JPH10161819A - Disk array control method and device therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To set a fixed data read/write time and also to improve the reliability of data. SOLUTION: The (n+1) pieces of disk devices are selected out of (m) pieces of disk devices in all to store the error correction data and (n) pieces of sub- blocks obtained by dividing the error correction data (S1). Then an outer circumference track where the error correction data and a prescribed sub-block are prepared, an inner circumference track where other sub-blocks are prepared and a start logical sector are decided for every selected disk device (S2). The error correction data and the size of every sub-block are decided so as to secure a fixed ratio between the number of sectors of a track and the error correction data prepared on the track or the sub-block size (S3).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array control method and apparatus, for example, by generating error correction data from sub-blocks having different sizes.
The present invention relates to a disk array control method and apparatus capable of keeping reading and writing times of error correction data and sub-blocks constant irrespective of data storage locations.

[0002]

2. Description of the Related Art In recent years, computer systems have been remarkably developed, and with the progress thereof, demands for external storage devices have been increasing. A disk array device has been developed to meet the demand. This disk array device incorporates a plurality of disk devices to increase capacity,
Each disk drive is operated in parallel to speed up reading or writing, and incorporates error correction data to increase reliability.

When data is stored in a disk array device, the data is divided into a plurality of sub-blocks, error correction data is generated based on the sub-blocks, and all the sub-blocks and the error correction data are stored in the sub-blocks. Write to different disk units.

Conversely, when data is taken out from the disk array device, a plurality of sub-blocks and error correction data are simultaneously read out from the disk device storing the sub-blocks constituting the data and the error correction data. The original data is constructed from the sub-blocks, and is transmitted as it is if there is no error. At this time, if the data cannot be read normally because the recording area where the sub-block is stored is broken or the like, correct data is read based on the other normally read sub-blocks and the error correction data. And then send it out.

In a disk array device, 1
Even if one disk device is completely broken, the function of replacing the broken disk device with a new one and using the data of another disk device to recover the damaged data can be provided.

There are several different types of disk array devices incorporating error correction data. David A. from UC Berkeley. Professor Patternson and others
The method is classified into five stages, and RAID (Redundant Arra
ys of Inexpensive Disks) for the first time. The contents are briefly introduced below.

[0007] RAID-1 duplicates data in a disk device and is also called a mirrored disk. In RAID-1, exactly the same data is stored in two disk devices. In RAID-2 and RAID-3, input data is divided in units of bits or bytes and stored in a plurality of disk devices. In RAID-2, the Hamming code is represented by R
In AID-3, parity is used as error correction data. RAID-4 and RAID-5 interleave data in sector units. In RAID-4, parity is stored in the same disk device, whereas in RAID-5, parity is distributed to a plurality of disk devices.

[0008] Of these RAID levels, the scheme most often employed in ordinary disk array systems is R
AID-3 and RAID-5. FIG. 18 shows the RAID
-3 represents an example of a configuration of a disk array device of a type-3,
FIG. 19 shows an example of the configuration of a RAID-5 type disk array device.

That is, in the RAID-3 type disk array device shown in FIG. 18, input data is divided into bytes, and each data divided into bytes is stored in a plurality of disk devices. Then, the parity as correction data is stored in a predetermined disk device. Here, the No. stored in a plurality of disk devices is used. Parities P _1-4 for the data of Nos. ₁ to ₄ and No. ₁ Parities P _{5 -8} for data _{5 to 8} are stored.

In the RAID-5 disk array device shown in FIG. 19, input data is divided into sectors, and the data is interleaved and distributed to a plurality of disk devices. In this case, the data A, E, and I are stored in the first disk device, the data B, F, and J are stored in the next disk device, the data C, G, and the parity P _IL for the data I to L are stored in the next disk device. Is stored. The next disk stores data D and K and parity P _EH for data E to H, and the last disk device stores parity P for data A to D.
_AD and data H and L are stored.

