JPH08161123A - Disk array device - Google Patents

Abstract

PURPOSE: To provide a disk array device capable of effectively using disk capacity while holding high reliability for evading a data loss properly correspondingly to a data access state from a host system and preventing the efficiency of access to data from being reduced. CONSTITUTION: The disk array device has data-multiplexed disk devices 5, 6 and a data-divided disk devices 7 to 9, a disk array control means 16 moves a data block with high writing frequency to the devices 5, 6 and moves a data block with high reading frequency to the devices 7 to 9 based upon information detected by an access frequency detecting means 12 in a disk controller 3. Since the data block with high writing frequency are not stored in the devices 7 to 9, overhead due to writing processing can be reduced, so that even if an access state to disk devices is changed, the disk array device can be operated without reducing access efficiency.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array device used as an external storage device of an electronic computer system.

[0002]

2. Description of the Related Art A conventional disk array device will be described below with reference to the drawings. FIG. 2 is a block diagram showing a configuration example of a conventional disk array device.
In FIG. 2, reference numeral 21 is a host system, and this host system 21 is connected to a disk array controller 23 via a host interface 22, and the disk array controller 23 is connected to a disk device 25 via a device interface 24. To 28.

The disk array controller 2
3 is a disk array control means 29 which controls the control.
And data from the host system 21 to the disk device 2
Writing means 30 for writing in 5 to 28, and disk device 2
Read-out means 31 for reading out data from 5-28. The operation of the disk array device configured as above will be described.

Here, a RAID that is commonly used
Level 1 (mirror disk), RAID level 3, and RAID level 5 will be described. In addition, RAID
(Redundant Arrays of Inex
"Pensive Disk" is a disk array technology, and is divided into five levels according to the mode of writing / reading data to / from a plurality of disk devices.

First, when set as a RAID level 1 (mirror disk), the disk devices 25 and 27 are data disks, and the disk devices 26 and 28 are mirror disks which are backups of the data disks (disk devices 25 and 27). It will be positioned. The data output from the host system 21 is input to the disk array controller 23, and the writing means 30 writes the disk devices 25 and 26 or the disk device 27.
The same data is written in the same data block address of No. 28 and No. 28. Each disk device 2 is shown in FIG.
An example of data block storage in 5 to 28 is shown, and Block
0, Block1, ... Show data in logical block units. In a normal read operation, the disk device 25 as a data disk is read by the read means 31.
Alternatively, the data is read only from 27.

Then, the data disk (disk device 2
5 or 27) or an error is detected at the time of reading data, the processing is switched to the mirror disk (disk device (26 or 28) side to read the data, and the system operation is continued. Also,
When set as RAID level 3, the disk devices 25, 26 and 27 are data disks and disk devices 28.
Is a disk that stores parity data as redundant data.

After the data output from the host system 21 is input to the disk array controller 23,
The data is divided into byte units and written on the data disk (disk devices 25, 26, 27) by the writing means 30. Also, parity data through the same block address on all data disks is written to the disk device 28. FIG. 3B shows the disk devices 25-2.
8 shows an example of data block storage in 00, 00h, 11h,
..., AAh, BBh indicate data divided into byte units, and Parity indicates parity data. In a normal read operation, all the data disks (disk devices 25, 26, 27) are accessed by the read means 31 to read data, and are recombined into data blocks.

Then, the data disk (disk device 2
5), 26, 27) or when an error is detected during data reading, the disk device 2
The parity data can be read from 8 and the data can be restored. Further, when the RAID level 5 is set, the data output from the host system 21 is input to the disk array controller 23, and then the data is distributed in logical block units, and the writing unit 30 writes each data disk (disk device). 25, 26, 27, 2
8) is written. Also, parity data through the same block address on all data disks is written to the disk storing redundant data. At this time, the disk in which the parity data is written is not fixed, but is written by being dispersed cyclically in all the disk devices 25 to 28. FIG. 3C shows an example of data block storage in each of the disk devices 25 to 28. In a normal read operation, the read means 31 accesses a disk storing a necessary data block to read data.

If any one of the data disks fails or an error is detected when reading the data, the data can be read from the disk device containing the parity data and the data can be restored.

[0010]

However, the conventional disk array device has the following problems. In the RAID level 1 setting, a highly reliable system can be realized by holding multiple data, but a disk capacity twice as large as the data capacity to be stored is required.

In the setting of RAID level 3, in order to read data in parallel from all the data disks (disk devices 25, 26, 27) at the same time, it is possible to read the data substantially faster than only a single disk. You can However, at the time of writing, since the amount of data (physical block) to be accessed at one time is as large as at least the sector capacity of the disk multiplied by the number of data disks, the unit of data block to be accessed from the host system 21 (logical block). ) Is smaller than this, there is an overhead that the physical block including the desired logical block is once read, the parity data is updated in the disk array controller 23 and updated, and the data is written to the disk device.

