JP3615250B2 - Disk array device - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a disk array apparatus in which access time is shortened in a redundant structure parallel distributed disk array.
[0002]
[Prior art]
With the recent advancement of semiconductor technology and the adoption of a computer architecture such as a pipeline system, the performance and reliability of each component of a computer system have been dramatically improved. On the other hand, the improvement in performance and reliability of the secondary storage device centering on the mechanism device does not reach the other components, and has become a bottleneck of the processing capability of the entire computer system.
Incidentally, a disk array of RAID (Redundant Arrays of Inexpensive Disks) architecture has been proposed in order to improve the performance and reliability of secondary storage at a realistic cost. Such a disk array is composed of a plurality of secondary storage devices, and divides and stores data in a plurality of devices. The data transfer is speeded up by performing access in parallel. On the other hand, parity is stored in one secondary storage device so that any one of the secondary storage devices can be dealt with.
[0003]
[Problems to be solved by the invention]
However, the conventional techniques described above have the following problems.
That is, a hard disk drive generally used as a secondary storage device is accompanied by a rotation waiting time when accessed. For example, when four devices are used, a rotation waiting time for four devices occurs. In this case, in an expensive system, the spindle synchronization function can synchronize the rotation of each disk to shorten the waiting time, but many small and inexpensive popular products do not have such a spindle synchronization function.
Therefore, for example, when waiting for rotation of four data disks at the time of reading, the rotation waiting time increases as compared to access to a single hard disk. Further, if there are four data disks at the time of writing, it is necessary to write the parity to the parity disk, so it is necessary to wait for the five disks to rotate, and the rotation waiting time further increases. As a result of experiments using simulations, it has been found that the ratio of these increase times is about 60% at the time of reading and about 66% at the time of writing.
[0004]
[Means for Solving the Problems]
The disk array apparatus of the present invention is characterized by the following points in order to solve the above-described problems.
(1)
1) A data read request issuing unit that issues a data read request to all of the plurality of data holding storage devices and the parity holding storage device when a data read request is made.
2) After the data read request is issued by the data read request issuing unit, when the data read operation of all the storage devices except the one with the slowest data read operation is completed, the operation is not yet completed. A determination unit that determines whether the device is a data holding storage device or a parity holding storage device is provided.
3) When the operation of the parity holding storage device is delayed according to the determination result of the determination unit, the data from the plurality of data holding storage devices are combined and transferred, while the data holding storage device When the operation is delayed, the data read operation is completed while reproducing the data missing from the plurality of other data holding storage devices and the parity from the parity holding storage device. A data transfer unit for combining and transferring data is provided.
[0005]
(2)
1) A data buffer provided corresponding to each of a plurality of data holding storage devices and parity holding storage devices is provided.
2) A data write request issuing unit that issues a data write request to all of the plurality of data holding storage devices and the parity holding storage device when a data write request is made.
3) Immediately after the data write request is issued by the data write request issuing unit, the divided data and parity are transferred to a plurality of data holding storage devices and parity holding storage devices, and these are also transferred to the data buffer. A transfer unit is provided.
4) After the data transfer by the data transfer unit, when the data write operation of all the storage devices except the one with the slowest data write operation is completed, the fact that the write operation is completed is reported to the host device and A write report unit is provided that stores at least the identifier of one of the removed storage devices in a nonvolatile memory.
5) After the completion of the data write by the write report unit, the data in the data buffer corresponding to the storage device that has already completed the data write operation is discarded, and the data write operation is also completed for the remaining one storage device. In some cases, a post-processing unit that discards the data in the corresponding data buffer is provided.
6) When the data in the data buffer corresponding to the remaining one storage device is lost, the lost data is recovered by the data from the storage devices other than the remaining one storage device, and the nonvolatile memory The restored data is written to the remaining one storage device by referring to the memory.
[0006]
[Action]
(1)
At the time of reading, for example, if access to any four of the five storage devices is completed, this is determined by the determination unit, and if the data is missing, the data lost in the data transfer unit is restored by parity. , Combined and forwarded. As a result, when reading data, a storage device that operates as fast as possible can be used, and the access speed can be improved.
