JPH07261946A - Array type storage device - Google Patents

Abstract

PURPOSE:To prevent the wire penalties for an array type storage device by decreasing the transfer quantity of data for an interface and reducing the rotation waiting time of a disk device. CONSTITUTION:The array type storage device consists of a disk controller 1, an interface 2 which connects together the disk devices 3 in a parity group and these devices 3. Each device 3 contains a function to calculate an exclusive OR and a buffer memory 5 which stores the address of a data transfer destination. When data are stored in the device 3, the old data are updated to the new data and at the same time, an exclusive OR is calculated between the old and new data for generation of the updating position data. This position data is directly transferred to the device 3 which stores a parity. This device 3 calculates an exclusive OR between the old parity and the updating position data to generate a new parity and updates the old parity to the new one. Furthermore, plural actuators 32 are provided in the device 3 so that a writing operation is carried out without waiting for one rotation after the data are read.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an array type storage device, and more particularly, it reduces the write penalty by reducing the data transfer amount of the interface between the disk control device and the disk device and the disk rotation wait. The present invention relates to an array type storage device.

[0002]

2. Description of the Related Art For the purpose of improving the reliability of a storage device, in order to realize data recovery in the event of a disk device failure, a parity (data recovery There is a disk array device for generating and storing redundant data). As for such a disk array device, the case-for-endant array array off-in-expensive disk (rate), procedural off-off 1
988 ACMC Mod Conf on Management Management Data, 1988
A Case for Redundant Arrey of Inexpensive Disk (RAI
D), Proceedings of the 1988 ACM SIGMOD Conf. On the
Management of Data, 1988. In this paper, five types of data storage methods, RAID1 to RAID5, are specified.

RAID4, RAID specified in this paper
In 5, the data is divided into blocks and stored in different disk units. In such a disk array device, parity is generated and stored for each parity group formed by a plurality of disk devices for storing data.
Therefore, when updating data, it is necessary to update the parity at the same time as updating the data.

The generation of the new parity at the time of updating the parity is usually performed by calculating the exclusive OR of the old data, the new data, and the old parity. Also,
The generation of this new parity is performed by the disk controller. Therefore, it is necessary to transfer old data and old parity to the disk controller.

FIG. 11 is a schematic configuration diagram of a conventional array type storage device. This array type storage device includes a disk control device 1, five disk devices 3-1 to 3-5 for storing data and parity, the disk control device 1 and the disk devices 3-1 to 3-5. It is composed of SCSI buses 2-1 to 2-5 as an interface for connecting to each other.

The disk controller 1 has SCSI buses 2-1 to 2-1.
2-5 SCSI bus control circuits 4-1 to 4-5,
An interface control circuit 15 for controlling an interface with a host device, a microcomputer 12, an internal bus 11, a cache memory 13, and a data division / combination control unit 14 are provided.

FIG. 12 shows a sequence for writing new data and updating a parity in such a conventional array type storage device. First, old data and old parity are read from the disk device and transferred to the disk controller (step 151). Next, in the disk controller, the exclusive OR calculation of the old data, the old parity, and the new data is performed to generate the new parity (step 15).
2). Further, the new data and the new parity are transferred from the disk controller to the disk device and written (step 153).

As described above, in the disk array device, by generating and storing the parity for each parity group, for example, even when one disk device fails, the parity is used to cause the error. The data stored in the disk device can be restored, and a highly reliable storage device can be realized.

However, when updating data, in addition to updating the data, it is necessary to update the corresponding parity. Therefore, the old data and old parity are read from the disk unit and transferred to the disk controller, the new parity is generated in the disk controller, and then the new data and new parity are transferred from the disk controller to the disk unit. Need to write. Therefore, for one data update, four reads / writes occur between the disk control device and the disk device. Moreover, after reading the old data or the old parity, at least until the new data or the new parity can be written,
It will wait for the disk to spin once.

As described above, in the RAID4 and RAID5 disk array devices, a so-called write penalty occurs in which the processing performance at the time of writing is lower than that at the time of reading data. Further, since the data transfer amount of the interface increases due to the generation of the new parity, the idle waiting time of the interface increases and the data update processing time increases, which is one of the factors causing the write penalty.