On the other hand, with the development of computer systems in recent years, it has become possible to handle multimedia data typified by images and sounds, and the storage for these multimedia data is called a multimedia server. There is. Since the multimedia server requires a large capacity and a high transfer rate, it is often realized as a disk array device as a hardware system. However, while the conventional computer server emphasizes the average performance of processing, the multimedia server emphasizes how to minimize the worst time required to complete the processing. .

For example, in the case of a moving image, the movement becomes awkward unless 30 images per second are displayed one after another at regular intervals. Also, if the capacity of the disk device cannot keep up and the sound data is insufficient, the sound is interrupted and unpleasant noise is emitted. In the worst case, real-time performance is guaranteed if the upper limit that processing can be performed in this amount of time is guaranteed. It is also important for a multimedia server whether this real-time property can be guaranteed.

[0013]

Meanwhile, a disk storage device such as a hard disk has a format method called zone bit recording (ZBR). Under the condition that the recording density is constant and the rotation speed of the disk is constant, as shown in FIG. 20, the number of sectors per track can be larger on the outer periphery than on the inner periphery. With ZBR, the recording surface is divided into a plurality of zones according to the distance from the center of the disk, and the number of sectors per track is increased in the outer zone than in the inner zone to increase the recording capacity of the disk. Formatting method.

FIG. 21 shows the relationship between the cylinder address and the number of sectors per track in a disk storage device formatted using the ZBR method. In this case, the outer track has, for example, 176 sectors, the inner track has, for example, 132 sectors, and the outer track has one more track than the inner track. The number of sectors per track is increasing.

In the disk storage device of the ZBR system, the rotation speed is always constant even though the number of sectors is smaller on the inner circumference than on the outer circumference. There is a problem that stored data takes longer to read. This means that ZB
The same applies to a disk array device using an R-type disk storage device. Since the read time varies depending on the storage location of the sub-blocks obtained by dividing the data, the time during which the disk array device can transmit data also increases. Is determined by Similarly, the data writing time differs depending on the storage location.

FIG. 22 shows how data is distributed to the disk storage devices when data is stored in the RAID-3 system. In this example, the fixed-length data and the error correction data are written so as to be sequentially filled from the outermost periphery of each disk. In this case, since the sub-blocks constituting the first data and the error correction data are stored in the outermost periphery, the read time or the write time is the shortest. Conversely, since the sub-block and the error correction data that constitute the last data are stored in the innermost circumference,
The read or write time is the slowest. As described above, the tendency that the access time varies depending on the storage location is not particularly suitable for applications in the multimedia field where real-time guarantee is required.

Further, as shown in FIG. 17, in a server using a disk storage device of the ZBR system, in order to make the reading time or the writing time uniform,
It is conceivable to divide the sub-blocks that compose the data into those that are stored from the outer circumference of the disk and those that are stored from the inner circumference, and determine the size of each of them so that the data read time and data write time are constant. Can be

When determining the size of each sub-block, the size is determined so as to be proportional to the number of sectors per track in the track where the sub-block is stored. In that case, the size of each sub-block is not constant. On the other hand, RAID-1 to RAID-5
In the method of generating error correction data in the conventional method including the method, there is a condition that all sub-blocks have the same size. Therefore, error correction data cannot be generated for sub-blocks having different sizes in the conventional method.

The present invention has been made in view of such a situation. In a disk array device, the time for reading data or the time for writing data can all be made uniform, and error correction data is added to each data. And improve the reliability of the data.

[0020]

According to a first aspect of the present invention, there is provided a disk array control method, comprising: generating error correction data from second data having different sizes; Device for storing the second data, the address on the disk device, the size of the second data, and the disk for storing the error correction data so that the time required for reading and writing the data for use becomes uniform. And determining an address on the device and the disk device.

According to another aspect of the present invention, there is provided a disk array control device, comprising: a generation unit configured to generate error correction data from second data having different sizes;
The disk device for storing the second data, the address on the disk device, the size of the second data, the error correction data, A disk device for storing data and a determination unit for determining an address on the disk device are provided.