[0007]
(2)
At the time of writing, as in the case of reading, for example, when access to any four of the five storage devices is completed, the data stored in the four storage devices is read out to the other one The data to be stored in the storage device can be reproduced. Therefore, the write operation can be completed at that time. As a result, as in the case of data read, the access speed can be improved by using a storage device that operates as fast as possible.
[0008]
【Example】
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram of an embodiment of a disk array device of the present invention.
The illustrated apparatus includes a data read request issuing unit 1, a data write request issuing unit 2, a determination unit 3, a data transfer unit 4, data buffers 51 to 55, a write report unit 6, and a post-processing unit 7.
The data read request issuing unit 1 and the data write request issuing unit 2 are provided in the main control unit 10. The main control unit 10 controls the entire disk array device 11. The main control unit 10 is provided with a nonvolatile memory 12.
[0009]
The data read request issuing unit 1 issues a data read request to all of the data holding storage devices 21 to 24 and the parity holding storage device 25 via the data control unit 13 at the time of a data read request from the host computer 20. .
Similar to the data read request issuing unit 1, the data write request issuing unit 2 uses the data holding storage devices 21 to 24 and the parity holding storage device 25 via the data control unit 13 when a data write request is issued from the host computer 20. The data control unit 13 includes a determination unit 3, a data transfer unit 4, a write report unit 6, and a post-processing unit 7 that issue data write requests to all. The data control unit 13 processes the data according to the RAID level 3 and controls access to the data holding storage devices 21 to 24 and the parity holding storage device 25.
[0010]
After the data read request is issued by the data read request issuing unit 1, the determination unit 3 monitors completion reports of data read operations from the data holding storage devices 21 to 24 and the parity holding storage device 25. Then, when the data read operation of all the storage devices except the one that is the slowest is completed, it is determined whether the storage device that has not completed the operation is a data holding storage device or a parity holding storage device To do. That is, the completion reports from the five storage devices are sequentially stored, and when there is a completion report for up to four storage devices, a storage device that does not yet have a completion report is detected.
[0011]
When the operation of the parity holding storage device 25 is delayed according to the determination result of the determination unit 3, the data transfer unit 4 combines the data from the data holding storage devices 21 to 24 and combines the data from the main control unit 10. To the host computer 20. On the other hand, when the operation of any of the data holding storage devices 21 to 24 is delayed, the data missing due to the parity from the parity holding storage device 25 is reproduced. Then, each data including the reproduced data is synthesized and transferred to the host computer 20.
The data buffers 51 to 55 are each composed of a nonvolatile memory or the like, and are provided corresponding to the plurality of data holding storage devices 21 to 24 and the parity holding storage device 25, respectively. The data buffers 51 to 55 are divided into a plurality of data holding storage devices 21 to 24 and a parity holding storage device 25 by the data transfer unit 4 immediately after the data write request is issued by the data write request issuing unit 2. Data and parity are transferred, and the same data and parity are transferred.
[0012]
After the data transfer by the data transfer unit 4, the write report unit 6 indicates that the write operation has been completed at the time the data write operation is completed for all the storage devices except the one with the slowest data write operation. Report to a host computer 20. That is, the completion reports from the five storage devices are sequentially stored, and when there is a completion report for up to four storage devices, a storage device that does not yet have a completion report is detected. At the time of detection, a completion report is sent to the host computer 20 and the detection result is passed to the post-processing unit 7.
After the completion of the data write by the write report unit 6, the post-processing unit 7 discards the data in the data buffers 51 to 55 corresponding to the storage device that has already completed the data write operation. Then, when the remaining one storage device completes the data write operation, the data in the corresponding data buffer is also discarded.
[0013]
The data holding storage devices 21 to 24 are secondary storage devices made up of a hard disk or the like, and store data distributed in byte units.
Similarly to the data holding storage devices 21 to 24, the parity holding storage device 25 is also a secondary storage device made up of a hard disk or the like, and stores parity for data distributed in the divided data A to D.
[0014]
Next, the operation of the above-described disk array device will be described.
First, the general operation of a RAID level 3 disk array device will be described in detail.
FIG. 2 is a diagram showing a data and parity distribution format of the RAID level 3 disk array device.
