JPH09265357A - Disk array device - Google Patents

Abstract

(57) Abstract: There is provided a disk array device which eliminates variations in reading speed and efficiently performs reading processing. SOLUTION: A CPU 1 is provided for A to D disk devices 20 to 23 connected to a SCSI interface 18.
For 0, the time until the access end signal is returned after the signal is transmitted is measured using the clock inside the CPU 10, the D disk device 23 with the slowest access speed is assigned to the parity disk, and the remaining 3 units are data. Assign to disk.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to RAID, for example.
(Redundant Arrays of Inexpensive Disks) 3 and RA
As typified by ID4 and the like, a plurality of data disks and one parity disk are provided, and if an error occurs when reading data from the data disk, it should be read from the remaining data disks and parity disks. Disk array device capable of generating the above data.

[0002]

2. Description of the Related Art
A disk array has been proposed as a storage device that enables high-speed access and guarantees high reliability. The disk array is systematized as a method to counter a large-sized disk device by including a plurality of small-sized disk devices such as a cache memory and a hard disk device. This systematized method is called RAID, and RAID 1 to RAID 5 are considered as the basic configuration. Here, RAID3 and RAI related to the present invention are described.
The D4 disk array will be briefly described.

RAID 3 and RAID 4 are composed of a plurality of data disks and one parity disk. The data is divided and recorded on a plurality of data discs. Also, the parity data corresponding to the data group is calculated by calculating the corresponding data between the data disks and recorded in the parity disk. As a result, if a read error occurs when reading data from one data disk, an error occurs from the data read from the remaining data disks in which no error has occurred and the parity data read from the parity disk. It is possible to generate data that should be read from the data disc. Furthermore, if one disk fails according to the same principle, the contents of the failed disk can be restored from all the remaining data disks and parity disks after the disk is replaced.

In RAID 3, recorded data is divided into 1-byte units. Therefore, in RAID3, it is necessary to access all the drives both when reading data and when writing data. Also, RAID3
Then, the parity data is always read when the data is read. If no read error occurs in any data disk, the read parity data is not used. If a read error occurs in a certain data disk, the data in which the error occurred by using the respective data read from the remaining data disks in which no error has occurred and the parity data read from the parity disk Generate the data that should be read from disk. Then, when the data from all the data disks are gathered, these data are combined and output as one data.

In the above method in which the occurrence of an error is detected and the data in which the error is generated is generated, the reading speed is lowered depending on the time until the error is detected. Therefore, when the parity data reading is completed without waiting for error detection and there is only one piece of data that has not been read from the data disk, the data from the read data disk and the parity disk are read. There is also proposed a method of generating the above-mentioned one piece of unread data by using the read parity data.

On the other hand, in RAID 4, data division is performed in units of a plurality of bytes. In dividing into a plurality of bytes, it is possible to divide into any number of bytes such as 10 bytes, 50 bytes, 100 bytes or 1 kilobyte.
Dividing data into multiple bytes is called block division. By dividing into blocks, it is not necessary to access all disks. That is, only the data disk on which the data requested to be read is recorded may be accessed. This indicates that even if the read requests occur at the same time, the data recorded in another data disk can be simultaneously read. That is, RAID 4 can perform parallel processing for reading.

In RAID 4, when an error occurs in the read data disk, all data disks except the disk and the parity disk are accessed for the first time and the corresponding data is read to generate the target data. Therefore, compared to RAID3, which always reads the parity disk when reading the data and can immediately generate the targeted data at the time of error, the targeted data generation becomes slower.

By the way, when configuring RAID 3 and RAID 4, one of the plurality of disk devices is appropriately assigned as a parity disk and the rest as data disks. As described above, even when the disk devices are properly assigned to the parity disk and the data disk, there is no problem if there is almost no difference in access speed between the disk devices.