In the disk array control method according to the first aspect, error correction data is generated from the second data having different sizes, and the second data and the error correction data are read out to the disk device. And a disk device for storing the second data and an address on the disk device so that the time required for writing is uniform,
The size of the second data, the disk device for storing the error correction data, and the address on the disk device are determined.

In the disk array control device according to the present invention, the generating means generates error correction data from the second data having different sizes, and the determining means sends the second data to the disk device. And a disk device for storing the second data, an address on the disk device, a size of the second data, and the like, so that the time required for reading and writing the error correction data becomes uniform.
The disk device storing the error correction data and the address on the disk device are determined.

[0024]

FIG. 1 is a block diagram showing a configuration of an embodiment of a multimedia server to which the present invention is applied.

Arrangement part (Block Allocato)
r) 1 (determining means), based on the input format parameters and the data from the ZBR (zone bit recording) table (ZBR Table) 8,
A block map (Block Map) 3 described later is created. The format parameters include the size S of one block of data, the number n for dividing the data, an optimal skew value (θ _skew ), and the like.

On the other hand, the ZBR table 8 is a table in which data arrangement locations are associated with the number of sectors (Sectors) per track (Track) at that location, and as a graph, as shown in FIG. Become. Therefore, the number of sectors on one track can be checked from the data arrangement location. In addition, a disk storage device (Disk) 6
A process of converting a physical sector address from -1 to 6-m to a logical sector address is required. In this embodiment, a SCSI (ANSI Small Computer System) described later is used.
(Interface) The conversion is performed by directly inquiring the disk storage devices 6-1 to 6-m via the drivers 5-1 to 5-m. In the following, the SCSI drivers 5-1 to 5-m are simply referred to as the SCSI driver 5 and the disk storage devices 6-1 to 6-m are simply referred to as the disk storage device 6, unless it is particularly necessary to distinguish them. I will.

The relationship between the logical sector address and the physical sector address of the disk storage device 6 will be described later with reference to FIG.

An error correction controller (ECC Cont)
(Roller) 9 (generating means) reads data from the data buffer 4, generates error correction data, and supplies the data to the data buffer 4 or when there is an error in a sub-block constituting the data read from the data buffer 4. The sub-block is corrected using the error correction data constituting the data.

After arranging the data, the arranging unit 1 determines
The block map 3 is created. Block map 3
Is, as shown in FIG. 2, where n sub-blocks obtained by dividing the k-th (arbitrary natural number not exceeding the total number of data) data and error correction data are stored on the disk. It shows. At the top of each data,
The error correction data is arranged, and the arrangement location can be determined from the disk storage device number D _kp , the starting logical sector address L _kp on the disk storage device, and the number S _{kp of} the sectors constituting the sub-block. The location of each sub-block can be determined from the disk storage device number D _ki , the starting logical sector address L _ki on the disk storage device, and the number S _{ki of} sectors constituting the sub-block.
Here, i is a natural number satisfying 1 ≦ i ≦ n.

When reading or writing data continuously, first, a plurality of data access requests are generated, a data number as information for identifying the requests, and a data buffer (Data Bu) in which the data is placed.
address) and a flag indicating read or write are stored in the access request buffer (Acses).
s Request Buffer 7).

Scheduler (Scheduler) 2
Reads the plurality of pieces of information from the access request buffer 7 at a predetermined time and refers to the block map 3 to
The sub-blocks constituting the requested data and the locations of the error correction data are checked. The location includes the disk storage device 6, the starting logical sector address on the disk storage device 6, and the number of sectors. The scheduler 2 issues an access request to the SCSI driver 5 that drives the relevant disk storage device 6. ing.

The data buffer 4 temporarily stores the input multimedia data and the like, and temporarily stores the multimedia data and the like read from the disk storage device 6.

Next, the operation of each unit when writing data to the disk storage device 6 will be described. First, data to be written to the disk storage device 6 is sequentially written to the data buffer 4. At this time, scheduler 2
Has information on how to divide data from the block map 3. FIG. 3A shows an example in which k-th data is divided into sub-blocks of different sizes. That is, the division size is the data of S sub-blocks # 1 of size S _k1, subblock # 2 size S _k2, subblock # 3 of size S _k3, and size in the sub-block # 4 in the S _k4 I do.