In the setting of RAID level 5, each disk device 25, 26, 27, 28 can be accessed independently in logical block units, but like RAID level 3, at the time of data writing. In order to maintain the consistency of redundant data, the existing data and parity data are read from each of the data blocks desired by the disk device, and new parity data is calculated in the disk array controller 23 together with the new data to be written. , Overhead of writing new data and new parity data to each data block desired by the disk device.

As described above, each level of RAID has advantages and disadvantages. Normally, the disk array device is set to any one level of RAID and operates, but the size of the data block to be accessed and the write / write operation are controlled by the application software handled by the host system 21.
Since the read ratio and the like are different, depending on the compatibility with the selected RAID level, not only the effect as a disk array device cannot be exhibited, but also the efficiency may be reduced as compared with the case of connecting a single disk.

According to the present invention, the disk capacity is effectively used and the data access efficiency is reduced while maintaining the high reliability of avoiding the data loss by responding to the data access state from the host system in a timely manner. First, it is an object of the present invention to provide a disk array device that continues to exert its effects sufficiently.

[0015]

DISCLOSURE OF THE INVENTION A disk array device of the present invention is a disk array device comprising a plurality of disk devices and a disk array controller for writing and reading data to and from the disk device from the outside. Therefore, the plurality of disk devices are composed of a plurality of data multiplexing disk devices for storing the same data and a plurality of data division disk devices for storing the divided data and its redundant data. Is a writing means for writing data to the disk device,
The reading means for reading data from the disk device, the access frequency detecting means for detecting and storing the number of read requests and the number of write requests externally requested for each data block of the disk device, and the data read by the reading means. Storage means for temporarily storing and sending to the writing means, and data multiplexing disk for storing the data blocks stored in the data division disk device and having a large number of write requests detected by the access frequency detecting means via the reading means, the storing means and the writing means A disk which is moved to the device and is moved to the data division disk device through the reading means, the storage means and the writing means, the data block being stored in the data multiplexing disk apparatus and detected by the access frequency detecting means and having a large number of read requests. The provision of array control means And butterflies.

[0016]

The disk array device of the present invention uses a plurality of disk devices as a data multiplexing disk device and a data dividing disk device, detects the number of write / read requests in external data access, and writes the write requests. A data block having a large number of times is moved to a data multiplexing disk device, and a data block having a large number of read requests is moved to a data dividing disk device. That is, the data for which the number of write requests is large is stored in the data multiplexing disk device instead of being stored in the data dividing disk device, so that it is possible to prevent the efficiency from being lowered due to the overhead generated in the writing process.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A disk array device according to an embodiment of the present invention will be described below with reference to the drawings. FIG.
FIG. 1 is a block diagram showing the configuration of a disk array device according to an embodiment of the present invention.

In FIG. 1, reference numeral 1 is a host system, which is connected to a disk array controller 3 via a host interface 2, and the disk array controller 3 is a device system.
5 disk devices 5-9 via interface 4
Connected to. Of these five, two disk devices 5 and 6 are used as a data multiplexing disk device similar to RAID level 1 in which the same data is written, and the three disk devices 7, 8 and 9 divide the data. Then, each disk device is used as a data division disk device similar to RAID level 5 in which data blocks and parity data are stored.

The disk array controller 3
Is a host interface means 10 for decoding an instruction from the host system 1, receiving / sorting command signals and data signals, and transmitting status information and data signals to the host system 1, and from the host interface means 10. Disc information 5
Access frequency detecting means 12 for detecting read / write requests for the data blocks 9 to 9 and storing the number of times in the internal directory 11 and writing to write data and parity data to the five disk devices 5 to 9. Means 13, read means 14 for reading data and parity data from the five disk devices 5-9, and read means 14 for reading the data stored in each of the disk devices 5-9 in order to replace the storage location. Based on the information of the storage unit 15 that temporarily holds the obtained data and the host system 1, the writing unit 13 and the reading unit 14 are controlled based on the information of the access frequency detecting unit 11, and the data writing process is performed. In order to prevent the occurrence of overhead at the time, a disk array system that operates the data storage in the disk devices 5-9 And a means 16.

The operation of the disk array device of this embodiment thus constructed will be described. First, as an initial setting, the data from the host system 1 is sent to the writing means 13 via the host interface means 10, is divided into data blocks according to the instruction of the disk array control means 16, and is generated together with the generated parity data. , Are written in a distributed manner in the data-partitioned disk devices 7-9. If the disk devices 7-9 are full,
It is written in the data multiplexing disk devices 5 and 6.
As described above, the same data is written in the disk devices 5 and 6.