However, if there is a difference in access speed between disk devices due to differences in manufacturers, the read speed may vary. For example,
In a disk array in which data is frequently read, when a disk device with a slow access speed is assigned to the data disk, the read speed is significantly slower than when a disk device with a fast access speed is assigned to the data disk. .

The present invention comprises a disk array device represented by RAID 3 or RAID 4, that is, a plurality of data disks and one parity disk, and when an error occurs during data reading from the data disk, the remaining In a disk array device capable of generating data that should be read from the errored data disk from the data disk and the parity disk of
It is an object of the present invention to provide a disk array device that eliminates the above-mentioned variations in reading speed and efficiently performs reading processing.

[0011]

Means for Solving the Problems and Effects of the Invention
In the case where the disk array device described in (3) has at least three disk devices, one of the disk devices is used as a parity disk and the other is used as a data disk to read data from the data disk. When a read error occurs in the disk, a disk that can generate the data to be read from the data disk in which the read error occurred based on the data read from the remaining error-free data disk and the parity disk In the array device, the disk device having the slowest access speed among the plurality of disk devices is assigned to the parity disk, and the remaining disk devices are assigned to the data disks.

For example, the allocation of the disk device is
As will be described later, it can be considered to be performed by the disk array device itself, or can be performed by the user. When the user performs it, the access speeds of all the disk devices are measured one by one. Then, if the parity disk can be allocated in a hardware connection order, etc., reconnect, and if the parity disk can be allocated in a soft memory table, etc., perform an operation based on it. Good.

Conventionally, the parity disk has been appropriately allocated among a plurality of disk devices. However, there may be a difference in access speed between the disk devices due to differences in manufacturers. In this case, for example, when a disk array that frequently performs read processing is assigned as a data disk having an access speed slower than that of other disk devices, the read speed is significantly reduced.

On the other hand, in this disk array device, one of the plurality of disk devices having the slowest access speed is assigned to the parity disk, and the rest is assigned to the data disk. As a result, the most efficient allocation of the parity disk and the data disk can be realized in the disk array device that repeatedly reads the data. Data reading using disk devices having different access speeds will be described in detail below with reference to FIG. 4 by taking the above-mentioned RAID3 and RAID4 as an example.

FIG. 4 shows the access time of each disk device when reading one data from the RAID 3 disk array device. The horizontal axis of FIG. 4 indicates time. This disk array has four disk devices A to D. In the description here, in order to distinguish the four disk devices, they are described as A disk device, B disk device, C disk device, and D disk device.
In RAID3, when accessing one data, A
Access all disk devices of ~ D. FIG. 4 shows a case where data is read simultaneously from the four disk devices A to D, and the hatched portion in the drawing is the read time. Here, the read time is A → B → C →
It is assumed that the disk devices of D are delayed in order. It should be noted that the data generation time indicates the time for generating data from a data disk that cannot be read based on the data read from the parity disk and the data disk. The portion indicated by a white square in the figure is the data generation time.

(1) Consider reading from a RAID 3 disk array. In RAID 3, after detecting a read error in one data disk, each data read from the remaining data disks and parity data read from the parity disk are used to read from the data disk in error. Generate the data that should be. However, in the above method, it takes time to detect an error in RAID3, and the reading speed decreases, so that the reading of data from the parity disk ends, and the number of data disks that have not finished reading becomes one. In some cases, the unread data is generated by the parity data and the data from the other disk when it becomes.

First, even if the reading of the parity disk is completed before the reading of the data disk, the data from the unread data disk is not generated.
Consider AID3. Consider a case where a D disk device having a slow access speed is assigned to a parity disk as described in claim 1. Considering the reading of certain data, reading of all data is completed at time t1 in FIG. 4 at which reading of data from the C disk device having the slowest access speed in the data disk is completed.