Data from the host computer 20 is stored in a plurality of data holding storage devices 21 to 24 in the order of d0 to d3 and d4 to d7 in byte units. With respect to the data in byte units of these storage devices, the parities p 1 and p 2 are generated at the time of data write and stored in the parity holding storage device 25. In general, in a RAID 3 disk array device, data is distributed and stored in a plurality of storage devices in units of bytes as described above, so that data can be read / written at high speed and the parity holding storage device 25 is provided. Therefore, even if one of the plurality of storage devices fails, the disk array device 11 does not lose data, and high speed and reliability are exhibited.
[0015]
Next, the data write operation, data read operation, single disk drive, and data write operation at the time of failure of the RAID level 3 disk array apparatus will be described with reference to FIGS.
In response to a data write request from the host computer 20, the data write request issuing unit 2 of the main control unit 10 issues a data write request to the data holding storage devices 21 to 24 and the parity holding storage device 25. Then, when data storage preparation for these storage devices is completed, the write data from the host computer 20 is transferred to each storage device by the data transfer unit 4. At this time, the write data is interleaved in units of bytes with respect to the data holding storage devices 21 to 24, and at the same time, parity bytes for the data bytes of these storage devices are generated according to the formula in FIG. Forward to.
[0016]
Next, the operation of reading the data written in this way by the procedure according to the present invention will be described.
FIG. 3 is an explanatory diagram of a data read procedure according to the present invention.
In the figure, a, b, c, d, and p are data bytes and parity bytes obtained from the data holding storage devices 21 to 24 and the parity holding storage device 25. In response to a data read request from the host computer 20, the data read request issuing unit 1 of the main control unit 10 issues a data read request to all of the data holding storage devices 21 to 24 and the parity holding storage device 25. In response to the data read request, the disk drives of the data holding storage devices 21 to 24 and the parity holding storage device 25 are ready to execute data transfer after the positioning of the seek and spindle is completed.
[0017]
Here, the disk drive that is the latest in positioning of the spindle, or that has generated a seek error or selector change, is referred to as a data holding storage device 24. In this case, as shown in FIG. 3A, the data transfer unit 4 of the data control unit 13 is connected to the bus at the time when the disk storage of the data holding storage devices 21 to 23 and the parity holding storage device 25 is ready for transfer. The upper data bytes a, b, c, and p are fetched. The parity generator in the data transfer unit 4 generates the missing data byte d from the obtained data bytes a, b, c, and p, and the data bytes a, b, c, and d are converted into the main control unit 10. To the host computer 20 that is the read request source.
In this way, the influence of the slowest disk drive due to a delay in positioning of the spindle, seek error, selector change, or the like can be eliminated, and the data read operation can be speeded up.
[0018]
Next, an operation for writing data by the procedure according to the present invention will be described.
FIG. 4 is an explanatory diagram of a data synchronous write procedure according to the present invention.
In the figure, a, b, c, d, and p are data bytes and parity bytes to be transferred to the data holding storage devices 21 to 24 and the parity holding storage device 25. In response to a synchronous write request for data from the host computer 20, the data write request issuing unit 2 of the main control unit 10 writes data to all the disk drives in the data holding storage devices 21 to 24 and the parity holding storage device 25. Issue a request. In response to this data write request, each disk drive starts taking in the write data. Then, the data bytes a to d and the parity byte p are transferred to the respective disk drives, and at the same time, these bytes are written in the corresponding data buffers 51 to 55.
[0019]
After transferring all the data bytes a to d and the parity byte p, each disk drive performs writing to the medium through positioning of the seek and spindle, and reports completion of data writing. The timing of the completion report differs depending on each disk drive due to the difference in the positioning of the spindle, seek error, or sector change. For this reason, for example, when the completion report of the disk drive of the data holding storage device 24 is delayed, the write reporting unit 6 reports completion from the disk drives of the remaining data holding storage devices 21 to 23 and the parity holding storage device 25. Upon completion, a completion report is sent to the host computer 20.
[0020]
At this stage, the data in the data buffers 51 to 53 and 55 are no longer needed, so they can be discarded or the area can be reused. At this time, the non-volatile memory 12 stores the identifier of the disk drive that has not reported completion, the request address, the data length, and a flag indicating that there is no completion report. Thereafter, the post-processing unit 7 receives the completion report from the disk drive, discards the data in the buffer, or allows the area to be reused, and invalidates the contents of the nonvolatile memory 12. Thus, a series of processing for the write request from the host computer 20 is completed.