In particular, regarding the interface for connecting the disk device and the disk controller, when a plurality of disk devices in the same parity group are connected by the same interface, the interface becomes a bottleneck of the entire system. , Becomes a factor that limits performance. Further, when the interface fails, the system will be down, and a highly reliable system cannot be realized. For the above reasons, a plurality of disk devices in the same parity group are often connected to the disk control device by using their individual interfaces.

As one of the conventional methods for reducing the write penalty described above, for example, Japanese Laid-Open Patent Publication No. 4-2305.
There is a method disclosed in Japanese Patent No. This is a method of reducing the waiting time for rotation by changing the write position of new data when writing new data to the disk device.

As another method for reducing the write penalty, for example, Japanese Patent Laid-Open No. 5-158621.
There is a method disclosed in the publication. This is a method of generating parity in the disk device that stores the parity, and reduces the transfer amount and write processing time of the disk interface by eliminating the need to transfer the old parity to the disk controller. .

Specifically, first, the old data is transferred from the disk device storing the data to the disk controller,
The new data is transferred from the disk controller to the disk device that stores the data, and the disk controller calculates the exclusive OR of the new data and the old data to obtain the updated position data, and the updated position data is set to the parity from the disk controller. Transfer to the storage disk device. Then, in the disk device that stores the parity, an exclusive OR of the old parity and the updated position data is obtained to generate a new parity, and the new parity is written to the disk.

[0015]

However, in the method of changing the write position of new data (Japanese Patent Laid-Open No. 5-158621), it is not possible to reduce the rotation waiting time after reading the old data until writing the new data. Although possible, the transfer of old data and old parity to the disk controller is still necessary since the new parity is generated in the disk controller. Therefore, the increase in the data transfer amount in the interface is not improved.

Further, in the method disclosed in Japanese Patent Laid-Open No. 5-158621, it is still necessary to transfer the updated position data from the disk control device to the disk device. Therefore, transfer of new data from the disk controller to the disk device that stores data, transfer of old data from the disk device that stores data to the disk controller, and transfer of old data from the disk controller to the disk device that stores parity. It is necessary to transfer the update position data three times in total.

Further, a plurality of disk devices in the same parity group are usually connected to the disk control device by using respective individual interfaces as described above, so that a large number of systems are required for the interfaces and the cost increases. .

An object of the present invention is to improve an array type storage device. In particular, an object of the present invention is to reduce the write penalty by further reducing the transfer amount of the interface in the array type storage device.

Another object of the present invention is to reduce the number of rotation waits for one rotation that occurs after reading old data and writing new data in an array type storage device.
To reduce the light penalty.

[0020]

A disk controller divides data, generates redundant data for data recovery from the divided data, and divides the divided data and the redundant data into the plurality of disk devices. In an array-type storage device that stores the data, the disk device stores the storage position information of redundant data corresponding to the data stored in the disk in the own disk device, and the disk controller instructs to update the data. At this time, the old data to be updated is read from the disk, and the updated position data representing the changed data is generated using the read old data and the new data given from the disk controller. Also,
The old data is updated to the new data, and the updated position data is directly transferred to the disk device that stores the redundant data corresponding to the data. Then, in the disk device storing the redundant data that is the transfer destination of the update position data, the old redundant data to be updated is read from the disk, and the new redundant data is read using the read old redundant data and the transferred update position data. Is generated and written to the disk to update the old redundant data.

As redundant data for data recovery, for example, exclusive OR (parity data) of divided data is used.
And the exclusive-OR (intermediate parity) of the old data and the new data is used as the update position data. However,
Other redundant data for data recovery may be used.

In the present invention, the update position data used for generating new redundant data is directly transferred from the disk device storing the data to be updated to the disk device storing the redundant data. As the interface, for example, the same interface that connects a plurality of disk devices and a disk control device and connects these plurality of disk devices to each other is used. The interface may have a plurality of systems.

Further, the plurality of disk devices and the disk control device are connected by an interface provided individually for each of the plurality of disk devices, and separately, a plurality of disk devices are connected to transfer the update position data. An interface for connecting the disk devices to each other may be provided. A plurality of interfaces may be provided to connect the plurality of disk devices to each other.