On the other hand, consider the case where the C disk device is assigned to the parity disk. Considering the reading of certain data, reading of all data is completed at time t2 in FIG. 4 at which reading of data from the D disk device having the slowest access speed in the data disk is completed. Here, even if the A or B disk device is assigned to the parity disk, the time t2 in FIG. 4 at which the reading from the D disk device is completed is similar to the above.
The data reading is completed with. That is, when the parity disk is assigned to a device other than the D disk device having the slowest access speed, an extra time is required by the time α which is the difference between the times t1 and t2 in FIG.

Next, when the reading of the parity disk is completed before the reading of the data disk, and when the number of the data disks for which the reading is not completed becomes one, the unread data is generated. RAID
Think about 3. Consider a case where the D disk device having the slowest access speed is assigned to the parity disk as described in claim 1. When considering reading of certain data, the parity data is used because the reading of the data from the C disk device is completed before the reading of the parity data from the D disk device having the slowest access speed is completed. No data is generated. Therefore, reading of data is completed at time t1 in FIG.

On the other hand, for example, consider the case where the C disk device is assigned to the parity disk. Considering the reading of certain data, when the data from the A disk device and the B disk device and the parity data from the C disk device are read (time t1 in FIG. 4), the data reading from the D disk device is performed. It operates to generate data that should be read from the D disk device without waiting. This state is shown by a white square in FIG. In this way, the data to be read from the D disk device is generated without waiting for the latest D disk device to read, and the data reading is completed at time t3 in FIG.
At this time, an extra time is required by the time β which is the difference between the times t1 and t3 in FIG. Even when a parity disk is assigned to the A or B disk device, data generation is started at the time when the reading of the C disk device is completed (time t1 in FIG. 4). Is completed, and an extra time is required by the amount of time β in FIG.

By the way, it is possible that an error occurs during reading from the data disk. in this case,
If the D disk unit is assigned to the parity disk, D
When the reading of the parity data from the disk device is completed (time t2 in FIG. 4), the error data is generated. Therefore, it takes more time than when other A to C disk devices are assigned to the parity disk. However, there is no problem because the number of times of error is usually extremely smaller than the number of times of normal reading.

As described above, in RAID3, by setting the disk device with the slowest access speed to the parity disk, it is possible to minimize the time required for one read, on average. For example, the effect becomes remarkable when it is used for RAID 3 which is frequently read.

(2) Consider reading from a RAID 4 disk array. In RAID4, when reading certain data, reading is performed only from the data disk on which the data is recorded. Further, the data disk and parity disk other than that are not accessed except when an error occurs.

For example, it is assumed that RAID 4 is composed of four disk devices A to D having different access speeds as shown in FIG. Consider a case where the D disk device having the slowest access speed is assigned to the parity disk as described in claim 1. When reading certain data, at least one of the three disk devices A to C, which are data disks, is accessed. And AC
Even if one of the disk devices is accessed,
Even if a plurality of units are accessed at the same time, the reading is completed before the time t1 in FIG.

On the other hand, consider the case where the C disk device is assigned to the parity disk. When reading certain data, at least one of the A, B, and D disk devices is accessed. And one of A, B, D disk device
Considering the case where a plurality of units are accessed and the case where a plurality of units are simultaneously accessed, the reading is completed before the time t2 in FIG.

Considering that both the RAID 4 in which the D disk device is assigned to the parity disk and the RAID 4 in which the C disk device is assigned to the parity disk are used to read and use data frequently, three data disks are equivalent. It is considered to be accessed. Therefore, there is no difference in the time for accessing the A and B disk devices, but when reading data from the C and D disk devices, the D disk device requires extra time by the time α in FIG. 4 for each access. If the three data disks are accessed to the same extent, the read speed becomes faster when the D disk device with the slowest access speed is assigned to the parity disk.

Now, consider the case where an error occurs in reading in RAID4. In RAID4, when a read error occurs in the data disk, the data necessary for data generation is read from the parity disk and the remaining data disks for the first time. That is, even if there is a difference in access speed between disk devices, there is no difference in read speed when an error occurs. Therefore, it suffices to consider a setting that maximizes the read speed in a state where no error occurs.

