JP3411451B2 - Disk array device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array device, and more particularly to reducing the amount of data transfer on the interface between the disk controller and the disk device in the disk array device when data is duplicated. The present invention relates to a disk array device that improves the processing performance of a disk control device.

[0002]

2. Description of the Related Art In a disk array device for the purpose of improving the reliability of a storage device, in order to prevent data read / write from becoming impossible due to a failure, data is multiplexed on a plurality of magnetic disk devices. It uses a redundant configuration to record. Several types of redundant configurations are known as "ACase for Redundant Arra
y of Inexpensive Disk (RAID) '' (Proceedings of the 1
988 ACM SIGMOD Conference on the Management of Dat
a 1988). Among the several types of redundant configurations specified in this paper, especially in the case of the redundant configuration specified by the name of RAID1 in which data is redundantly written to two magnetic disk devices, Even if one magnetic disk device fails, the same data is recorded in the other normal magnetic disk device, so that the data can be read.

FIG. 8 is a schematic diagram showing an example of the configuration of a disk array device using a conventional data duplication system. The disk array device is a disk controller 801,
And disk devices 802a, 802b, 803a, 80
3b. The disk control device 801 is a SCSI control circuit 809 that controls a communication path with the disk device.
a, 809b, 809c, 809d, interface control circuit 8 which is an interface with the host computer
05, microprocessor 806, cache memory 8
08. Further, FIG. 9 is a flowchart showing a sequence of new data writing in such a conventional disk array device. 8 and 9, data writing from a host computer in a disk array device using a conventional data duplication method,
The operation for a read request will be described below.

First, in the case of a write request, the disk control device 801 in the disk array device which has received the write request from the host computer is a redundant configuration of two magnetic disk devices 802a and 802b. To the respective magnetic disk devices 802a and 802 by receiving the WRITE command or parameter (step 901).
02b starts the seek search operation of moving the position of the magnetic head to the position where the data should be transferred on the disk and waiting until the position where the data should be written rotates right under the head (step 902). In parallel with the magnetic disk devices 802a and 802b performing the seek search operation, the disk control device 801 transfers the data to the buffer memory in the magnetic disk device and ends the data transfer processing ( Step 903). When the seek search operation is completed, the magnetic disk drive 80
2a and 802b write the data in the buffer memory to the disk (step 904), and notify the disk controller 801 that the writing has been completed normally. Upon receiving this notification, the disk control device 801 confirms that the two magnetic disk devices have normally written, and notifies the host computer that the write request has been normally completed.

Next, there is a case where a data read request is received from the host computer. In this case, there are two methods.

In the first method, the disk controller 801 in the disk array device issues a read command and a parameter to one of the predetermined magnetic disk devices 802a. Receiving this, the magnetic disk device 802a starts the seek search operation, and when the data transfer becomes possible, sends the data to the buffer memory in the magnetic disk device 802a. In addition, the magnetic disk device 802a notifies the disk control device 801 that the data transfer is possible in parallel with the operation of transferring the data to the buffer memory, and the disk control device 801 that received this notification The data is read from the buffer memory in the magnetic disk device 802a and transferred to the host computer.

The second method is a method of issuing a command to the faster completion of the seek / search operation of a pair of magnetic disk devices having a redundant configuration. In the case of this method, since the magnetic disk device of the two magnetic disk devices which is in a state where data transfer is possible earlier is selected, the read time of the magnetic disk device can be shortened. However,
This method requires the disk control device to manage the queue of commands already issued to the respective magnetic disk devices.

Here, a problem with the conventional data duplication system is that, in the case of new data writing processing, the same data must be written to two magnetic disks, so that a normal magnetic disk device is simply used. That is, the data transfer processing time of the microprocessor is twice as much as that in the case of writing to the memory. Due to this problem, the load on the disk control device is large and the processing performance of the disk control device is degraded.

In the method disclosed in Japanese Unexamined Patent Publication No. 7-281959, which is considered to be one example for solving this problem, means for communicating the respective operating states between two magnetic disk devices is provided in the magnetic disk device. With this contact method,
A method is provided in which two magnetic disk devices simultaneously perform a write operation. When this method is used, the data transfer amount between the magnetic disk device and the disk control device is the same as the data transfer amount for one magnetic disk device.