Further, in a preferred mode of the present invention, an interface of a plurality of systems is provided between a plurality of disk devices and a disk control device, and a plurality of disk devices are connected to each system, the interface Separately, a plurality of systems of interfaces for connecting a plurality of disk devices to each other are provided to transfer the updated position data. In that case, as the device address of the disk device, in any interface of a plurality of systems to which the disk device is connected, one device address that does not overlap with the device addresses of all other disk devices connected to the interface is used. Try to set it.

The disk device itself may be a second array type storage device composed of a plurality of disk devices and a disk controller for controlling them. That is, the plurality of disk devices themselves, which are units for generating redundant data connected by the same interface, are the second array-type storage device configured by the plurality of disk devices. When the disk control device is directly connected to the interface, the interface device is a disk device that generates update position data in the disk control device that stores data and stores the generated update position data as redundant data. The disk device that directly transfers the data to and stores the redundant data has a function of generating and storing new redundant data from the update position data and the old redundant data.

Further, the disk device is provided with at least readable first data access means and at least writable second data access means for the disk in the own disk device, and the first data The old redundancy data is read by the access means, and new redundancy is made by using the old redundancy data and the update position data until the second data access means can write the same address. It is preferable to perform a calculation for generating data and write the new redundant data by the second data access means.

The disk device includes write data temporary storage means for temporarily storing new data given from the disk control device, read data temporary storage means for temporarily storing old data read from the disk in the own disk device, and generation. The update position data temporary storage means for temporarily storing the updated position data is provided, and new data is stored in the write data temporary storage means, old data is stored in the read data temporary storage means, and old data and new data are stored. The update position data generated using the data and the data may be stored in the update position data temporary storage means in parallel and simultaneously.

[0028]

The update position data is generated in the disk device storing the data, and the generated update position data is directly transferred to the disk device storing the redundant data. In the disk device storing the redundant data, the update position data is updated. Since the new redundant data is generated and stored from the position data and the old redundant data, the redundant data for data recovery can be updated without the intervention of the disk controller. Therefore, it is not necessary to transfer the old data and the old parity (old redundant data) to the disk controller as in the conventional case, and the data transfer amount at the time of updating the data is reduced.

In addition, independent reading,
A plurality of writable data access means are provided, after the old data is read by the first data access means, new data is generated in the disk device, and the new data is written by using the second data access means. Therefore, it is not necessary to wait for the disk to make one revolution after reading the old data.

[0030]

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

1 and 2 show an embodiment of an array type memory device according to the present invention. In the array type storage device of this embodiment, a magnetic disk device (hereinafter referred to as a disk device) is used as a storage device, and a SCSI bus is used as an interface for connecting the disk device and the disk control device. It is also possible to connect a device other than the magnetic disk device as the storage device, or connect using an interface other than the SCSI bus as the interface.

As shown in FIG. 1, the array type storage device of this embodiment is roughly divided into a disk controller 1 and five disk devices 3-1 to 3-3 for storing data and parity.
-5, the disk control device 1 and the disk device 3-1
3 to 5 and a SCSI bus 2 as an interface for connecting to each other.

Disk controller 1 and disk device 3-1
3-5 are connected by the same SCSI bus 2.
The disk control device 1 includes a SCSI bus control circuit 4 that controls the SCSI bus 2, an interface control circuit 15 that controls an interface with a host device, a microcomputer 12, an internal bus 11, a cache memory 13, and a data division / synthesis control unit. 14 are provided.

When a write command for writing data is issued from the host device, the data division / combination control unit 14 divides the write data into storage units of the disk device. The microcomputer 12 issues a write command for the divided data to the SCSI bus control circuit 4. Further, at this time, the microcomputer 12 simultaneously transfers to the write destination disk device the ID number and address of the disk device necessary for storing the parity corresponding to the divided data.

Next, a disk device (for example, disk device 3) that has received a write command issued by the disk controller 1
-1) writes new data to the disk inside the disk device, and updates position data which is the exclusive OR calculation result of the old data read from the disk and the new data to be written. A disk device that generates the updated position data and stores the parity (for example,
Transfer to the disk device 3-5). This update position data transfer is performed from the disk device 3-1 to the disk device 3-5.
Is done directly to.

The disk device 3-5 storing the parity calculates the exclusive OR of the transferred updated position data and the old parity (the old parity is read from the disk in the own machine) to generate the new parity. , Write the new parity to disk.

