG.
One of the triple-written data is stored at the physical address A1 of the disk drive # 1, and the other one is
Stored in the physical address A2 of the disk drive # 2,
The other one is the physical address A of the disk drive # 4
3 is stored. Since the data is the same, the data capacities of the areas 130-1, 130-2, and 130-3 are equal and D1 bytes.

A plurality of areas 105 are connected to the area 100, a plurality of areas 120 are connected to each area 105, and a plurality of areas 130 are connected to the area 120. The entire data address management table 50 is a tree. It has a structure. In this embodiment, the entire data address management table 50 has a tree structure, but may have another list structure or table structure.

As described above, one piece of data is multiplexed and, as indicated by the data address management table 50, retains past history data. Therefore, the data capacity of each user can be calculated by summing the data capacity multiplied by the multiplicity and the number of history records.
For example, N represents the number of data managed by the user number j, and C represents the data capacity, multiplicity, and history retention number of each data.
i, Mi, Hi, the data capacity U of the user number j
Cj is expressed by the following equation (1).

[0034]

(Equation 1) Further, assuming that the number of users is L, the data capacity UC used by all users is represented by the following equation (2).

[0035]

(Equation 2) Therefore, as a whole of a plurality of storage units, at least UC
Data capacity is required.

In the preferred embodiment, which can improve the performance and reliability of reading and writing data, all multiplexed data must be written to different disk devices. Therefore, it is desired that the number of disk drives is larger than the multiplicity. However, if it is only necessary to improve the performance at the time of reading, it can be realized by writing to different positions in the same disk drive. In that case, the number of disk drives may be smaller than the multiplicity.

Next, referring to FIG.
Will be described.

In FIG. 4, reference numerals 200-1 to 200-n denote disk drive numbers, and each disk drive 16-
1 to 16 to n. 205-1 to 20
Reference numeral 5-7 denotes an area for registering a physical address indicating an empty area where data of each of the disk drives 16-1 to 16-n is not stored, and a data capacity that can be stored in the empty area. In FIG. 3, the free space 20 is stored in the disk drive number # 1.
5-1 and 205-2 exist, and the disk drive number #
3 has free areas 205-3 and 205-4, the free area 205-5 is provided for the disk drive number # 4, and the free areas 505-6 and 205 are provided for the disk drive number #n.
-7 exists. Note that there is no free space in the disk drive # 2. Empty area management table 55
The whole has a list structure. In this embodiment, the entire empty area management table 55 has a list structure, but may have a tree structure or a table structure. The capacity of the empty area of the plurality of storage units is determined by each disk drive 16-1.
It can be calculated by adding the data capacity of the area 205 for every .about.16-n and taking the total sum for each disk drive. EC of the free space capacity of each disk drive
Assuming that k is the number of disk drives and D is the number of disk drives, the free space capacity EC required for the entire plurality of storage units can be expressed by the following equation (3).

[0039]

(Equation 3) If the capacity of the data in the empty area is EC, the storage capacity AC required for the entire plurality of storage units is the sum of the data capacity UC and is expressed by the following equation (4).

[0040]

AC = EC + UC (Equation 4) The plurality of storage units in this embodiment improve input / output processing performance by writing input data to an empty area of another disk drive that can be used most quickly. For this reason,
In order to bring out the performance sufficiently, it is necessary to secure a sufficient capacity ECk of the empty area of each disk drive. If this free space is insufficient or not present, data will be written to the disk drive that has already received other I / O commands, but even at this time, the I / O processing performance will be lower than that of a normal multiple storage unit. It does not decrease. If there is no empty area, it is also possible to reduce a part or the multiplicity of the history data shown in FIG. 3 and turn the resulting area into an empty area.

Actually, in addition to these, a save area of memory such as the data address management table 50, the empty area management table 55, and the empty area management table 60, a confirmation area of the data recovered from the failure, and the like are required.

FIG. 5 shows a disk device management table 60.
The structure of is shown.

In FIG. 5, the disk device management table 60 includes a disk drive number registration area 401 for identifying each of the disk drives 16-1 to 16-n, and a processing wait for each of the disk drives 16-1 to 16-n. An area 402 for registering the number of input / output instructions is included. In FIG. 5, the number of input / output requests is represented by a three-digit binary number in the input / output instruction number registration area 402. If a failure occurs and there is a disk drive that cannot be used in the registration area 402 for the number of input / output instructions to be processed, a flag indicating that the disk drive cannot be used can be set in the area 402 corresponding to the disk drive. In FIG. 5, the maximum value “111” represented by a three-digit binary number is used as a flag indicating that it cannot be used. That is, the disk drive number # 5 in FIG. 5 cannot be used. FIG. 5 shows the disk drive number # 2
And the disk drives 16-2 to 16-n of #n have no input / output commands waiting for processing, and the disk drive 16-1 of disk drive number # 1 has six input / output commands in the processing waiting state. It shows that there is.
Also, disk drive 1 of disk drive number # 3
6-3 indicates that the number of input / output instructions to be processed is three.