[0010]

In the data duplication method described as the prior art, the data transfer processing load on the disk controller increases because data is transferred to two magnetic disk devices. Also, in the method disclosed in Japanese Patent Laid-Open No. 7-281959 cited as an example for solving this problem, the magnetic disk device must be provided with a dedicated connecting means and a connecting line, and two magnetic disks are used when writing data. Since the device is occupied, the utilization efficiency of the magnetic disk device is reduced in the disk array device in which a large number of magnetic disk devices are connected and the multiplicity of data transfer is increased to improve the performance.

It is an object of the present invention to provide a method for reducing the amount of data transfer between a disk control device and a magnetic disk device in data duplication in a disk array device using a RAID 1 redundant configuration, and to provide an existing magnetic field. It is to realize this method without adding any hardware such as a dedicated control means for this method to the disk device.

[0012]

In order to achieve the above object, a plurality of disk devices, a disk controller for controlling data input / output to / from the disk device in response to an input / output request from a host computer, and a plurality of the disk controllers are provided. Disk array device and an interface for connecting the disk control device, a plurality of the disk devices form a set, and a disk array device using a redundant configuration in which data is redundantly stored Means for writing data from the control device to the first disk device via the interface, and the first disk device is of the same redundant configuration group existing on the interface to which the first disk device is connected. A means for transferring the data is provided to the second disk device.

In addition, temporary storage means for temporarily holding a plurality of commands and data in the disk device, and commands and data from the disk device of the same redundant configuration group during execution of other commands are stored in the disk device. A means for receiving and holding the temporary storage means and a means for changing the execution order of a plurality of commands waiting for execution held in the temporary storage means are provided.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION A method of data duplication in a disk array device using a RAID 1 redundant configuration provided by the present invention will be described in detail with reference to FIGS.

FIG. 1 schematically shows an embodiment of a disk array device according to the present invention. In the embodiment shown here, FC-AL (Fibre) is used as an interface for connecting the disk control device and the magnetic disk device.
Although a case where a channel arbitrated loop) is used is shown, the present invention can be realized by connecting a storage device other than a magnetic disk device (such as a magneto-optical disk device) or by using an interface other than FC-AL. Can be implemented.

As shown in FIG. 1, the disk array device is a disk control device 101 and magnetic disk devices 102a, 1a which form a RAID1 redundant configuration group in pairs.
02b, the magnetic disk devices 103a and 103b, and an FC connecting the disk control device and the magnetic disk device.
-It is composed of the AL interface 104. The disk control device 101 includes an interface control circuit 105 that controls an interface with a host computer, a microprocessor 106, an internal bus 107, a cache memory 108 that temporarily holds data transferred between the host computer and the magnetic disk device, and an FC. -A
FC-AL control circuit 1 for controlling the L interface
09. The microprocessor 106 is operated by a program in the memory not shown in FIG.

When a write request is issued from the host computer, the write data is first written in the cache memory 10.
Written in 8. The microprocessor 106 performs FC conversion of the data held in the cache memory 108 according to the parameters of the write request from the host computer.
While passing the data to the -AL control circuit 109, the FC-AL control circuit 109 is instructed to write the passed data to the magnetic disk device. At this time, the microprocessor 106 issues a write command to the destination magnetic disk device (here, temporarily the magnetic disk device 102a),
And a magnetic disk device 102b paired with it
The ID number for identifying the (data duplication disk) from other disk devices and the control information such as the logical block address where the data is to be written are FC-A at the same time.
Transfer to the L control circuit 109. The FC-AL control circuit 109, which has received this write command, adds DD based on the above control information.
Create a WRITE command (described later) and add this DDW
By issuing the RITE command to the magnetic disk device via the FC-AL interface, the data from the host is transferred to the magnetic disk device.

The magnetic disk device 10 which has received the write command
The data 2a writes the transferred data to the disk and transfers the data to the second magnetic disk device 102b forming a set.

In the embodiment of FIG. 1, magnetic disk devices 102a, 102b, 103 are compared with FIG. 8 which is a conventional example.
FC-AL interface 10 with the same a and 103b
4 are connected to each other, it is possible to perform data transfer for duplication between the magnetic disk devices proposed by the present invention without going through the disk control device 101, and the disk control device 101 in the data duplication. It is possible to reduce the amount of data transfer between the disk and the magnetic disk device.