Hereinafter, such parity update will be described in detail with reference to FIGS. The following is
The data update will be described. When the data is first written, the formatted data is normally "0", so "0" is stored as the old data and the old parity. By doing so, writing is possible under the same control as described below.

FIG. 2 shows an example of the configuration of the disk device 3 shown in FIG. Disk device 3 of FIG.
Each of 1 to 3-5 is assumed to have the configuration of FIG.

The disk device 3 includes a disk controller 40,
A disk 33 for storing data and an actuator 32 for reading / writing the disk 33 are provided.

The disk control unit 40 is an SC for controlling the SCSI bus 2 which is an interface with the disk control device 1.
SI bus control circuit 4, microcomputer 41 for controlling the entire disk device, hard disk controller 43 for controlling the actuator 32 and disk 33, read / write control circuit 44 for controlling read / write to the disk 33, temporary data It has a buffer memory 5 for temporarily storing it and an internal bus 45.

In this embodiment, as the buffer memory 5,
The write data buffer memory 52, the read data buffer memory 51, the exclusive OR data buffer memory 53, and the exclusive OR data transfer destination address 5
4 and disk devices 3-1 to 3-5
However, it is characterized in that they are connected by the same interface 2 as shown in FIG.

As a result, the write data is stored in the write data buffer memory 52, the read data is stored in the read data buffer memory 51, and the exclusive logic generated from the write data and the read data. Exclusive-OR data buffer memory 53 for sum data
It becomes possible to store the data in the same time. Further, it becomes possible to directly transfer the update position data generated by the disk device storing the data to the disk device storing the parity without going through the disk controller 1.

FIG. 3 shows the SCSI bus 2 of this embodiment.
The structure of the new command is shown. The new command instructs the disk control device 1 to the disk device 3 to read the old data and write the new data, and to update the parity from the disk device to the disk device storing the parity. This is a command instructing to issue a command to instruct.

Hereinafter, the command shown in FIG. 3 will be referred to as a DPW command. In this embodiment, in order to newly establish the DPW command, the group 6 command defined in the SCSI standard
The operation code C1H released to the vendor is used.

The DPW command includes an operation code of BYTE0, a logical block address of a disk device for writing BYTE2 to BYTE5, a write data length of BYTE6 to BYTE7, and an exclusive OR calculation result (update position data). The transfer destination address information is characterized in that it includes an ID number of a transfer destination disk device of BYTE 8 and a transfer destination logical block address for accessing BYTE 10 to BYTE 13.

As a result, the disk device 3 receiving this DPW command can recognize the transfer destination address of the update position data, and the old data read in the disk device 3 and the transferred data from the disk controller 1 are transferred. The exclusive OR with the new data is calculated, the calculation result (update position data) is transferred from the disk device 3 to the device at the transfer destination address (that is, the disk device storing the parity), and the update of the parity is instructed. can do.

FIG. 4 is a flow chart showing the operation of the disk device 3 which receives the DPW command. This flow chart is shown in FIG.
The DPW routine of the microcomputer 41 in the disk device 3 of FIG. Here, the disk device 3-1 of FIG.
The description will be given assuming that the command has been received.

When the disk device 3-1 receives the DPW command (step 801), the disk device 3-1.
The microcomputer 41 is a hard disk controller 43
Is instructed to read the old data already stored at the write address. Under the control of the microcomputer 41, the hard disk controller 43 starts seeking to the designated address in order to perform the read processing (step 802).

On the other hand, the SCSI bus control circuit 4 transfers the new data to the write data buffer memory 52 in the disk device 3-1 (step 803). After the transfer is complete
After releasing the SCSI bus 2 and regaining the SCSI bus 2, the disk device with the transfer destination ID number (here, the disk device 3
-5 is specified as the transfer destination), and an instruction to read the old parity of the transfer data length is issued from the transfer destination logical block address (step 804). Then, the SCSI bus 2 is released to wait for the old data to be read.

In step 804, the disk device 3
The command issued from -1 to the disk device 3-5 to instruct generation and writing of new parity is called a PW command. Actually, the reading of the old parity is instructed in step 804, and the update position data required for generating the new parity is transferred in step 811 described later. The processing of the disk device 3-5 which receives the PW command will be described later with reference to FIG.

