JP2003150323A - Information storage device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to separately use individual disks constituting RAID (redundant array of independent disks) and change the constitution of the RAID without any pre-processing. SOLUTION: In this information storage device using a RAID system, forms, the contents such as whether pieces of data to be recorded are character data or video data or sound data of the data are discriminated, size of individual pieces of data to be separately recorded in the respective disk devices a109 to d112 to constitute the RAID, namely, a separation unit is changed to preset size in accordance with the discriminated form, contents of the data so that the individual pieces of data recorded in the respective disks can be reproduced to original data forms even when the individual pieces of data are separately reproduced, the individual pieces of data are separated and recorded by the separation unit.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array used as a storage system, and more particularly to an information storage device for recording / reproducing audio / video, document data, etc. using the disk array.

[0002]

2. Description of the Related Art Generally, the performance of a magnetic disk device alone,
For example, when high speed and high reliability are required, a disk array device that uses a plurality of disk devices that are driven independently of each other is configured, and this disk array device is used to configure a RAID (Redandent Array of Inexpensive Disk).
sks). As a result, it is possible to ensure the performance that is higher than that of one magnetic disk device. That is, high speed and high reliability can be ensured by using a large number of disk devices and distributing access.

Generally, there are six types of RAID levels constituting RAID, from RAID0 to RAID5, each having different reliability and high speed. Below, RAID0
And RAID 5 will be outlined.

First, a data recording method of RAID0 composed of four disks (disk A to disk D) will be described. Input / output data is written in order from the disk A to the disk D by a specific size by the disk array controller. Here, the size of the data recorded on each of the disks A to D is called a striping unit. RAID 0 is called striping and handles only data to be recorded, does not consume extra disk area due to redundant data such as parity information, and does not require extra I / O.
Does not occur. Therefore, the storage area of the disk can be used most effectively. However, since there is no redundancy in the data, if even one of the plurality of disk devices that make up the RAIDO fails, the data cannot be read or written, and the reliability is lower than other RAID levels.

For this reason, there is a disk array such as RAID 5 in which the number of magnetic disk devices is made redundant and the parity information is recorded in the redundant part. Below, 4
A data recording method of RAID5 composed of one disk (disk A to disk D) will be described. When RAID 5 is composed of four disks, the input / output data is sequentially written by the disk array controller to three disks A to C for each striping unit. On the disk D, the parity information obtained by the exclusive OR for each byte of the data from the disk A to the disk C is recorded. Next continuous data is recorded on three disks from disk D to disk A and disk B in order, and parity information is recorded on disk C. In this way, with four disk devices RAID5
When configured, the disc for recording data is always 3
The disk on which the parity information is recorded is shifted one by one. RAID 5 is called distributed data guarding because the disks on which the parity information is recorded are distributed by sequentially recording the parity information on each disk.

In the case of RAID 5, theoretically three magnetic disk devices are required. Further, in RAID5, even if the constituent magnetic disk device is missing, if the number of missing disk devices is one, data can be restored by using the parity information. On the contrary, when one disk device is removed, the data of that disk device is meaningless when viewed as a single disk device, and correct data cannot be obtained and is destroyed. Is the same.

[0007]

Generally, in RAID, the striping unit is fixed, and the size of the striping unit once fixed cannot be changed later.
This is because if the striping unit is changed, the order of reading the data recorded up to that time will be changed, and correct data cannot be obtained. Also,
In the case of RAID0, correct data cannot be obtained if even one disk device constituting RAID0 is missing for some reason. Therefore, in order to insert and remove the disk device, it is necessary to replace the data array in all the disks constituting RAID0.

Further, in RAID 5, even if the constituent magnetic disk device is missing, if the number of the missing disk devices is one, data can be restored by using the parity information. However, if there are two or more missing disks, correct data cannot be obtained as in RAID0.

When attention is paid to individual disk devices constituting a RAID, what is recorded in each disk device is a part of the entire data, but it is recorded in striping units regardless of the contents of the data. Since it is data, it has no meaning. Therefore, let's not get correct data when handling the disc alone.
You can't even get meaningful data.

Therefore, when it is desired to change the RAID configuration such as inserting and removing the disk device, the RA
Since it is necessary to perform processing such as data rearrangement on each disk constituting the ID, the user cannot freely and easily change the ID. That is, even if you try to exchange data or exchange disks, RAID
It is not possible to freely insert and remove the disk device that constitutes the device, and there is a problem in terms of usability.