Here, the division of the sub-block will be described in more detail. For example, if the data of one frame data size S, 1 frame data size respectively of Sk _1, Sk _2, Sk ₃ and 4 sub-blocks # 1 of Sk _4, # 2, # 3 and # 4 To divide. As shown in FIG. 4, when one frame of data is composed of M × N pixels, one frame is divided into packets each having a predetermined number of pixels (L pixels). Then, as shown in FIG. 4, each divided packet is sequentially allocated to four sub-blocks. For example, packet A0 is assigned to subblock # 1, packet B0 is assigned to subblock # 2, packet C0 is assigned to subblock # 3, packet D0 is assigned to subblock # 4, and packet A1 is assigned to subblock # 1. B1 is allocated to sub-block # 2, packet C1 is allocated to sub-block # 3, and packet D1 is allocated to sub-block # 4. In this way, by allocating each packet to each sub-block in order, it is divided into sub-block # 1, sub-block # 2, sub-block # 3 and sub-block # 4 as shown in FIG.

The error correction controller 9 reads these sub-blocks from the data buffer 4 in accordance with an instruction from the scheduler 2, and generates error correction data based on the read sub-blocks. In most cases, the size of the sub-block is not constant, and the size of the error correction data is set to be equal to the size of the largest sub-block among the divided sub-blocks.

Basically, the j-th error correction data P
_j is generated from the j-th data of each sub-block. At this time, as shown in FIG. 3 (B), if the value of j is small, because j-th data corresponding to all the subblocks exist, the data of all the subblocks, error correction data P _j Is generated. On the other hand, when j becomes larger than a certain level, the j-th data may not exist in some sub-blocks as shown in FIG. In this case, in sub-blocks # 1 and # 3, j
No corresponding data exists.

In such a case, the j-th data is read from the sub-blocks # 2 and # 4 having a size of j or more,
The error correction data _Pj is generated based on the data. Alternatively, predetermined data is added to the sub-blocks # 1 and # 3, and the error correction data _Pj is generated from the data of all the sub-blocks as in the case of FIG. You can also. Although there are various types of algorithms for generating error correction data, they are not described here in detail because they need not be described in detail. FIG. 5 shows parity generated from a sub-block composed of a plurality of packets.

The error correction data generated by the error correction controller 9 is written to the data buffer 4. After that, the SC instructed by the scheduler 2
When the SI driver 5 writes the sub-blocks on the data buffer 4 and the error correction data to the corresponding disk storage device 6, a series of write processing ends.

Next, the operation of each unit when reading data will be described. First, an access request is stored in the access request buffer 7. The scheduler 2 reads an access request from the access request buffer 7 and supplies it to the corresponding SCSI driver 5. SCSI
The driver 5 reads the sub-block and the error correction data designated by the access request from the corresponding disk storage device 6 based on the access request supplied from the scheduler 2, and writes them to the data buffer 4.

The error correction controller 9 reads out these sub-blocks and the data for error correction written in the data buffer 4 in accordance with the instruction from the scheduler 2 and determines whether there is an error in the sub-block based on the error correction data. Judgment is made, and if there is an error, correct it.

For example, when the j-th data of a predetermined sub-block is incorrect, the error is basically corrected using the j-th data of another sub-block and the error correction data. For example, as shown in FIG. 6A, when the value of j is small, the error correction data is generated from the j-th data of all the sub-blocks. Error-free sub-blocks # 1 to #
3 using all j-th data, sub-block # 4
J is corrected.

On the other hand, as shown in FIG. 6B, when the value of j is increased to some extent, error correction data is generated from the j-th data of the sub-block having a size of j or more. Correction is performed using the above-described sub-blocks having no error and the error correction data. The sub-block in which the error has been corrected is written to a designated location on the data buffer 4, and sent out as correct data outside the disk array device.