The interface for connecting the disk control device and the magnetic disk device shown in FIG. 1 is equipped with a function for arbitration (arbitration) of the bus. As the device, it is preferable to use a device that can be easily mounted as hardware and to use a device having a high multiplicity of data transfer per unit time. The arbitration function of this bus is used when the disk control device sends a command to the magnetic disk device, or when the magnetic disk device sends a command to the duplicated disk device forming the duplicated data.

FIG. 2 shows the magnetic disk device 102a of FIG.
2 shows an example of the configuration of FIG. However, the magnetic disk devices 102a, 102b, 103a, 10 of FIG.
It is assumed that all 3b have the same configuration. The magnetic disk device 102a includes a magnetic disk control unit 201 that controls a magnetic disk, a disk 202 that stores data, and an arm 204 that includes a magnetic head 203 that transfers data to and from the disk 202. In addition, the magnetic disk control unit 2
01 is an FC-AL control circuit 21 for controlling FC-AL which is an interface with the disk control device 101.
3. Microprocessor 20 for controlling the magnetic disk
5, a disk controller 206 for controlling the arm 204 and the disk 202, a read / write control circuit 207 for controlling writing / reading, and a buffer memory 208 as a means for temporarily holding the received commands and data. .

In FIG. 2, the buffer memory 208 includes a read data 209 buffer, a write data 210 buffer memory, and a duplex data transfer destination address buffer memory. In the buffer memory for the data transfer destination address for duplication, the magnetic disk device is used as a DDWR.
When receiving the ITE command, the ID number of the duplication magnetic disk device, which is the duplication data transfer destination address indicated in the DDWRITE command, is stored. This transfer destination address is used when transferring data to the magnetic disk device of the duplication destination.

Further, by increasing the capacity of the buffer memory 208, it is possible to temporarily hold a plurality of commands and data from the interface in the buffer memory. For this reason, it is possible to receive and hold commands and data from the magnetic disk devices of the same redundant configuration group in the buffer memory even during execution of commands other than the DDWRITE command. As a result,
It is possible to change the execution order of a plurality of commands waiting for execution held in the buffer memory, and if necessary, execute the duplication data write processing first, or vice versa. It becomes possible to write the duplicated data in consideration of the occurrence of normal processing, such as after the processing, and it is possible to configure a flexible disk array device.

Next, the command will be described with reference to FIGS. 3 and 4.

FIG. 3 shows the structure of a new command (called a DDWRITE command) in the FC-AL interface of this embodiment. This command instructs the magnetic disk device 102a to write data from the disk control device 101, and the magnetic disk device 1
02a, which has a content to instruct the magnetic disk device 102b which is a data duplication destination to transfer the data, and is created by the EC-AL control circuit 109.
publish. FC-AL interface is SCSI-
It uses the operation code C0H (byte 0) that supports the three standards and may be independently defined by the vendor of the group 6 defined by the three standards. This command includes a "transfer destination magnetic disk device SISI ID number" (FIG. 3) indicating the magnetic disk device 102b which is the duplication destination.
8th byte). Magnetic disk device 102a
By referring to this, the disk controller 10 is added to the magnetic disk device 102b which is the data duplication destination.
The data received from 1 is transferred.

FIG. 4 shows a DDWR using an existing command.
XO used when you want to realize the function of ITE command
The structure of the R command is shown. The XOR command (operation code 80H) is specified by SCSI-3,
By setting the unused bits (the 5th and 6th bits of byte 1 in FIG. 4) in this command to the presence or absence of the execution of the exclusive OR operation and setting the execution of the exclusive OR operation to no, the DDWRITE The function equivalent to the command can be realized.

Exclusive OR (XOR: eXclusiv)
e OR) is an operation required to generate parity data (redundant data for data recovery at the time of failure) necessary for RAID4 and RAID5 among the redundant configurations existing in the disk array device described above. is there. RA that stores data in blocks on multiple magnetic disk devices
In ID4 and RAID5, data in a plurality of magnetic disk devices having a redundant configuration is subjected to exclusive OR operation in block units, and the operation result is stored as parity data in the magnetic disk device.