The hard disk controller 43 waits for seek and rotation (step 806) and then reads old data (step 807). The read old data is stored in the read data buffer memory 51. After the read is completed, the rotation waits until the same address can be written (step 808), and then new data is written (step 809).

The microcomputer 41 generates updated position data by calculating the exclusive OR of the new data in the write data buffer memory 52 and the old data in the read data buffer memory 51 (step 81).
0). The updated position data is stored in the exclusive OR data buffer memory 53.

Further, under the control of the microcomputer 41, SCSI
The bus control circuit 4 uses the disk device 3-5 of the transfer destination ID number.
To the disk device 3-5 to transfer the updated position data, calculate the exclusive OR of the updated position data and the old parity, and write the new parity (step 811).

Further, the microcomputer 41 of the disk device 3-1 receives the completion report of the write of the new parity from the disk device 3-5, and then reports the completion to the disk controller 1.

Next, the disk device 3-1 issued
An example of a sequence for generating a new parity in the disk device 3-5 which receives the PW command will be described.

FIG. 5 is a flow chart of processing in the disk device 3-5 which receives the PW command. This flow chart shows that the PW routine of the microcomputer 41 in the disk device 3-5
7 shows a flow of operations performed by controlling each unit in the apparatus.

When the PW command issued in step 804 of FIG. 4 is received (step 901), the disk device 3
The microcomputer 41 of -5 is the hard disk controller 4
3 is instructed to read the old parity. Under the control of the microcomputer 41, the hard disk controller 43
Starts a seek to a designated address in order to perform a read process (step 902).

On the other hand, the SCSI bus control circuit 4 receives the update position data transferred from the disk device 3-1 in step 811, and stores the received update position data in the write data buffer memory 52. (Step 9
03). Then, it waits for the reading of the old parity (step 905).

The hard disk controller 43
Seeks and waits for rotation (step 904),
The old parity is read (step 906). The read old parity is the read data buffer memory 5
It is stored in 1.

Next, the microcomputer 41 generates new parity by calculating the exclusive OR of the update position data in the write data buffer 52 and the old parity in the read data buffer memory 51 (step 90
7). The new parity is stored in the exclusive OR data buffer memory 53. Next, after waiting for rotation until writing to the same address becomes possible (step 908),
The new parity is written (step 909).

Finally, the microcomputer 41 of the disk device 3-5 sends a completion report to the disk device 3-1. The end report may be directly reported to the disk controller 1.

As described above, the update position data is generated in the disk device 3-1 for storing the data, the update position data is transferred to the disk device 3-5 for storing the parity, and the update position data is stored in the disk device 3-5. Since it is possible to generate new parity with and write it to the disk, the old data,
It is not necessary to transfer the old parity and the updated position data to the disk controller 1.

Therefore, at the time of data update and parity update, old data and old parity are transferred to the disk controller, new parity is generated, and then new data and new parity are transferred from the disk controller to the disk device and written. Compared to the conventional method, the transfer amount of SCSI bus 2 is half (2 transfer of new data and update position data
Times). Further, since it is not necessary for the disk controller 1 to generate the updated position data, it is possible to prevent the processing capacity of the disk controller 1 from being lowered.

FIG. 6 is a block diagram showing another example of the interface structure of the array type memory device according to the present invention.

The array type storage device of this embodiment comprises a disk control device 1 having SCSI bus control circuits 4-1 and 4-2, and disk devices 3-1 to 3-2 accessible by two SCSI buses. 3-5 and SCSI buses 2-1 and 2-2 which are interfaces for connecting the disk control device 1 and the disk devices 3-1 to 3-5.

The present embodiment is characterized in that the array type storage device shown in FIG. 1 is provided with two SCSI buses. Other than that, the first of FIGS.
It is similar to the embodiment of. In this embodiment, since two SCSI buses are provided, the reliability of the interface can be improved and the transfer amount per interface can be reduced.

FIG. 7 is a block diagram of a hierarchical disk array device which is another example of the array storage device according to the present invention.

The array type storage device of this embodiment comprises a global disk controller 61, local disk controllers 63-1 to 63-5, and a common bus 62. The common bus 62 is the global disk controller 61.
And local disk control devices 63-1 to 63-5.