Next, the input / output management unit 70, the CBUF 7-1 (or 7-2), and the DB will be described with reference to the flowchart of FIG.
The processing procedure of the UF 8-1 (or 8-2) will be described.

When the input / output management unit 70 starts the input / output processing, first, it checks the state of the input / output command and data held in the CBUF 7-1 and DBUF 8-1 (step 50).
5). Here, the number of queues in the CBUF 7-1 and the DBUF 8-1 is counted, and it is checked whether the number of queues exceeds a preset threshold. CBUF7-1, DBUF8-
If the number of queues in 1 is equal to or smaller than the threshold value, it is checked in the next step whether there is an input / output command sent from the CPU 1 (step 510). If a new input / output command has been received from the CPU 1, the input of the input / output command is accepted (step 515). Next, the input / output command accepted in step 515 is stored in a queue in CBUF7-1 and DBUF8-1 (step 520). Next, CBUF7
-1, sort the queues in the DBUF 8-1 according to the priority order attached to the input / output instruction (step 5).
25). Then, the process returns to step 505.

If it is determined in step 505 that the queue is equal to or larger than the threshold value, the command processing management unit 75 checks whether the input / output command can be processed (step 530). The command processing management unit 75 is performing other input / output processing,
If the process is not possible, the process returns to step 505. If it is determined that the command processing management unit 75 can perform new input / output processing, the number of input / output instructions that can be processed is increased.
The input / output command is sent to the processing management unit 75 in the sorted order (step 535). Thereafter, the process returns to step 505 to continue the input / output instruction. If the input / output processing is not continued, the input / output instruction processing is terminated (step 545).

As can be seen from FIG. 6, in this embodiment, CBUF 7-1 (or 7-2), DBUF 8-1
If the queue in (or 8-2) is equal to or smaller than the threshold value, the input / output management unit 70 sends the command processing management unit 75
The input of the input / output instruction from the CPU 1 is performed with priority over the processing of transferring the input / output instruction held in the BUF 7-1 (or 7-2) and the DBUF 8-1 (or 8-2). .

Next, regarding the processing in the command processing management unit 75, each of the input command processing and the output command processing will be described.

FIG. 7 is a flow chart showing the output command processing (data reading) procedure of the command processing management section 75.

The output command from the CPU 1 is transmitted to the input / output
When it is sent from 0, the command processing management unit 75 starts output command processing (data read from the secondary storage device 101 to the CPU 1) (step 600). The data address management table 50 is searched using the logical address of the output instruction specified by the CPU 1 and requested by the instruction.
A search is made for all disk drive numbers equal to the number of multiplicity of data (step 605). In step 605, first, the user number is confirmed, a logical address managed by the user number is searched from the area 105, and the storage destination of the data having the history number which is usually the youngest among the data corresponding to the logical address is checked. Is read from the area 130. (For example, the disk drive numbers # 1, # 2, and # 4 are read.) Next, the disk drive management table 60
Disk drive 1 not performing input / output processing
6, the area 40 of the disk drive management table 60
All disk drive numbers of disk drives without a processing wait instruction in which 2 is 0, ie, "000" in binary, are searched (step 610). Steps 605 and 6
10 can perform parallel processing in addition to performing the processing in this order. Next, among the disk drives searched in steps 605 and 610, a disk drive having the requested data and a disk drive having no processing wait instruction is selected (step 615). In the next step 620, it is confirmed whether or not the number of selected disk drives is equal to or larger than an integer of a predetermined threshold value l (l <multiplicity m). In order to enable the present invention, the multiplicity m must be 3 or more, and the total number of drives must be m or more.

If the number of disk drives selected in step 615 is equal to or smaller than the threshold value, the process waits for processing in the area 402 of the disk drive management table 60 among the disk drives having data intended for the shortage. An instruction with a small number of input / output instructions is selected until the threshold value 1 is reached (step 625). Next, the logical address of the output instruction of the CPU 1 is converted into the physical address of the selected disk drive using the data address management table (step 630). At this time, output commands to be issued to the drive processor 17 are generated by the number of selected disk drives.