When the host system 1 reads data from the data storage state as described above, the host system 1
The host interface means 10 which has received the read command from the above transmits the information to the access frequency detection means 12 and the disk array control means 16. The disk array control means 16 reads the data from the disk device in which the corresponding data is stored via the read means 14 and sends it to the host system 1 via the host interface means 10. At this time, the access frequency detecting means 12
Then, the accessed data block number is stored in the entry of the internal directory 11, and the value of the counter (read counter) for counting the number of read accesses is increased by one.

Next, when the host system 1 writes data, the host interface means 10 receiving the write command from the host system 1 transmits the information to the access frequency detecting means 11 and the disk array control means 16. Here, the access frequency detecting means 12 stores the accessed data block number in the internal directory 1
The value of a counter (write counter) that counts the number of write accesses is incremented by one.

Thereafter, each time an access is made to a data block that has already been stored, the value of the read counter or the write counter of each entry of the directory 11 in the access frequency detecting means 12 increases. In addition, when new data is added and written in the empty data block, the data division disk device 7,
If 8 and 9 have a margin, they are stored there. However, when there is no room, the disk array control means 16 extracts the information of the directory 11 of the access frequency detection means 12,
A block which is stored in the data division disk devices 7, 8 and 9 and has the largest write counter value is selected, is read out by the reading means 14, and is read out by the storage means 15; By writing to the devices 5 and 6 via the writing means 13, the data is moved, and new data is written to the empty blocks of the data-partitioned disk devices 7, 8 and 9 via the writing means 13.
At this time, the disk array control means 16 also updates the physical block address of the entry of the directory 11 of the access frequency detection means 12 for the moved data block.

As in this embodiment, since the data blocks that are frequently written are not made to reside in the data-partitioned disk devices 7, 8 and 9, the overhead due to the write processing can be reduced, so that the access to the disk device is possible. Even if the situation changes, the disk array device can be operated without reducing the efficiency. In particular, it is effective for a server type computer system that handles various application software. Further, the data stored in the data multiplexing disk devices 5 and 6 has high reliability.

There is a limit to the number of entries in the directory 11, and when all data blocks cannot be recorded, a known algorithm such as LRU (Leas) is used.
(e.g., tRecently Used), the oldest accessed block among the blocks stored in the entry is replaced for the newly accessed block, or the stored contents are erased at a certain time. It manages by the way to go.

In this embodiment, RAID level 5 is adopted as the data division disk system, but RAID level 3 may be used. Further, in this embodiment, the movement from the data division disk device 7, 8, 9 to the data multiplexing disk device 5, 6 is carried out only when new data is added. When the access to the data block is low, or periodically, the information in the access frequency detecting means 12 is appropriately referred to, and the data block having a high write frequency is stored in the data multiplexing disk device 5,
If the disk array control means 16 is operated so as to move the data so that the data blocks that are frequently read are stored in the data partitioning disk devices 7, 8 and 9, the efficiency can be further improved. it can.

[0027]

As described above, the present invention uses a plurality of disk devices as a data multiplexing disk device and a data dividing disk device, detects the number of write / read requests in external data access, and writes the data. A data block with a large number of requests is moved to a data multiplexing disk device, and a data block with a large number of read requests is moved to a data dividing disk device. In this way, the storage locations of frequently written data blocks and frequently read data blocks can be automatically adjusted to reduce the overhead due to the writing process, so that the access status to the disk device fluctuates. Also,
There is an effect that the disk array device can be operated without reducing the efficiency. In particular, it is effective for a server type computer system that handles various application software. The data stored in the data multiplexing disk device has high reliability.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a disk array device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing the configuration of a conventional disk array device.

FIG. 3 is a diagram showing an example of data block storage in a disk device in a conventional example.

[Explanation of symbols]

 1 Host System 2 Host Interface 3 Disk Array Controller 4 Device Interface 5, 6 Data Multiplexing Disk Device 7, 8, 9 Data Partitioning Disk Device 10 Host Interface Means 11 Directory 12 Access Frequency Detecting Means 13 Writing Means 14 Reading means 15 Storage means 16 Disk array control means

Claims (1)

[Claims] 1. A disk array device comprising a plurality of disk devices and a disk array controller for writing and reading data to and from the disk device, wherein the plurality of disk devices are the same. The disk array controller comprises a plurality of data multiplexing disk devices for storing data and a plurality of data partitioning disk devices for storing the divided data and redundant data thereof, wherein the disk array controller writes data to the disk device. Means, read means for reading data from the disk device, access frequency detection means for detecting and storing the number of read requests and the number of write requests externally requested for each data block of the disk device, and the read The data read by the means is temporarily stored and Storage means for sending data to the data division disk device, and the data block stored in the data division disk device and having a large number of write requests detected by the access frequency detection means, through the read means, the storage means and the write means. While moving to the disk device, the data block stored in the data multiplexing disk device and having a large number of read requests detected by the access frequency detecting means is divided into the data through the reading means, the storing means and the writing means. A disk array device comprising a disk array control means for moving the disk array device.