As described above, by setting the disk device with the slowest access speed to the parity disk even in RAID4, it is possible to minimize the read time on average. From the consideration of (1) and (2) above, setting the parity disk to the disk device with the slowest access speed in either RAID3 or RAID4 minimizes the read time on average. Can be
Especially, a disk array specialized for reading has a great effect.

Although RAID 3 and RAID 4 have been considered here, when an error occurs during data reading from a data disk, the operation of generating data that should be read from the remaining data disks and parity disks is performed. The same effect can be obtained with a disk array device.

According to a second aspect of the present invention, there is provided a disk array device, based on access speed measuring means for measuring an access speed of the disk device and access speeds of the respective disk devices measured by the access speed measuring means. A disk allocating means for allocating the disk device having the slowest access speed to the parity disk and allocating the remaining disk devices to the data disk is provided.

In this case, the access speed of each connected disk can be measured by the access speed measuring means. Then, based on the measured access speed of each disk device, the disk allocation means allocates the disk device with the slowest access speed to the parity disk and the remaining disk devices to the data disks. It is sufficient to perform this processing once when all disks are initialized, such as when a new disk array is connected or when a disk is replaced due to a failure or the like.

Thus, the access speeds of all the disk devices to which the disk array device itself is connected are measured, the disk device with the slowest access speed is assigned to the parity disk based on the result, and the remaining disk devices are assigned to the data disks. As described above, there is no need for the user to measure the disk access speed one by one and determine the disk device allocation as described above.

[0033]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an electrical configuration when the present disk array device is applied to RAID3 or RIAD4.

The disk array device has a CPU 10 as a control means, a RAM 12, and an R as a storage means.
OM 14, data transfer interface 16, SC
SI interface 18 and four disk units 2
0, 21, 22, 23 are provided. In the following description, in order to distinguish the four disk devices, the A disk device 20, the B disk device 21, and the C disk device 2
2 and D disk device 23. Each of the disk devices 20 to 23 is a so-called physical hard disk drive and a control board for controlling the same integrated.

The CPU 10 uses the A to D disk devices 2
The entire disk array device is controlled by instructing 0 to 23 to read and write data. The CPU 10 corresponds to "access speed measuring means" and "disk allocation means".

The ROM 14 stores in advance a program started by the CPU 10. RAM
12 operates as a cache memory for temporarily recording data read from the A to D disk devices 20 to 23, data to be written, and the like, and when a program previously stored in the ROM 14 is executed, the program is stored. It also works as a work area for executing. Further, the RAM 12 is provided with an allocation table 12a for allocating the disk devices 20 to 23 of A to D as parity disks or data disks.

Further, the data transfer interface 16
Performs data read / write signal transfer with a main computer connected to the outside, and read / write data itself transfer. On the other hand, the SCSI interface 18 is
~ D disk devices 20 to 23 are connected.

Next, the allocation table 12a will be described with reference to FIG. FIG. 2 shows the allocation table 12
It is an explanatory view of a, and the upper part shows three data disks and one parity disk. Data discs are D1-D
At 3, the parity disk is shown at P. The lower part shows A to D disk devices 20 to 23 assigned to the three data disks and one parity disk in the upper part.
It is described as A to D in the figure. FIG. 2 shows that the A to C disk devices 20 to 22 are assigned to the data disks and the D disk device 23 is assigned to the parity disks. For example, the allocation of the A to D disk devices 20 to 23 is changed by rewriting the order of A to D in the lower stage.

Next, based on the flow chart of FIG. 3, the disk array device itself has four disk devices 2 of A to D.
A process of assigning 0 to 23 to one parity disk and three data disks will be described. This process is performed by, for example, the A to D disk devices 20 to 2 which were connected to the disk array device when a new disk array was constructed.
It is set to be performed when all the disk devices 20 to 23 of A to D are initialized, for example, when 3 is replaced due to a failure.