Next, the output command generated in step 630 is sent to the disk drive management unit 85 (step 63).
5). When a connection report with the disk drive 16 that has sent the output command is received from the disk drive management unit 85 (step 640), the disk drive is sent to another disk drive except the first connected disk drive (for example, 16-2). Is canceled, and the process is continued only for the first connected disk drive (step 645).

In the next step 650, data reading from the disk drive 16 is completed, and processing of other commands is performed until the end of output command processing is reported from the disk drive management unit 85 (steps 650 and 655).
If the end report has been received from the disk drive management unit 85, the status of the processing of the output command is checked and recorded (step 660), and the end of the output command is reported to the CPU 1 (step 665). When the above processing is completed, the command processing management unit 75 waits for the next input / output command sent from the input / output management unit 70 (step 670).

FIG. 8 is a flowchart of the input command processing (multiplexed data writing from the CPU 1 to the secondary storage device 101) in the command processing management section 75.

The input command from the CPU 1 is sent to the input / output
When sent from 0, the command processing management section 75 starts input command processing (step 700). First, an input / output command is sent to the multiple write failure management unit 80, and the multiplicity and the number of history records are received from the multiple write failure management unit 80 (step 705). Next, the data address management table 50
A disk drive corresponding to the logical address of the input data designated by the CPU 1 and waiting for data satisfying the oldest condition is searched for (step 710).
Subsequently, the area 40 of the disk drive management table 60
A search is made for an empty disk drive in which the number of waiting input / output instructions recorded in 2 is equal to or less than a certain number (step 715). Further, the free area of each disk drive is secured by the size of the data from the free area management table 55 (step 7).
20). Steps 710, 715 and 720
Can be performed in this order, and can also be performed in parallel. Step 710 is necessary for multiple writing in the secondary storage device 101 having a small free space, but may be used only for erasing the oldest history data in the secondary storage device having a sufficient free space. is there.

Next, from the empty disk drives 16 searched in steps 715 and 720, the disk drive detected in both steps is selected (step 725).
It is confirmed whether or not the number of the selected disk drives can be ensured to satisfy the multiplicity received from the multiple write failure management unit 80 (step 730).

If the disk drive selected in step 725 alone is not enough, a drive whose number of input / output instructions exceeds a certain number among the empty disk drives searched in step 715 is empty. A certain disk drive is selected (step 735), and further, steps 725 and 735 are selected, and it is confirmed whether or not the number of disk drives can be secured by the multiplicity received from the multiple write failure management unit 80 (step 74).
0). If there are still not enough disk drives, the drive having data that satisfies the conditions found in (Step 710) is selected (Step 74).
5).

Thus, step 730 or 7
After satisfying the condition 40 or securing the disk drive 16 at step 745, the logical address of the input command is converted to the physical address of the selected disk drive by using the data address management table 50 (step 75).
0). At this time, the input commands are generated by the number of the selected disk drives and issued to the drive processor 17 corresponding to the selected disk drive. The input command generated in step 750 is sent to the disk drive management unit 85 in step 755 (step 75).
5) Processing of other input / output commands is performed until a completion report of the input command processing is received from the disk drive management unit 85 (steps 760 and 765).

When a completion report is received from the disk drive management unit, the physical address of the history older than the predetermined number of history records received from the multiple write failure management unit 80 is used as the old physical address using the result of step 710. (Step 770) The old address of the data is read from the data address management table 50, the area where the old history data is stored is registered as the empty area in the empty area management table 55, and the empty area where the new data is stored is registered. If there is an area, the area is deleted from the empty area management table 55 (step 775). Next, a new physical address of the data is read out and registered in the data address management table 50 (step 780). Next, the processing status of the input command is checked and recorded (step 785), and a completion report is sent to the CPU 1 (step 790). When the above processing is completed, the command processing management unit 75 waits for the next input / output command sent from the input / output management unit 70 as in the case of the output command (step 795).

FIG. 9 is a flowchart of the input / output command processing in the disk drive management unit 85.

First, the disk drive management unit 85 sends a command processing management unit 75 to each drive processor 17.
Processing is started by receiving input / output instructions for -1 to 17-n (step 800). Upon receiving the input / output command from the command processing management unit 75, the disk drive management unit 85
0 is retrieved, and the number of input / output instructions waiting for processing of the disk device (area 402) used by the input / output instructions is increased by one (step 805). The input / output command received from the command processing management unit 75 is sent to the disk drive control unit 90 in the drive processor 17 of the disk drive 16 to be used next (step 810). Thereafter, it is confirmed whether or not a report of the end of the input / output processing has been received from the drive processor 17 (step 815).