FIG. 6 is a diagram simply showing the operation of the disk array device by the original XOR command and the transfer route of the command and data. In FIG. 6, it is assumed that four groups of magnetic disk devices 602a, 602b, 602c, and 602d form one group having a RAID 5 redundant configuration. The operation of the disk array device by the XOR command will be briefly described below.

Upon receiving the data write request from the host computer, the disk controller 601 issues an XOR command to the magnetic disk device 602a to which the data should be written. The magnetic disk device 602a is a magnetic disk device that stores old data to be rewritten, and new data is written to a buffer memory that is a temporary holding unit (path 604). At the same time, the magnetic disk device 602a copies old data from the disk. This is the result of the exclusive OR operation of the new data and the old data after reading and performing the exclusive OR operation of the new data and the old data in the magnetic disk device 602a and writing the new data to the disk. The intermediate data is transferred to the magnetic disk device 602d storing the old parity data (path 605).

The intermediate data is stored in the buffer memory of the magnetic disk device 602d which stores the old parity data. After that, similar to the above-mentioned operation, the exclusive OR of the intermediate data and the old parity data is calculated, the new parity data as the result of the calculation is written to the disk, and the normal completion status of the writing of the new parity data is set. Magnetic disk device 602d for storing parity data
From the magnetic disk device 602a that stores the data and receives the normal end status.
02a returns to the disk controller 601 that the XOR command has been completed normally.

Here, among the operations by the XOR command,
If the old data and the old parity data for generating the parity data are not read from the disk and the exclusive OR operation is not executed, the disk device 602a in which the old data is stored is transferred from the disk controller 601. The new data is the disk control device 602a.
Is written to the magnetic disk device 602 and the old parity data is directly stored as intermediate data.
will be transferred to d. And the magnetic disk device 6
In 02d, the intermediate data is directly recorded in the magnetic disk device 6
02d will be written.

That is, in the XOR command shown in FIG. 4, since the 6th and 5th bits of the first byte are unused, the exclusive OR operation is performed depending on whether or not the exclusive OR operation is executed. Set to no execution of operation and
When the command is executed, the data transferred from the disk controller is the magnetic disk device 602a in which the old data is stored and the magnetic disk device 602d in which the parity data is stored, similar to the DDWRITE command.
Will be transferred to.

FIG. 5 shows a DDWRITE command.
FIG. 3 is a diagram showing the operation of the disk array device and the transfer paths of commands and data. Commands or data are transferred in the order of the routes 501 to 505. Also, FIG. 7 shows DDWRI.
7 is a flowchart showing a sequence of the magnetic disk device that has received the TE command. The operation by the DDWRITE command will be described below with reference to FIGS. 5 and 7.

Upon receipt of the data write request from the host computer, the disk controller 101 writes the DDWRITE command to the magnetic disk device 10 to which the data should be written.
Issue to 2a. Here, the disk control device 101
The magnetic disk devices that issue the DDWRITE command may be either of the two magnetic disk devices that are paired. It may be decided to always issue to one of them, or a method of randomly selecting either of the two may be used. A preferred embodiment is a method in which the disk controller 101 manages a queue of commands that have already been issued to the magnetic disk device and are waiting to be executed, and the command is issued to the shorter one of the queues. .

The queuing function defined in the SCSI specification is used as the function for rearranging the execution order of the queue of pending commands already in the queue, and the logical block address and the number of transfer blocks indicated by the command parameters are used. Therefore, it may be realized by using a method for optimizing the seek search operation to be the minimum.

The magnetic disk device 102 which received the command
a starts the seek search operation (step 70).
1, 702). At the same time, the new data transferred from the disk controller is stored in the buffer memory, which is a temporary holding means (step 703), and waits until a data writable state is reached (step 705). When the seek search operation is completed (step 704) and the data transfer to the disk becomes possible, the magnetic disk device writes the new data from the buffer memory to the disk (step 706). Next, for duplication of data, the magnetic disk device 102
The a transfers the new data to the other magnetic disk device 102b forming the same RAID1 group. At this time, the command issued by the magnetic disk device 102a to the magnetic disk device 102b may be the same as the normal write command used by the disk control device to instruct the magnetic disk device to write (step 707). The operation until the new data is written to the disk is the same as that of the magnetic disk device which first received the command (step 708-).
713).