When a write command is issued from the host device, the global disk controller 61 divides the write data given by the host device into the storage units of the local disk controller, and the write command of the divided data is locally stored. Issued to the disk controller. In particular, the parity is generated from the divided data and stored in the local disk control device (here, the local disk control device 63-5) for storing the parity.

When a read command is issued from the host device, the global disk controller 61 combines the data read from each local disk controller and passes it to the host device. Since the local disk controller 63-5 stores the parity, the data can be recovered even if any of the local disk controllers fails.

In this way, the global disk controller 6
Reference numeral 1 manages data by RAID control to the local disk control devices 63-1 to 63-5.

Local Disk Controllers 63-1 to 63
Each of -5 includes a plurality of disk devices 3 therein, and controls division / combination processing of data to the respective disk devices 3. Any method may be used to control the plurality of disk devices 3 in the local disk control device, but RAID control may be performed, for example. In that case, RAID control is performed by the global disk control device 61 and the local disk control devices 63-1 to 63-5, and each local disk control device 63-
RAID control is also performed inside each of 1 to 63-5, and hierarchical RAID control is performed.

In this embodiment, the local disk control devices 63-1 to 63-5 use the read data, write data,
Buffer memories 65-1 to 65-6 for storing the exclusive OR data and the exclusive OR data transfer destination address
It is characterized by having 5-5 respectively. Also,
The local disk control devices 63-1 to 63-5 generate update position data and parity in the microcomputer 41.

Also in the case of the hierarchical disk array device shown in this embodiment, since each local disk control device is connected to the same interface by the common bus, the sequence shown in FIGS. 4 and 5 is used. , Old data and old parity to the global disk controller 61
Data and parity can be updated without having to be transferred to.

FIG. 8 shows another example of the interface configuration of the array type memory device according to the present invention. This is provided with an interface in the row direction in addition to the conventional interface configuration for connecting a disk device in the column direction.

In the interface configuration of this embodiment, in addition to the column-direction SCSI buses 2-1 to 2-5, the row-direction SCSI bus 2-for connecting disk devices in the same parity group is used.
6 is provided. The SCSI bus 2-6 is for transferring the PW command and the update position data in the above-mentioned first embodiment. As a result, the data and the parity can be updated without transferring the old data and the old parity to the disk controller 1.

Furthermore, when a failure occurs in the SCSI bus 2-1, it becomes possible to access the disk device 3-1 via the SCSI bus 2-2 and the disk device 3-2, for example. Therefore, it is possible to secure high reliability at a low cost as compared with the case where two systems of column-direction SCSI buses are provided.

FIG. 9 shows still another embodiment of the present invention, showing an example of a method of setting the ID numbers of the disk devices when the disk devices are arranged in a matrix of 3 rows and 5 columns. The ID number of # 1 in the disk device 3a-1 is "1".
Is shown.

This embodiment is characterized in that the ID number of each disk device is set so as not to overlap with the ID numbers of other disk devices connected to the two SCSI buses to which the disk device is connected. There is.

For example, the ID number of the disk device 3a-1 is set to "1", and the other disk devices 3a-2 to 3a-5 connected to the SCSI bus 2-6a are set to ID numbers other than "1". The other disk device 3b connected to the SCSI bus 2-1.
ID numbers other than "1" are set for -1 to 3c-1. As a result, the disk device 3a-1 has an ID number of "1" regardless of whether the SCSI bus 2-1 system or the SCSI bus 2-6a system.
become. That is, it is not necessary to set a plurality of ID numbers for each SCSI bus in a disk device connected to a plurality of SCSI buses.

FIG. 10 shows still another embodiment of the present invention and shows an example of a case where a disk array device is constituted by a disk device having a plurality of data access means.

The disk device 3 includes independently readable / writable actuators 32-1 and 32-2 and corresponding hard disk controllers 43-1 and 43-1.
43-2 read / write control circuits 44-1 and 44
-2 and. Others are the same as in the embodiment of FIG. The overall configuration is the same as in FIG.

Here, as an example, a case of updating the parity will be described. The update position data generated by the disk device 3-1 that stores data is transferred to the disk device 3-5 that stores parity, and the disk device 3
The data is stored in the write data buffer memory 52 within -5. Meanwhile, the old parity is read using the first actuator 32-1 and the old parity is stored in the read data buffer memory 51.