If the end report has not been received, the status of each disk drive is monitored and the disk drive 16-
1 to 16-n, drive processors 17-1 to 17-n
For checking whether a failure has occurred (820,
825). If no failure has occurred, step 81
Return to 5. Here, the disk drives 16-1 to 16-
n, and when a failure occurs in any of the drive processors 17-1 to 17-n, for the failed disk drive, the area (area 402 ) Is set to an unusable flag and the use of the corresponding disk drive is prohibited (step 830). Then, the failure is reported to the multiple write failure management unit 80 (step 835). In step 815, if a completion report of the input / output processing has been received, the processing status is checked, and the number of waiting input / output of the disk drive for which the input / output processing has been completed is reduced by one in the disk drive management table (step 84).
5). Finally, the end of the input / output command processing is reported to the command processing management unit 75, and the processing is ended (step 847).
When the above processing is completed, the disk drive management unit 85
Waits for the next input / output command from the command processing management unit 75 (step 850).

FIG. 10 is a flowchart showing a processing procedure of the ordinary multiple write fault management unit 80 in the input / output command processing.

When an input / output command is sent from the input / output management unit 70, the multiple write fault management unit 80 receives the input / output command from the CPU 1 or stores the multiplicity and the history retention according to parameters determined by initialization. The number is determined (step 965). Next, the multiplicity and the number of history records are sent to the command processing management unit 75 (step 970).
Next, it waits for an input command (step 975).

The method for controlling multiple writes of the same data according to the present invention is also effective for data recovery when a failure occurs in a disk drive.

Next, the processing when a failure occurs in the disk drive during the input / output processing will be described with reference to FIG. When the disk drive management unit 85 detects that a failure has occurred in any one of the disk drives (step 825 in FIG. 9) and receives the report (step 9).
00), the multiple write fault management unit 80 temporarily suspends the processing of the input / output command, and reports the occurrence of the fault to the CPU 1 (step 905). Next, the disk drive number or the physical address of the failed disk drive is deleted from the data address management table 50 and the use is prohibited (step 910). Next, the empty area management table 55
Within the disk drive, delete the failed disk drive number or physical address and prohibit use (step 9)
15). Then, it is confirmed whether the fault that has occurred can be recovered (step 920).

If the failure can be recovered, the next step 925
Then, it is confirmed whether or not to perform the failure recovery. When performing a fault recovery, the fault recovery process is performed as it is. The details will be described with reference to FIGS. 12 and 13 (step 930). If the failure recovery is not to be performed, the processing of the input / output instruction is resumed as it is (step 935). If it is determined that the recovery from the failure is not possible, then the CPU 1 is inquired of the use of the history data (step 940). When the history data is used, it is confirmed whether or not the data having the old history number can be used (step 945). If available, the history number of the next oldest data is updated, and the number of history records is reduced accordingly. Then, the process proceeds to step 925. If the data with the oldest history number is not available, the data is regarded as a loss, the fact is notified to the CPU 1 and the system is stopped (step 955).

Failure recovery processing (step 930 in FIG. 11)
FIG. 12 is a flowchart showing the detailed processing procedure of.

If it is determined in step 925 in FIG. 11 that recovery from a failure is to be performed, the multiple write failure management unit 80 starts recovery from a failure (step 1000). Next, the data address management table 50 and the empty area management table 55 are reorganized (step 1010). Here, if it is necessary to make up for the lack of capacity, step 1005
As shown in (1), an unused spare disk drive may be incorporated. However, if the capacity EC of the empty area is sufficiently larger than the capacity of the data in the failed disk drive, the method of the present invention can recover from the failure without a spare disk device. Next, a threshold value of the number of input / output instructions waiting to be used, a data recovery capacity for one time, a time for giving priority to input / output processing, or the number of input / output processing to be used for the judgment of switching between priority control of input / output processing and failure recovery processing are set. (Step 1015). In the present embodiment, as described below, the failure recovery processing is executed prior to the input / output processing. However, when the number of input / output instructions in the processing standby state exceeds a certain number, the processing in the processing standby state is performed. The input / output processing is preferentially executed in order to prevent an abnormal increase in the number of input / output instructions. In other words, in this embodiment, data recovery processing can be performed while executing other input / output processing. Each parameter set in step 1050 is used to determine the switching of the process to be executed with priority. Next, the number of input / output instructions waiting to be processed is counted from the disk drive management table 60 (step 1020). This is performed by calculating the total sum of the number of input / output instructions waiting to be processed, corresponding to each disk drive number registered in the disk drive management table 60. As the next step, it is confirmed whether the number of input / output instructions to be processed is equal to or smaller than a threshold value (step 1025).