In a device such as a mobile device or a wearable device which is used while being carried by an individual,
Unlike conventional stationary devices that are always supplied with power from outside, reading and writing a large amount of data that occurs in necessary processing each time the RAID configuration is changed is a very large problem from the viewpoint of power consumption. . In addition, in a portable RAID such as a mobile device, there are more opportunities to replace a magnetic disk device that constitutes the RAID and to independently use the disk device than in a conventional stationary RAID.

An object of the present invention is to enable individual use of individual disks constituting a RAID or change the RAID configuration without preprocessing.

[0013]

[Means for Solving the Problems] In a RAID system, the striping unit is not a fixed amount of data but a data format, that is, data format and content such as character data, video data, audio data, etc. By changing the striping unit accordingly and recording, it is possible to obtain meaningful data that is decipherable even when the disk device forming the array changes. The striping unit is not a fixed amount of data,
By changing and recording according to the format and content of the data, even if one of the plurality of disk devices that make up the RAID system fails, the data is saved and saved with some data missing. The data may be decipherable, although the data may not be recovered intact.

Further, when the data is viewed by each individual magnetic disk device, the striping unit is not fixed in the amount of data but is changed according to the format and content of the data and recorded, so that the data recorded therein is changed. Is not irreproducible (the same as being destroyed), but a part of the data is missing, but it is recorded in a reproducible state.

Therefore, even if one or more disk devices are missing in RAID0 or RAID5, or even if only one disk device constituting the RAID is used, all data cannot be obtained, but It is possible to obtain meaningful data that is a part of

Therefore, by appropriately selecting the unit for dividing the data, the recorded data can be discriminated even if only one of the plurality of disk devices constituting the RAID is taken out and used. , It is easier to use. In addition, RAI can be performed without performing pre-processing (pre-processing).
Even if the configuration of D is changed, there is no fear that the data recorded in each disk device will become unusable.

Note that changing the division unit according to the data format means changing the division unit depending on whether it is character data, video data, audio data, or the like, and it changes according to the content of the data. This means that, for example, when character data is divided into paragraphs, the data size changes for each paragraph.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described with reference to the drawings. FIG. 2 shows a portable information storage device of this embodiment, in which operation buttons and a monitor for image display are arranged on the front surface, and a card slot into which a memory card is inserted is provided on the upper surface, which is not shown. The speaker is installed. FIG. 3 shows the main configuration of this information storage device.

The illustrated information storage devices are independent of each other,
And a plurality of removable small-sized magnetic disk devices, magnetic disk device interfaces provided for each of the small magnetic disk devices, and division control means connected to the magnetic disk device interfaces. RAID controller, main controller connected to this RAID controller, communication interface and small memory device interface connected to this main controller, and small memory connected to small memory device interface It is configured to include a device, the operation buttons connected to the main controller, an image display monitor, and a speaker (not shown). The card slot is a portion into which a memory card of a small storage device is inserted.

The data read from the memory card inserted in the small storage device is taken into the RAID controller as original data via the main controller, divided and recorded in a plurality of small magnetic disk devices. Data received from outside via wireless or communication line is
The original data is taken into the RAID controller via the communication interface and the main controller, divided and recorded in a plurality of small magnetic disk devices. The image display monitor can display the information read from the small magnetic disk device or the small storage device, the data received from the outside, and the RAID configuration at that time. There is.

The operation buttons are used by the operator to set the RAID configuration and division unit when dividing and storing information in the plurality of small magnetic disk devices.

The configuration and operation of the RAID controller will be described below with reference to FIG. In this embodiment,
It is assumed that RAID 0 has four small magnetic disk devices (disk a109 to disk d112). The RAID controller includes a data discriminator (complexer) 102 connected to the main controller and a data divider / combiner 1 connected to the data discriminator (complexer) 102.
04, a data discriminating (combining) device 106 connected to the data dividing / combining device 104, and a controller 108.
It is connected to the disk d.

A series of original data 1 from the main controller
When 01 is sent, the RAID controller uses the data divider / combiner 104 to divide the data and
Disks a109 to d112 that compose D
Record the data in. The data is not divided into a fixed size, but is divided into meaningful units even after the division according to the type of data to be handled. Therefore, the data discriminator (composite) device 102 uses the original data 10
The format and content of No. 1 are discriminated, and the original data image 103 obtained by the discrimination (decoding) is subjected to data division using the data division / combiner 104.