Further, a new parity is generated by calculating the exclusive OR of the updated position data and the old parity, and the new parity is stored in the exclusive OR data buffer memory 53. The time required to create a new parity is
It is on the order of microseconds, and the time required for the disk 33 to make one rotation is sufficiently short as compared with ten and several milliseconds. Therefore, after the old data is read by the first actuator 32-1, the second actuator 32 -2 is used to write the new parity.

As described above, after the old parity is read and before the new parity is written, the rotation waiting time for the disk 33 to wait for one rotation can be reduced.

[0086]

As described above, according to the present invention,
New data is transferred to a disk device that stores data, update position data is generated from the new data and old data in the disk device, and the updated position data is recovered from the disk device in correspondence with the data. Directly to the disk device that stores the redundant data (parity) for
Since the new parity is generated and then updated in the disk device that stores redundant data, interface data transfer is performed compared with the case where old data and old parity are transferred to the disk controller and new parity is generated. The amount can be halved to half, and the write penalty can be reduced. Further, as another effect, since the write penalty can be reduced with a small number of interfaces, there is an effect that an inexpensive array type disk device can be realized.

When the disk device for storing the data and the disk device for storing the redundant data for data recovery corresponding to the data are connected to the same interface, the update position data is transferred by using the interface. Can be transferred. Also, when applied to an array-type disk device having an interface configuration in which a row-direction interface is provided to the column-direction interface configuration of a conventional array-type disk device, it is necessary to transfer old data and old parity to the disk controller. Is eliminated, the write penalty is reduced.

Also, when the present invention is applied to the hierarchical disk array, it is not necessary to transfer the old data and the old parity to the highest-level disk controller, so that the write penalty can be reduced. .

Further, in the preferred embodiment of the present invention, the disk device for storing the data and the disk device for storing the redundant data corresponding to the data are connected by the interfaces of a plurality of systems. Even if a failure occurs, there is an effect that a highly reliable system can be realized without bringing down the entire system.

Further, since the ID number of each disk device does not overlap with the ID numbers of other disk devices connected to a plurality of systems, for example, SCSI buses to which the disk device is connected, each disk device is , Multiple IDs per SCSI bus
There is no need to set a number, and the cost can be reduced.

Further, in the disk device, independent reading,
A plurality of writable data access means are provided, after the old data is read by the first data access means, new data is created in the disk device, and the new data is written by using the second data access means. Because
After reading the old data, it is not necessary to wait for the disk to make one revolution, and the write penalty is reduced.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 2 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 3 is a configuration diagram showing a configuration example of a SCSI command according to the present invention.

FIG. 4 is a flowchart showing an operation example of an array disk device according to the present invention.

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

FIG. 6 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 7 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 8 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 9 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 10 is a configuration diagram showing a configuration example of an array type storage device according to the present invention.

FIG. 11 is a configuration diagram showing a configuration of a conventional array type storage device.

FIG. 12 is a flowchart showing the operation of a conventional array disk device.

[Explanation of symbols]

1 ... Disk control device 2, 2-1, 2-2, 2-3, 2-4, 2-5, 2-
6, 2a-6, 2b-6, 2c-6, ... SCSI bus 3, 3-1, 3-2, 3-3, 3-4, 3-5, 3a-
1, 3a-2, 3a-3, 3a-4, 3a-5, 3b-
1, 3b-2, 3b-3, 3b-4, 3b-5, 3c-
1, 3c-2, 3c-3, 3c-4, 3c-5, ...
Disk device 4, 4-1, 4-2, 4-3, 4-4, 4-5 ... SC
SI bus control circuit 5 ... Buffer memory 11, 45 ... Internal bus 12, 41 ... Microcomputer 14 ... Data division / synthesis control unit 32, 32-1, 32-2 ... Actuator 33 ... Disk 40 ... Disk control unit 43, 43-1, 43-2 ... Hard disk controller 44, 44-1, 44-2 ... Read / write control circuit 51 ... Read data buffer memory 52・ ・ ・ Write data buffer memory 53 ・ ・ ・ Exclusive OR data memory 54 ・ ・ ・ Data transfer destination address buffer memory 61 ・ ・ ・ Global disk control device 62 ・ ・ ・ Common bus 63-1, 63-5・ ・ ・ Local disk controller 301 ・ ・ ・ SCSI command

Claims (12)