If the number of input / output instructions waiting to be processed is equal to or greater than the threshold value, the input / output processing is executed for a predetermined time or the number of input / output processings set in step 1015 (step 1035). Proceed to 1040.

If the number of input / output instructions waiting to be processed is equal to or smaller than the threshold in step 1025, the data recovery processing is executed (step 1030), and then step 1040 is executed.
In step 1040, it is confirmed whether a logical address for which data recovery has not been completed remains. If any error remains, the process returns to step 1020 to continue failure recovery. When the data recovery processing is completed for all the failed logical addresses, the failure recovery ends and the CP
A report is made to U1 (step 1045).

It is also possible to use the threshold value of the number of input / output instructions stored in the input / output buffer for the determination of switching the control to the input / output processing priority. Reference numeral 1020 counts the number of input / output instructions stored in the input / output instruction buffer. In addition, although the failure recovery here is intended to perform the failure recovery without stopping the processing of the input / output instruction, it is also possible to perform the recovery after stopping the processing of the input / output instruction.

Next, the details of the data recovery process (step 1030) in FIG. 12 will be described with reference to the flowchart in FIG.

When the data recovery process is started (step 1100), first, the logical address of the data to be recovered is set (step 1105). The logical address of the data to be recovered is obtained by referring to the data address management table 50 and extracting the logical address registered in the data address management table 50 based on the disk drive number of the failed disk drive. Next, an output instruction is issued from the command processing management unit 75 for each of the logical addresses set in step 1105 (step 1110). Next, the same data as the data stored in the failed disk drive is read from another disk drive by performing the output command processing described in FIG. 7 (step 1115). Next, referring to the disk drive management table 60, a free disk drive number in which the number of waiting input / output instructions is equal to or less than a certain number is searched (step 1120). A search is made for the amount of data read in step 1115 (step 112).
5). Note that the steps 1120 and 1125 can perform parallel processing in addition to performing the processing in this order. Next, a disk drive having data satisfying the conditions found in step 1120 and having an empty area also found in step 1125 is selected (step 1130). In the next step 1135, it is checked whether the multiplicity can be secured only by the selected disk drive.

If the number of disk drives is not enough, a disk drive having an empty area is selected at step 1140 from among disk drives having a certain number of input / output instructions, and then the process proceeds to step 1145. Next, the logical address of the data to be recovered is converted to the physical address of the disk device secured in step 1130 using the data address management table (step 1145). Next, the input command is sent to the disk drive management unit 85 (step 1150), and the data read in step 1115 is written to the disk drive secured in steps 1130 and 1140. Upon receiving the processing report of the input command from the disk drive management unit 85 (step 1155), the empty area in which the newly recovered data has been input is stored in the empty area management table 5.
5 (step 1160), and registers the new physical address and logical address of the recovered data in the data address management table 50 in association with each other (step 1165). Next, the processing status is checked and recorded (step 1170). Finally, the data recovery processing ends (step 1175).

In this embodiment, the reading of the recovery data is as follows:
Multiple write failure management unit 80 to command processing management unit 7
5, the multiplex failure management unit 80 independently writes the read recovery data. However, the multiplex failure management unit 80 also executes command processing for writing the recovery data. It is also possible to send an input / output command to the management unit 75 so that the processing from step 1120 to step 1150 is performed by the command processing management unit 75.

In the above-described embodiments, the magnetic disk drives are listed as the storage units in the rows. However, the storage units may be configured using an optical disk drive, a floppy disk drive, or a semiconductor memory. Also, the present invention relates to the control unit 2 of the secondary storage device 101, the disk drive control units 90-1 to 90-n, the disk drive 16-
1 to 16-n and the like are stored in one housing, but in addition, the control unit 2 of the secondary storage device 101, the disk drive control units 901-90-n, and the disk drives 16-1 to 16-n
16-n can be distributed on a network and the operations described in the present invention can be applied to them.

According to the embodiment of the present invention, when processing an input command, if the disk drive on which the original data exists cannot be used for processing another input / output command, another available disk drive is used. Store the data in the empty area. This eliminates the need for the CPU to wait for an input command despite the fact that there is a usable disk drive, thereby improving the input / output processing performance. Also, when recovering from a disk drive failure, the data to be recovered is a copy of the multiplexed data written to several disk drives, and the data is in a writable state without being written to a specific disk drive. Since recovery data is written to the disk drive at any time, it is possible to realize a failure recovery process that minimizes the deterioration of the input / output processing performance. Also, if the user specifies only the logical address of the requested data, the actual data address, multiplexing of data, and updating of the data history are converted and processed inside the device, so the user does not need any extra operation. .