In this process, in step S100, the access speed of each of the four A to D disk devices 20 to 23 connected to the SCSI interface 18 is measured. The measurement is performed by, for example, four A to D disk devices 20 connected to the SCSI interface 18.
To 23 are sequentially accessed, and the time from when the CPU 10 outputs a signal to when each of the disk devices 20 to 23 of A to D returns an access end signal is measured by using a clock in the CPU 10. . Here S1
As a result of the measurement of 00, for example, assuming that the access speed becomes slower in the order of A → B → C → D disk devices 20 to 23, the following description will be continued.

Next, in S110, the D disk device 23 having the slowest access speed is assigned to the parity disk. Then, in S120, the remaining three A to C disk devices 20 to 22 are assigned to the data disks. Four A to D in S110 and S120
The allocation of the disk devices 20 to 23 to the parity disk or the data disk is performed by, for example, the RAM 12 described above.
The allocation table 12a for deciding the allocation of the four A to D disk devices 20 to 23 is stored in the storage device (see FIG. 2). If there is no difference in speed among the A to D disk devices 20 to 23 whose access speeds have been measured, an appropriate one is allocated as a parity disk and the rest are allocated as data disks.

As a result, the D disk device 23 having the slowest access speed is assigned as the parity disk. Then, the remaining three A to C disk devices 20 to 22 are allocated to the data disks. At this time, the fact that the reading efficiency is the highest will be specifically described with reference to FIG.

As a first specific example, consider the case where the disk array device of this embodiment constitutes RAID3. As described in the above description of the prior art, in RAID 3, recorded data is divided into 1-byte units. Therefore, in RAID3, it is necessary to access all the drives both when reading data and when writing data.

FIG. 4 shows access times of the A to D disk devices 20 to 23 when reading one data. Now, the four disk devices 20 to 23 of the above A to D
The data is simultaneously read out from, and the shaded portion in the figure is the read time, and the horizontal axis is the time. It should be noted that the data generation time indicates the time for generating data from a data disk that cannot be read based on the data read from the parity disk and the data disk. The portion indicated by a white square in the figure is the data generation time.

In RAID 3, after a read error of one data disk is detected, the error data is obtained by using the respective data read from the remaining data disks and the parity data read from the parity disk. Generate the data that should be read from disk.
However, in the above method, it takes time to detect an error in RAID3, and the reading speed decreases, so that the reading of data from the parity disk ends, and the number of data disks that have not finished reading becomes one. In some cases, the unread data is generated by the parity data and the data from the other disk when it becomes.

First, consider RAID 3 in which even if the reading of the parity disk is completed before the reading of the data disk, data is not generated from the unread data disk. Consider the case where the D disk device 23 shown in FIG. 4, which has the slowest access speed, is assigned to the parity disk as in the disk array device of this embodiment. At this time, the remaining A to C disk devices 20 to 22 are assigned to the data disks. When reading data, the A to C disk devices 20 to 22
The divided data is read from. Then, the time t when the reading of the data from the C disk device 22 is completed.
In 1 all data is read and each data is combined 1
It becomes one data.

On the other hand, let us consider, for example, allocating the C disk device 22 which is second from the slowest access speed to the parity disk. At this time,
The A, B and D disk devices 20, 21 and 23 are assigned to the data disks. If you read some data,
At time t2 when the reading of the data from the D disk device 23 is completed, the reading of all the data is completed. Here, even if it is considered that the A or B disk device 21 is assigned to the parity disk, the D disk device 23 similarly
The reading of data is completed at time t2 when the reading of the data is finished. That is, D, which has the slowest access speed
When a parity disk other than the disk device 23 is set, an extra time is required by the time α which is the difference between the times t1 and t2 shown in FIG.