When the writing of the new data to the disk by the DDWRITE command is completed, the normal end status of the writing is written from the magnetic disk device 102a for storing the data to the magnetic disk device 102 for storing the data.
The magnetic disk device 102b, which has been transferred to the disk b, receives the normal end status and returns to the disk control device 101 that the DDWRITE command has been normally ended. In this case, the disk control device 101 can know that the dual data writing is completed only by receiving the status indicating normal completion, and compared with the case where the completion status is received from each magnetic disk device. , The number of times the status is received is reduced, the management becomes easier, and the overhead can be reduced. Alternatively, the two magnetic disk devices 102a and 102b may directly return the normal end status to the disk control device 101 when the write operation to the disk ends normally. In this case, the disk controller 101 uses DDWRIT
The E command is determined to be normally completed when the normal completion status is received from the two magnetic disk devices 102a and 102b.

[0038]

According to the present invention, in a disk array device using a RAID 1 redundant configuration, the data transfer amount between the disk control device and the magnetic disk device can be reduced and the processing performance can be improved. Further, the present invention is effective with a small number of interfaces, and a high-performance disk array can be realized at low cost because no special control means is required for the present invention.

[Brief description of drawings]

FIG. 1 is a schematic diagram showing a configuration example of a disk array device according to the present invention.

FIG. 2 is a schematic diagram showing a configuration example of a magnetic disk device according to the present invention.

FIG. 3 is a configuration diagram showing a configuration example of a command for executing data duplication between magnetic disk devices according to the present invention.

FIG. 4 is a configuration diagram of an XOR command used in a disk array device having a redundant configuration of RAID 4 or 5 according to the present invention.

FIG. 5 is a schematic diagram showing an example of the operation of the disk array device according to the present invention.

FIG. 6 is a schematic diagram showing an example of the operation of the disk array device according to the present invention.

FIG. 7 is a flowchart showing an example of the operation of the disk array device according to the present invention.

FIG. 8 is a schematic diagram showing the configuration of a conventional disk array device.

FIG. 9 is a flowchart showing the operation of a conventional disk array device.

[Explanation of symbols]

101, 601, 801, ... Disk control device, 102
a, 102b, 103a, 103b, 602a, 602
b, 602c, 602d, 802a, 802b, 803
a, 803b ... Magnetic disk device, 804a, 804
b, 804c, 804d ... SCSI interface,
104, 603 ... FC-AL interface, 10
5, 805 ... Interface control circuit, 106, 20
5,806 ... Microprocessor, 107, 212, 8
07 ... internal bus, 108, 808 ... cache memory,
109, 213 ... FC-AL control circuit, 201 ... Magnetic disk control unit, 202 ... Disk, 203 ... Magnetic head, 204 ... Arm, 206 ... Disk controller,
207 ... Read / write control circuit, 208 ... Buffer memory, 209 ... Read data buffer memory, 210 ...
Write data buffer memory 211 ... Duplication data transfer destination address buffer memory 301 ... SCSI
Command, 401 ... XOR command, 809a, 809
b, 809c, 809d ... SCSI control circuit.

Claims (2)

(57) [Claims] 1. A plurality of disk devices, a disk controller for controlling data input / output to / from the disk device in response to an input / output request from a host computer, and an interface for connecting the plurality of disk devices to the disk controller. A disk array device using a redundant configuration in which a plurality of the disk devices form a set and store data redundantly, and when the data is written, the disk control device transfers the data to the first disk device. a disk controller for writing data through the interface over the face and sends a write command, said first disk device, the same redundant configuration present on the interface said first disk device is connected Send the second write command to the second disk device of the group and transfer the data That has <br/> a disk controller, the disk controller receives a status from the second disk device showing the write successful second disk device, also, the first to the disk controller A disk array device that sends a status indicating the normal end of a write command. 2. The disk device further comprises a temporary storage means for temporarily holding a plurality of commands and data, and a command and data from the disk device of the same redundant configuration group during execution of another command. 2. The disk array device according to claim 1, further comprising means for receiving and holding the temporary storage means, and means for changing the execution order of a plurality of commands waiting for execution held in the temporary storage means.