In determining the format and content of the original data 101 in the data discriminating (compositing) device 102, it is discriminated based on the extension and header information of the data, or discriminated based on the data format (structure) already given by the user. There is. The data format (structure) mentioned here is a classification such as character data, audio data such as music, or image data such as a moving image. The meaningful unit of the division unit described above is, for example, a paragraph unit or a page unit in the case of character data, a time unit of several seconds in the case of voice, and an image data of 1 second in the case of a moving image. is there. In this case, for example, in the case of character data, the size (size represented by the number of bits) of each divided data itself is not constant for each paragraph or page but changes.

In the case of FIG. 1, the original data image 103
In, since the value for each sample has meaning, the data is divided for each value for each sample. In consideration of general data, the sample of the original data image 103 indicates time and position (spatial coordinates). The data recorded on each of the disks a109 to d112 is the original data image 105 obtained by division. The original data image 105 divided by the data discriminating (decoding) device 106 is decoded into the divided data 107, and each data of the divided data 107 is sequentially recorded in the respective disks 109 to 112 by the controller 108. Go.

As a method of obtaining the divided data 107, a method of directly dividing the original data 101 by the data dividing / combining unit 104 using the original data image 103 can be considered. In addition, when the original data 101 is not compressed, the data size of each sample may be the same, and if the data size per sample is found by the data divider / combiner 104 using the original data image 103. The RAID may be configured with the striping unit as the unit.

In the case where a disk is lost in the conventional RAID 0, meaningful data cannot be restored, but by performing recording by the above method,
Even if the disc is lost, a part of the data (meaningful data) can be sufficiently reproduced although it is not the original complete data. For example, RAI with two disk devices
When recording a moving image by configuring D0, if recording is performed by dividing it into units of data every 1 second, since data is recorded every 1 second on each disc, the moving image is reproduced every 1 second. It is quite possible to do so. Considering uncompressed video data, the number of images per second is 30 frames, so 3
If RAID 0 is configured with 0 disk devices and recording is performed by the above method, even if only one disk device that is configured with RAID 0 is used, it can be sufficiently reproduced, and it is not so smooth, but 1 second. This means that one frame of video data is saved.

In the case of audio data, if it is divided into data every one second, it is difficult to determine what is recorded when reproduced by only one disk device. In the case of voice data, it is desirable to divide the data into longer time units in order to obtain meaningful data.

In any case, by recording by the above-mentioned method, the video (image) or sound at a specific time can be
It is possible to reproduce from the data of one disk device.

In RAID 5, in order to increase the reliability of data, parity information is recorded in a redundant disk device for recording data. In Figure 1, original data 1
When the size of the data 107 for each sample of 01 (the data size for each bold line frame) is the same, the striping unit is made equal to this data size, and the parity information can be created as in the conventional case. .

On the other hand, in the compressed data, since the data size of each sample is different, the parity information cannot be created for each byte as in the conventional case. Therefore, there are the following two types of methods for creating parity information. One is to add dummy data to the data so that the size of each piece of data for which parity information is made equal, and to create parity information for each byte for each piece of data to which dummy data is added,
The other is a method of creating parity information according to the content of each data.

A method of creating parity information according to the content of each data will be described. In general, data compression methods can be divided into lossless compression and lossy compression depending on whether or not the data is reversible. In lossless compression, the same data can be obtained even if the original data is compressed and decompressed (the reverse operation of compression is called decompression). On the other hand, in lossy compression, if the original data is compressed and decompressed, the same data cannot be obtained. Examples of lossy compression include JPEG and MPEG.

To create parity information for the original data compressed by lossless compression, as shown in FIG.
The (divided) original data 401 is decompressed by the decompressor 402, and the decompressed data is used to generate the parity information generator 40.
3, the parity information is created, the parity information is compressed by the compressor 404, and recorded as the parity information in the disk device 405.

However, in the case of irreversible compression, when the parity information is created by the above means shown in FIG. 4, the parity information itself is irreversibly compressed, so that the data recorded on the disk also includes the compressed parity information. And cannot be restored to. Here, FIG. 5 shows an image compression method based on the discrete cosine transform. The original image 501 is compressed into image data 503 by an encoder 510 based on discrete cosine transform. In the encoder 510, the discrete cosine transformer 511 performs the discrete cosine transform, and the quantizer 512 quantizes it based on the quantization table 531.
The data is irreversibly compressed at this point and cannot be decompressed into the same data. The quantized data is encoded by the entropy encoder 513 using the encoding table 532 and becomes the image data 503.