[Claims] 1. A plurality of disk devices each having a disk for storing data therein, and data sent from a host device are divided, and redundant data for data recovery is generated from the divided data, In an array type storage device having a disk controller that stores the divided data and the redundant data in the plurality of disk devices, the disk device is a redundant device that corresponds to the data stored in the disks in the own disk device. Storage means for storing data storage position information, and when the disk controller gives an instruction to update the data, the old data to be updated is read from the disk, and the read old data and the disk controller are provided. Update position data generating means for generating update position data representing a changed portion in the old data using the new data, A new data writing unit that updates the old data by writing new data to the disc; and a transfer unit that directly transfers the updated position data to a disk device that stores redundant data corresponding to the data, The disk device storing the redundant data of the transfer destination of the update position data reads the old redundant data to be updated from the disk and generates new redundant data using the read old redundant data and the transferred update position data. An array type storage device comprising: means and new redundant data writing means for updating the old redundant data by writing the generated new redundant data to a disk. 2. A plurality of disk devices each having a disk for storing data therein and an RA for the disk devices.
An array type storage device comprising a disk control device controlled according to an ID, wherein the disk device stores storage position information of redundant data for data recovery corresponding to data stored in a disk in the own disk device. And when the disk controller instructs to update the data, the old data to be updated is read from the disk, and the updated position data is obtained using the read old data and the new data given from the disk controller. Update position data generating means for generating, new data writing means for updating the old data by writing new data to the disk, and the updated position data directly to a disk device for storing redundant data corresponding to the data. Transfer means for transferring, storing redundant data of the transfer destination of the update position data. The disk device reads the old redundant data to be updated from the disk, generates new redundant data using the read old redundant data and the transferred update position data, and writes the generated new redundant data to the disk. An array-type storage device comprising: new redundant data writing means for updating the old redundant data. 3. The array type storage device according to claim 1, wherein the redundant data for data recovery is parity data which is an exclusive OR of the divided data, and the update position data is An array type storage device, which is update position data which is an exclusive OR of the old data and the new data. 4. The array type storage device according to claim 1, wherein the same interface is used to connect the plurality of disk devices and the disk control device and to connect the plurality of disk devices to each other. The array-type storage device, wherein the update position data is directly transferred via the interface. 5. The array type storage device according to claim 4, wherein the interface is provided with a plurality of systems. 6. The array type storage device according to claim 1, wherein the plurality of disk devices and the disk control device are connected by an interface provided individually for each of the plurality of disk devices. In addition to the interface, the array type storage device further comprises an interface for connecting the plurality of disk devices to each other in order to transfer the update position data. 7. The array type storage device according to claim 6, wherein a plurality of interfaces are provided to connect the plurality of disk devices to each other. 8. The array type storage device according to claim 1, wherein a plurality of systems of interfaces are provided between the plurality of disk devices and the disk control device.
A plurality of disk devices are connected to each system, and in addition to the interface, a plurality of systems of interfaces for connecting the plurality of disk devices to each other for transferring the update position data are provided. A characteristic array type storage device. 9. The array type storage device according to claim 8, wherein as the device address of the disk device, in any of the interfaces of a plurality of systems to which the disk device is connected, all other devices connected to the interface are connected. An array type storage device characterized by setting one device address which does not overlap with the device address of the disk device. 10. The array type storage device according to claim 1 or 2, wherein the disk device itself is a second array type storage device including a plurality of disk devices and a disk control device for controlling them. An array type storage device characterized by being present. 11. The array type storage device according to claim 1, wherein the disk device has at least a readable first data access unit and at least a writable first data access unit with respect to a disk in the own disk device. The second data access means is provided, and the old redundant data is read by the first data access means until the second data access means can write to the same address. An array type storage device characterized in that a calculation for generating new redundant data is performed using the old redundant data and the updated position data, and the new redundant data is written by the second data access means. 12. The array type storage device according to claim 1, wherein the disk device includes write data temporary storage means for temporarily storing new data given from the disk control device, and The read data temporary storage means for temporarily storing the old data read from the disk and the update position data temporary storage means for temporarily storing the generated update position data are provided. The present invention is characterized in that storage, storage of old data in read data temporary storage means, and storage of update position data generated using old data and new data in update position data temporary storage means are performed simultaneously in parallel. Array type storage device.