Further, in the secondary storage device to which the present invention is applied, it is possible to easily add a disk drive.

That is, for example, it is assumed that the (n + 1) th disk drive is added in the secondary storage device 101 shown in FIG. 1 or FIG. In this case, as indicated by a broken line in FIG. 4, a disk drive number # n + 1, all of which are empty areas, is added to the empty area management table 55. On the other hand, as indicated by the broken line in FIG. 5, the number of input / output instructions waiting to be processed is zero, that is, a disk drive number # n + 1 of "000" in binary is added.

When the control unit 2 executes an input / output command from the CPU 1 as shown in FIG. 8 in such a state, the added disk drive # n + 1 has sufficient free space and is Since the number of output instructions is small, there is a high possibility that the disk drive is selected as a multiplex writing disk drive as compared with other disk drives. Therefore,
The data capacity of the added disk drive # n + 1 increases rapidly, and becomes equal to that of other disk drives.

As described above, the secondary storage device to which the present invention is applied has high flexibility in both adding a storage unit and reducing the number of storage units due to a failure.

[0083]

As described above, according to the present invention, data is rearranged without imposing a burden on the user.
It is possible to realize a secondary storage device capable of improving the performance of the input / output processing performance and suppressing the performance degradation at the time of failure recovery.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a secondary storage device according to an embodiment of the present invention.

FIG. 2 is a configuration diagram showing a control unit 2 shown in FIG.

FIG. 3 is a configuration diagram showing an example of a data address management table of the present invention.

FIG. 4 is a configuration diagram illustrating an example of an empty area management table according to the present invention;

FIG. 5 is a configuration diagram showing an example of a disk drive management table according to the present invention.

FIG. 6 is a flowchart illustrating a processing procedure of an input / output management unit.

FIG. 7 is a flowchart illustrating an output instruction processing procedure of a command processing management unit;

FIG. 8 is a flowchart illustrating an input command processing procedure of a command processing management unit;

FIG. 9 is a flowchart illustrating a processing procedure of a disk drive management unit.

FIG. 10 is a flowchart illustrating a processing procedure of a normal multiple write failure management unit;

FIG. 11 is a flowchart illustrating a processing procedure of a multiple write failure management unit when a failure occurs.

FIG. 12 is a flowchart illustrating details of a failure recovery process.

FIG. 13 is a flowchart illustrating details of a data recovery process.

[Description of Signs] 1 CPU, 2 control unit, 3 external interface path, 4 interface port, 5 channel path switch, 6 channel port, 7 command buffer,
8 data buffer, 9 internal data path, 10 internal command path, 11 microprocessor, 12 management table, 13 drive port, 14 drive path switch, 15 arbitration processor, 16 disk device, 17 … Drive processor,
18: drive command path, 19: drive data path.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-63-56873 (JP, A) JP-A-2-280221 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 3/06 G06F 12/16

Claims (25)