Further, the image data 503 is reproduced by the decoder 520 based on the inverse discrete cosine transform and reproduced image 502.
Is converted to. In the decoder 520, first, the encoding table 532 is used for decoding in the entropy decoder 523. After that, the inverse quantizer 522 performs inverse quantization on the basis of the quantization table 531 and the inverse discrete cosine transformer 5
Inverse discrete cosine transform is performed at 21 to form a reproduced image 502. As shown in FIG. 5, the lossy compression is performed by a portion (discrete cosine transformer 511 and quantizer 51) that performs lossy compression.
2) and a part that performs reversible compression (entropy encoder 5
13) and. Therefore, in order to create a parity for lossy-compressed data, attention is paid only to the lossless compression part, and the decompressor 402 and the compressor 40 shown in FIG.
In step 4, reversible compression and decompression processing can be performed to create parity information.

In the image compression system shown in FIG. 5, the entropy encoder 513 using the encoding table 532,
Only the entropy decoder 523 is used to perform the compression and decompression of FIG.

Next, an example of a method of creating parity information when the data size of each sample is different will be described with reference to FIG. The decompressed data is the target. With respect to the divided data images 601, 602, the parity information generator 603 generates the parity information 604 not for each byte of data but for each value of a specific dimension indicated by data such as a pixel in image data. create. That is, the image of each frame is divided into a grid and the areas of the grid at positions corresponding to each other (in the case of FIG. 6,
The parity information 604 is sequentially created for the image data in the upper left corner area).

With the above method, it is possible to construct a RAID system having striping units according to the format and contents of data.

As a command given to the above RAID system by a user who uses the RAID system, the R at the time of recording is used.
Designation of AID configuration (RAID level, designation of disks constituting RAID), designation of a disk used at the time of reproduction, data restoration in RAID5, conversion from a recording state in RAID to a recording state in non-RAID, and For example, you can specify the unit to divide the data. R specified by the user
The RAID system configures a RAID and records data according to the configuration of the AID. The data recorded on each disk constituting the RAID is recorded as a separate file so that each disk can be reproduced alone.
However, the file name is the same as the original data, or RA
It is represented by a variable according to the ID configuration. Further, to each of the divided data recorded in each disk device, data indicating a RAID configuration at the time of recording is added to a header portion or recorded in a file distinguished by an extension. The recorded RAID configuration includes a RAID level and information necessary for individually recognizing the disk devices that configure the RAID, such as a serial number.
At the time of reproduction, data indicating the RAID configuration at the time of data recording is used, and the data read from one or a plurality of magnetic disk devices designated by the user is reconstructed for each sample, so that the original continuous Data can be reproduced. When the disc is designated so that the same data as the original cannot be reconstructed at the time of reproduction, the reproduction is performed excluding the parity information recorded on each disc. R
In the conversion from the recording state in AID to the recording state in non-RAID, the divided and recorded data is read from each disk device, the data is reconstructed by the data processing device for each sample, and one magnetic disk device is reconstructed. To record.

In the case of character data, the unit for dividing the data can be changed so that it is divided for each designated line number, each paragraph, or each designated page number. In the case of image data, it can be changed so as to be divided every designated time or every designated number of image frames, and in the case of audio data, it can be divided every designated time. The operator sets the division unit in the RAID controller in advance for each data format by using the operation unit (the operation button) for setting the division unit. Original data 101
Is a data discriminating (decoding) device 1 of the RAID controller.
The format is determined by 02, and the data division / combiner 104 divides the original data into the set division units according to the determined data format. The data discriminating (encoding) device 106 assembles the divided original data into the divided original data 107 shown in FIG.
Reference numeral 08 denotes the divided unit, which is sequentially allocated and stored in each of the magnetic disks a to d.

[0041]

As described above, according to the present invention, RAID
By changing the size of each piece of data that is divided and recorded on each disk that composes the data according to the format (content) of the original data, the size of the data has a meaningful value (for example, the data size of each sample). ), Data can be recorded on each disk constituting the RAID.
Even if the structure of the AID is freely changed without any pre-processing, it is possible to reproduce the data in each sample recorded on each disc. Moreover, since the configuration of the disks constituting the RAID can be changed without preprocessing, it is possible to reduce the time and power consumption required for preprocessing.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention.