Next, when the reading of the parity disk is completed prior to the reading of the data disk, and when the number of the data disks for which the reading is not completed becomes one, the unread data is generated. RAID
Think about 3. Like the disk array device of this embodiment, the D disk device 2 with the slowest access speed
Consider the case where 3 is assigned to the parity disk. At this time, the A to C disk devices 20 to 22 are assigned to the data disks. When reading data, the data including the parity data is read from all the disk devices 20 to 23 of A to D, but the C disk is read before the reading of the parity data from the D disk device 23 is completed. In this case, no data is generated using the parity data because the reading of the data from the device 22 is completed. Therefore, at the time t1 shown in FIG. 4, the reading of the data is completed, and the data are combined to form one data.

On the other hand, for example, the second C disk device 22 from the disk device having the slowest access speed is assigned to the parity disk. At this time, the data disks are assigned to the A, B, and D disk devices 20, 21, and 23. When the data from the A disk device 20 and the B disk device 21 and the parity data from the C disk device 22 are read (time t1 in FIG. 4), the D disk device 23 does not wait for the data to be read from the D disk device 23. It operates to generate the data that should be read from 23. This state is shown by a white square in FIG.
In this way, the data to be read from the D disk device 23 is generated without waiting for the slowest read from the D disk device 23, and the data reading is completed at time t3 in FIG. At this time, the times t1 and t in FIG.
An extra time is required for the time β which is the difference of 3. The parity disk is A or B disk device 20, 2
Even when it is set to 1, data reading is completed at time t1 in FIG. 4 when the reading of the C disk device 22 is completed, so that data reading is completed at time t3 in FIG. An extra time is required for the time β.

By the way, it is possible that an error occurs during reading from the data disk. in this case,
When the D disk device 23 is set to the parity disk as in the present embodiment, the time when the reading of the parity data from the D disk device 23 is completed (time t2 in FIG. 4).
The error data is generated from. Therefore, it takes more time than when the other A to C disk devices 20 to 22 are set as the parity disks. However, there is no problem because the number of times of error is usually extremely smaller than the number of times of normal reading.

As described above, when the disk array device of this embodiment configures RAID 3, by setting the D disk device 23 with the slowest access speed, on average, reading is performed once. This time can be minimized. For example, in the case of RAID3 in which reading is frequently performed, the effect becomes remarkable.

As a second specific example, let us consider the reading when the disk array device of this embodiment constitutes RAID4. In RAID4, when reading certain data, reading is performed only from the data disk on which the data is recorded. Further, the data disk and parity disk other than that are not accessed except when an error occurs.

For example, as shown in FIG. 4, RA is used in four disk devices 20 to 23 having different access speeds A to D.
It is assumed that ID4 is configured. Consider a case where the D disk device 23 having the slowest access speed is a parity disk as in this embodiment. In this case, the data disk is assigned to the A to C disk devices 20 to 22. A to C disk device 20 for reading data
At least one of .about.22 is accessed. And A
Even when one of the C disk devices 20 to 22 is accessed, or when a plurality of C disk devices are simultaneously accessed, the reading is completed before the time t1 in FIG.

On the other hand, let us consider, for example, a case where the C disk device 22 which is second from the disk device having the slowest access speed is set as the parity disk. In this case, the A, B and D disk devices 20, 21 and 23 including the D disk device 23 having the slowest access speed are assigned to the data disks. When reading data, at least one data disk of the A, B, and D disk devices 20, 21, and 23 is accessed. And A,
Even when one of the B, D disk devices 20, 21, and 23 is accessed, or when a plurality of devices are simultaneously accessed, the reading is completed before the time t2 in FIG. .