(57) [Claims] 1. A method for performing multiplex writing of the same data to a plurality of storage units, the method comprising: responding to a data write command issued from a host device, for each of the first plurality of storage units. A data read command and a data write command issued earlier from the higher-level device, which are already in a processing wait state for the storage unit, are detected. Based on the detection result, the preceding data in a process waiting state is detected. A first data read command or a data write command which has no read command or data write command or is in a waiting state for processing, and the total waiting time of processing of the data read command or the write command is smaller than the first time. more less the second plurality of storage units is selected, the selected second plurality of storage Multiple write how the same data respectively to the knit, characterized in that it comprises the step of writing data the data write command requires
Law. 2. A method for reading data multiplexed stored in a second plurality of storage units less than the first plurality of storage units in the first plurality of storage units, the method comprising: In response to the read command, for each of the first plurality of storage units equal to the second plurality of storage units storing data requested by the command, A data read command and a data write command issued earlier from the higher-level device, which are already in a process waiting state, and, based on the detection result, among the second predetermined number of storage units, Has no data write command or data read command in the state, or has a data write command or data read command already in the waiting state, and All processing latency data read or data write command selects fewer less than the plurality of second third plurality of storage units, each of the read life to the selected storage unit
Command and the first connection from the selected storage unit.
In response to a follow-up report, the storage unit that
The read command is cached for storage units other than
The data from the storage unit that first reported the connection.
Means for reading and outputting to the higher-level device
A data reading method characterized by the above-mentioned. 3. A method for performing multiplex writing of the same data to a plurality of storage units, wherein a storage address of each of the data in the first storage unit is registered in a data address management table; For each of the storage units, the free storage area address and the free capacity are registered in the free area management table. For each of the plurality of storage units, the total number of read instructions or write instructions in the processing waiting state is stored in the storage unit management table. In response to a data write command from a higher-level device, the device refers to the free space management table and the storage unit management table, has free space capable of storing data requested by the write command, and , The total number of read instructions or instructions waiting to be processed is less than the first plurality. Selecting a second plurality of storage units that are not present, referring to the data address management table, and writing data requested by the write command to each free space of the selected second plurality of storage units; after writing the data, characterized in that it comprises the step of updating the data address management table, the free area management table and the storage unit management table
Multiple writing method of the same data. 4. The method according to claim 1, wherein the step of selecting a second plurality of storage units includes the step of: if the number of storage units having free space capable of storing the write data is less than the second plurality.
With reference to the address management table, an area of the data having the oldest degree of history of the write data in the plurality of storage units is changed to a free area, and the storage unit including the area is changed to a storage area of the second plurality of storage units. Claims further comprising the step of selecting as one.
Item 4. A method for multiple writing of the same data according to Item 3. 5. The same data according to claim 3 , further comprising the step of adding a storage unit and updating the free space management table and the storage unit management table in accordance with the addition of the storage unit.
Multiple writing method. 6. A method for reading data multiplexedly stored in a second plurality of storage units less than the first plurality of storage units in the first plurality of storage units, the method comprising: For each of the data, the storage address, the degree of history, and the number of multiple writes are registered in the data address management table. For each of the plurality of storage units, the total number of read or write commands waiting to be processed is stored in the storage unit management table. Registering, referring to the storage unit management table in response to a data read command from a higher-level device, and setting the second read and write command waiting for processing to be smaller than the total number of the read and write commands.
And selecting the third plurality of storage units less than the plurality of storage units, sending the read command to each of the selected third plurality of storage units, and reporting the first connection from the selected third plurality of storage units. In response to, cancel the read command sent to a storage unit other than the storage unit that first reported connection, and refer to the data address management table for the data from the storage unit that first reported connection. reading,
Characterized by having a step of outputting to the host system
Data reading method. 7. A data recovery in a secondary storage device in which the same data is multiplexed and stored in a second plurality of storage units in the first storage unit, the number being smaller than the number of the first plurality of storage units. A method of registering each of the data in the first plurality of storage units in a storage address and a data address management table, and for each of the plurality of storage units, the free storage area address and the free capacity thereof Registering in the free space management table, and for each of the plurality of storage units, registering the total number of read and write commands waiting to be processed in the storage unit management table. When a failure occurs in a certain storage unit, the storage unit management Updating the table to disable the failed storage unit; Referring to a table, data stored in a normal storage unit having the same contents as data used for recovery of data (data to be recovered) in the storage unit in which the failure has occurred is determined. Referring to the table and the storage unit management table, select a storage unit having a free space capable of storing the data to be recovered and having the smallest total number of read instructions or write instructions waiting for processing, It refers to the address management table, data recovery side, characterized in that it comprises the step of writing the correct data to be used for the recovery to the selected storage unit
Law. 8. The step of determining data to be recovered,
8. The data recovery method according to claim 7 , wherein it is determined that data recovery is to be performed using the history data, and the degree of data history in the data address management table is reduced by one. 9. The multiplexing of the same data according to claim 1 , wherein said storage unit is a disk drive.
Writing method. 10. The data reading device according to claim 2 , wherein said storage unit is a disk drive.
Method. Wherein said selection claims, characterized in that it comprises the step of selecting a plurality of storage units of second in response to the total number of read instructions or write instructions waiting to be processed in each storage unit Multiple copies described in 1