FIG. 2 is a perspective view showing an appearance of an information storage device to which the present invention is applied.

FIG. 3 is a system diagram showing a main configuration of the information storage device shown in FIG.

FIG. 4 is a diagram showing a method of creating parity information for compressed data in the embodiment of the present invention.

FIG. 5 is a diagram showing an example of a conventional image compression method.

FIG. 6 is a diagram showing a method of creating parity information for the content of data in the embodiment of the present invention.

[Explanation of symbols]

101 Original data 102 data discriminator 103 Original data image 104 data divider / combiner 105 divided data images 106 data discriminating (encoding) device 107 Original data divided 108 controller 109-112 disk device 401 Original data 402 Decompressor 403 Parity information generator 404 compressor 405 disc 501 original image 502 Playback image 503 image data 510 encoder 511 Discrete Cosine Transform 512 quantizer 513 Entropy encoder 520 Decoder 521 Inverse Discrete Cosine Transform 522 Dequantizer 523 Entropy Decoder 531 Quantization table 532 encoding table 601, 602 divided data image 603 Parity information generator 604 Parity information

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Takashi Yamaguchi             502 Kintatemachi, Tsuchiura City, Ibaraki Japan             Tate Seisakusho Mechanical Research Center (72) Inventor Masaki Odai             502 Kintatemachi, Tsuchiura City, Ibaraki Japan             Tate Seisakusho Mechanical Research Center F-term (reference) 5B018 GA02 HA35 MA14                 5B065 BA01 CA30                 5B082 CA07                 5D044 AB05 AB07 BC01 CC05 CC09                       DE12 DE69 DE83 DE92

Claims (10)

[Claims] 1. An information recording / reproducing method using a disk array system in which data is divided into a plurality of disk devices and sequentially recorded, and at the time of reproduction, data of the plurality of disk devices is sequentially read and combined. A recording / reproducing method of information, characterized in that the size is changed according to the format and content of the data. 2. The information recording / reproducing method according to claim 1, wherein the format of the data is at least character data,
A method of recording and reproducing information, characterized in that it includes image data and audio data as formats. 3. The information recording / reproducing method according to claim 1, wherein the data is divided into a plurality of disk devices, and the data stored in the plurality of disk devices is read from any one of the disk devices. At times, the information recording / reproducing method is characterized in that the size is such that it can be reproduced in the original data format. 4. The information recording / reproducing method according to claim 1, wherein the size of each data is equalized to each data recorded in the plurality of disk devices. As described above, the information recording / reproducing method is characterized in that dummy data which is not necessary for individually reproducing the individual data is added. 5. A disk array system is used, in which data containing at least one of video and audio is divided and recorded in order in a plurality of disk devices, and the data in the plurality of disk devices are sequentially read and combined at the time of reproduction. In the information recording / reproducing method, the information recording / reproducing method is characterized in that, when the data is divided, the data is divided for each preset time unit. 6. The information recording / reproducing method according to claim 5, wherein the length of time is a length with which data separately recorded in each disk device can be reproduced in the original data format. A method for recording and reproducing information, which is characterized by being present. 7. The information recording / reproducing method according to claim 1, wherein the data is recorded in at least one of the plurality of disk devices corresponding to the data. An information recording / reproducing method characterized by recording information indicating a RAID configuration at the time. 8. The information recording / reproducing method according to any one of claims 1 to 7, wherein the reproduction of data recorded by two or more disk devices less than the plurality of disk devices is performed. A method for recording / reproducing information, wherein data is reproduced except for parity information for data recovery recorded in the plurality of disk devices. 9. A device for taking in information, and an interface in which a plurality of disk devices for dividing and storing the information are detachable, and RAID is used in the plurality of disk devices connected to the interface. In the constituent information storage device, display means for indicating the current RAID structure, and the information is divided into preset units according to the format of the data forming the information and stored in order in the plurality of disk devices. Dividing control means and dividing means RA
An information storage device comprising: an ID configuration and an operation unit for setting a division unit when the information is recorded in the plurality of disk devices. 10. The information storage device according to claim 8, wherein the data recorded by being divided into the plurality of disk devices can be reproduced in the original data format even when the disk devices are independent. A characteristic information storage device.