Considering that both the RAID 4 in which the D disk device 23 is assigned to the parity disk and the RAID 4 in which the C disk device 22 is assigned to the parity disk are both frequently read and used, 3
It is considered that the data disks of each unit are accessed to the same extent. Therefore, although there is no difference in the time for accessing the A and B disk devices 20 and 21, both are C disk devices 22 and D.
When reading data from the disk device 23, the D disk device 2 is used for the time α shown in FIG. 4 for each access.
3 takes extra reading time. If the three data disks are accessed to the same extent, the read speed becomes faster when the D disk device 23 having the slowest access speed is set as the parity disk.

Now, consider the case where an error occurs in reading in RAID4. In RAID4, when a read error occurs in the data disk, the data necessary for data generation is read from the parity disk and the remaining data disks for the first time. That is, A to D disk devices 20 to 2
Even if there is a difference in access speed in No. 3, if an error occurs, there is no difference in read speed. Therefore, it suffices to consider a setting that maximizes the read speed in a state where no error occurs.

As described above, even if the present embodiment is a disk array device that configures RAID 4, by setting the disk device D disk device 23 with the slowest access speed to the parity disk, on average, the read It is possible to minimize the time required for

As discussed above and in the above, the disk array device of the present embodiment is operated by RAID3 or RAID4.
In either case, by allocating the D disk device 23, which has the slowest access speed, to the parity disk, it is possible to minimize the time required for reading, on average, and especially for reading. If it is a specialized disk array, it has a great effect.

Further, in this embodiment, all the disk devices 20 to 20 of A to D to which the disk array device itself is connected.
The access speeds of the disk units 20 to 23 are measured by measuring the access speeds of the disk units 23 to 23, and the disk units 20 to 23 of the units A to D are automatically allocated by the user.
No need to change the connections of the A to D disk devices 20 to 23 by measuring the access speeds of No. 3 and changing their connections or rewriting the contents of the allocation table 12a of the SCSI interface 18. Become.

As described above, the present invention is not limited to such an embodiment at all, and can be implemented in various forms without departing from the gist of the present invention. For example, in the above-described embodiment, the disk array device itself uses the A to D disk devices 20 to 2 for the parity disk and the data disk.
However, as described above, the access speeds of all the disk devices 20 to 23 of A to D to which the user is connected are measured, and the disk devices 20 to 2 of A to D are allocated.
It is also conceivable to assign 3 to the parity disk and the data disk. As a method of allocating the A to D disk devices 20 to 23, it is conceivable to change the connection of the A to D disk devices by hardware, or to rewrite the contents of the allocation table 12a by software.

Further, in the above embodiment, so-called hard disk devices are used as the A to D disk devices 20 to 23, but it is conceivable that the disk devices of other forms such as silicon disks are used.

[Brief description of drawings]

FIG. 1 is a block diagram showing an electrical configuration of a disk array device.

FIG. 2 is an explanatory diagram of a disk device allocation table in a RAM.

FIG. 3 is a flowchart showing a process of assigning a plurality of disk devices to a parity disk and a data disk.

FIG. 4 is a time chart showing an access time when reading data from a disk device.

[Explanation of symbols]

10 ... CPU 12 ... RAM 14 ... ROM 16 ... Data transfer interface 18 ... SCSI interface 12a ... Allocation table 20 ... A hard disk device 22 ... B hard disk device 24 ... C hard disk device 26 ... D hard disk device

Claims (2)

[Claims] 1. A read error when data is read from a data disk using at least three disk devices, one of which is used as a parity disk and the other is used as a data disk. In the disk array device capable of generating the data to be read from the data disk in which the read error has occurred, based on the data read from the remaining error-free data disk and the parity disk. A disk array device, wherein a disk device having the slowest access speed among the plurality of disk devices is assigned to the parity disk, and the remaining disk devices are assigned to the data disks. 2. An access speed measuring means for measuring an access speed of the disk device, and a disk device having the slowest access speed is assigned to a parity disk based on the access speed of each disk device measured by the access speed measuring means. 2. The disk array device according to claim 1, further comprising disk allocation means for allocating the remaining disk devices to data disks.