How to write. 12. The method of claim 11, wherein selection is claims, characterized in that it comprises the step of selecting a plurality of storage units of second in response to the total number of read instructions or write instructions waiting to be processed in each storage unit Multiple copies of 2
How to write. 13. The method according to claim 13, further comprising: searching for a free storage capacity of each of the plurality of storage units in response to the write command from the higher-level device, and selecting the second plurality of storage units. Selecting the second plurality of storage units from storage units having free space capable of storing the data output from the higher-level device based on the search result .
2. The multiple writing method according to claim 1, wherein: 14. A control device for multiplex writing of the same data to a plurality of storage units, comprising: means for receiving a data write command from a host device; and responding to the input of the write command in the receiving means. Means for detecting, for each of the plurality of storage units, a read and write instruction from the higher-level device in a processing wait state, based on the detection result, the small storage unit, means for selecting a predetermined number smaller than the number of said plurality of storage units, control, characterized in that it comprises means for writing the data outputted from each of the host device to the selected storage unit apparatus. 15. A control device for reading data multiplexed stored in a first predetermined number of storage units in a plurality of storage units, comprising: means for receiving a data read command from a host device; Means for detecting, in response to a read command, a read and write command from the higher-level device in a processing wait state for each of the plurality of storage units; and reading and writing from the higher-level device based on the detection result. Means for selecting a second predetermined number of storage units having a short processing wait time for the instruction, the second predetermined number being less than the first predetermined number; means for sending the read command to the selected storage unit; and the selected storage unit In response to the first connection report from the storage unit, the storage unit other than the storage unit that first reported the connection is read. Means for canceling out the instruction, the first reading the data from the connection reported by the storage unit, the control apparatus characterized by comprising: means for outputting to the host device. 16. A control device for performing multiplex writing of the same data to a plurality of storage units, wherein for each of the data in the plurality of storage units, a storage address, a history degree, and the number of multiplex writes are managed. A data address management table; a free space management table for managing the free storage area address and free space for each of the plurality of storage units; and a process waiting read and write instruction count for each of the plurality of storage units. A storage unit management table to be managed; a unit for receiving a data write command from a higher-level device; and, in response to the write command, referring to the free space management table and the storage unit management table and output from the higher-level device. Has a free space capable of storing the data, and Means for selecting a predetermined number of storage units having a small number of read and write commands waiting for writing, and referring to the data address management table, the data output from the higher-level device to a free area of the selected storage unit, respectively. means for writing, after writing the data, the data address management table, the controller characterized in that it comprises a means for updating the free area management table and the storage unit management table. 17. The means for selecting a predetermined number of storage units refers to the address management table if a predetermined number of storage units having a free space capable of storing the data output from the host device cannot be selected. Changing an area of the data having the oldest data history in the plurality of storage units to a free area.
The control device according to claim 16. 18. A control device for reading data multiplexed stored in a first predetermined number of storage units in a plurality of storage units, wherein each of the data in the plurality of storage units is read out. A data address management table for managing a storage address, a degree of history, and the number of multiple writes; a storage unit management table for managing the number of read and write instructions waiting for processing for each of the plurality of storage units; Means for receiving from the device;
Means for, in response to the read command, referring to the storage unit management table and selecting a second predetermined number of storage units having a smaller number of the waiting read and write commands than the first predetermined number; Means for sending the read command to each of the selected storage units; and in response to a first connection report from the selected storage unit, the read command is sent to storage units other than the storage unit that first reported the connection. Means for canceling, reading the data from the storage unit that first reported the connection with reference to the data address management table,
Means for outputting to the higher-level device. 19. A controller in a secondary storage device in which the same data is multiplexly stored in a predetermined number of storage units in a plurality of storage units, wherein each of the data in the plurality of storage units is: A data address management table that manages the storage address, the degree of history, and the number of multiple writes; a free space management table that manages a free storage area address and free space for each of the plurality of storage units; A storage unit management table for managing the number of read and write instructions waiting to be processed, a unit for detecting the occurrence of a failure in a certain storage unit, and updating the storage unit management table to use the failed storage unit. Means for prohibiting the data, and in response to the occurrence of the fault storage unit, Means for determining data to be recovered by referring to an address management table; storing the data to be recovered by referring to the free space management table and the storage unit management table in response to the occurrence of the failed storage unit Means for selecting a predetermined number of storage units having a possible free space and having a small number of read and write instructions waiting for processing; and referring to the data address management table, the recovery is performed for each of the selected storage units. A control unit for writing data to be recovered, and a unit for updating the data address management table, the free space management table, and the storage unit management table after writing the data to be recovered. 20. A means for determining the data to be the recovery claims decides to perform data recovery using the history data, and wherein the lowering one history of the data in the data address management table Item 19. The control device according to Item 19 . 21. The control device according to claim 14 , wherein said storage unit is a disk drive. 22. The control device according to claim 15 , wherein said storage unit is a disk drive. 23. The control device according to claim 14 , wherein said processing waiting time is determined according to the number of read and write commands from said higher-level device. 24. The control device according to claim 15 , wherein said processing waiting time is determined according to the number of read and write commands from said higher-level device. 25. Further, in response to the write command from the higher-level device, means for searching for a free storage capacity for each of the plurality of storage units, and means for selecting a predetermined number of the storage units, 15. The storage system according to claim 14, wherein a predetermined number is selected from storage units having a free space capable of storing the data output from the host device.
The control device as described .