In this way, it is possible to eliminate the influence of the slowest disk drive due to a delay in positioning of the spindle, seek error, selector change, or the like, and to speed up the data write operation.
[0021]
If the disk array device 11 is down due to a power failure or the like after receiving a completion report from the disk drive whose access is delayed after the completion of the data write to the host computer 20, the write data in the data buffer 54 is written. Is lost. For this reason, after the power supply of the disk array device 11 is turned on thereafter, the missing data in the disk drive whose access is delayed is recovered based on the information remaining in the nonvolatile memory 12. This is done by the following procedure.
[0022]
First, information on the nonvolatile memory 12 is read to identify in which part of which disk drive missing data is present. Then, a write request is issued to the disk drive where the missing data exists, and a read request is issued to the other disk drives. With the data from the disk drive in which no data is lost, the parity generator in the data control unit 13 recovers the missing data according to the processing procedure shown in FIG. 3, and transfers it to the disk drive in which the missing data exists. Then, after transferring the missing data and receiving a write completion report from the disk drive where the missing data exists, the information in the nonvolatile memory 12 is invalidated and the missing data recovery process is finished.
In this way, it is possible to cope with a failure caused by power interruption.
[0023]
Of course, the present invention is not limited to the above-described embodiments, and various modifications are possible. For example, a hard disk device has been described as an example of a secondary storage device, but the same applies to a floppy disk device or the like. Further, the data buffer has been described so as to correspond to each secondary storage device, but it is also possible to physically divide and use one data buffer.
[0024]
【The invention's effect】
As described above, according to the disk array device of the present invention, the access request is issued to all the disk drives, and only the other disk drive except the slowest operation is completed. Since the completion report to the apparatus is performed, the waiting time for waiting for the disk to rotate can be reduced both at the time of reading and at the time of writing, and the access speed can be improved.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of a disk array device of the present invention.
FIG. 2 is an explanatory diagram of a data and parity distribution format of a disk array device of RAID level 3;
FIG. 3 is an explanatory diagram of a data read procedure according to the present invention.
FIG. 4 is an explanatory diagram of a data synchronous write procedure according to the present invention.
[Explanation of symbols]
1 Data Read Request Issuing Unit 2 Data Write Request Issuing Unit 3 Judgment Unit 4 Data Transfer Unit 6 Write Report Unit 7 Post-Processing Unit

Claims (2)

A data read request issuing unit that issues a data read request to all of a plurality of data holding storage devices and a parity holding storage device at the time of a data read request;
After the data read request is issued by the data read request issuing unit, when the data read operation of all the storage devices except the one with the slowest data read operation is completed, the storage device that has not yet completed the operation. A determination unit that determines whether the storage device is a data holding storage device or a parity holding storage device;
When the operation of the parity holding storage device is delayed according to the determination result of the determining unit, the data from the plurality of data holding storage devices are combined and transferred, while the data holding storage device Are delayed, due to the data from the other plurality of data holding storage devices for which the data read operation has been completed and the parity from the parity holding storage device are reproduced, A disk array device comprising a data transfer unit for combining and transferring data. A data buffer provided corresponding to each of a plurality of data holding storage devices and a parity holding storage device;
A data write request issuing unit that issues a data write request to all of the plurality of data holding storage devices and parity holding storage devices at the time of a data write request;
Immediately after the data write request is issued by the data write request issuing unit, the divided data and parity are transferred to the plurality of data holding storage devices and the parity holding storage device, and the data is also transferred to the data buffer. A transfer section;
After the data transfer by the data transfer unit, when the data write operation of all the storage devices except the one with the slowest data write operation is completed, the fact that the write operation has been completed is reported to the host device and the above is excluded. A write report unit for storing at least an identifier of one storage device in a nonvolatile memory;
After the completion of the data write by the write report unit, the data in the data buffer corresponding to the storage device that has already completed the data write operation is discarded, and then the data write operation is also completed for the remaining one storage device A post-processing unit that also discards the data in the corresponding data buffer,
When the data in the data buffer corresponding to the remaining one storage device is lost, the lost data is recovered by the data from the storage devices other than the remaining one storage device, and the nonvolatile memory is A disk array device that performs a write operation of the recovered data to the remaining one storage device with reference to the disk array device.

Images