Here, the relationship between the logical sector address and the physical sector address of the disk storage device 6 will be described with reference to FIG. As shown in FIG. 7 (B), the disk storage device 6 is generally designed to be able to access each area called a sector. A sector is usually about 512 bytes to 4 kilobytes in size, and a donut-shaped area in which the sectors are arranged on a track is a track. Further, as shown in FIG. 6A, a plurality of recording media (Media) are stacked. A cylindrical area in which tracks at the same distance from the center of a circle are collected is called a cylinder. As described above, the address including the cylinder number, the media number, and the sector number is the physical sector address.

On the other hand, in recent years, in a drive of the SCSI specification which is most widespread, a serial number is given to all sectors in the drive, and data is accessed using the serial number. This serial number is the logical sector address.
As described above, the arrangement unit 1 realizes the conversion between the physical sector address and the logical sector address by directly inquiring the disk storage device 6 via the SCSI driver 5. Of course, using another method, a correspondence table between the physical sector addresses and the logical sector addresses as shown in FIG. 8 is prepared, and the conversion between the physical sector addresses and the logical sector addresses is performed using the correspondence table. You may.

As shown in FIG. 8A, a logical sector address (Logical Sector) L _ki is a cylinder number (Cylinder) CYL _ki , a media number (Media) MED _ki , and a sector number (Sect).
or) Corresponds to the physical sector address composed of SEC _ki . FIG. 8B shows a specific example. Therefore, according to this correspondence table, a logical sector address can be converted into a physical sector address, or a physical sector address can be converted into a logical sector address.

FIG. 9 shows that the skew θ _skew which is the angle difference between the start positions of adjacent sub-blocks on the disk device and the gap θ _gap which is the angle difference between the end position and the start position of the sub-block are made constant. The example which performed allocation was shown. When such allocation optimization is performed, the rotation waiting time can be further reduced by combining with the SCAN algorithm for changing the order of disk access requests so as to reduce the moving amount of the head. For details on this, see “Japanese Patent Application No.
282175 ". Note that this allocation is merely introduced to make the above embodiment easier to understand, and the present invention can be sufficiently applied to an allocation that does not use a skew or a gap.

Next, with reference to the flowcharts of FIGS. 10 to 15, a description will be given of a processing procedure when the arranging unit 1 of FIG. 1 allocates the k-th data to a predetermined one of the disk storage devices 6.

First, in step S1, n + 1 disk storage devices 6 for storing error correction data and n sub-blocks are selected from the m disk storage devices 6 in all. In this example, first, a disk device for storing the error correction data is selected, and then, a disk device for arranging the sub-blocks on the inner periphery and a disk device for arranging the sub-blocks on the outer periphery are selected.

When sub-blocks are numbered 1 to n, for example, odd-numbered sub-blocks
The even-numbered sub-blocks are arranged from the outer periphery. Further, the disk storage devices 6 to be arranged according to the data are shifted and selected so that the sub-blocks are not concentrated on a small number of disk storage devices 6.

FIG. 11 is a flowchart showing details of this processing. That is, in step S11, the j-th disk storage device 6 represented by the following equation (1) is selected as the disk storage device 6 in which the error correction data is arranged.

J = MOD (k−1, m) +1 (Equation 1) where MOD is an operator for calculating a remainder obtained by dividing k−1 by m (the total number of disk storage devices 6). is there.

Next, in step S12, the disk storage devices 6-i (i = 1, 3, 5,
...), the j-th disk storage device 6-i represented by the following equation (2) is selected.

J = MOD (k + i−1, m) +1 (Equation 2)

Then, the process proceeds to a step S13, wherein the disk storage devices 6-i (i = 2, 4, 6,.
..), the j-th disk storage device 6-i represented by the above equation (2) is selected.

In the flowchart shown in FIG. 11, the sub-blocks arranged from the inner periphery and the sub-blocks arranged from the outer periphery are determined alternately. However, the sub-blocks may be arranged alternately if the balance is maintained. That is, the sub-blocks are sequentially arranged in the disk storage device 6 from the inner circumference, and then
The remaining sub-blocks may be sequentially arranged in the disk storage device 6 from the outer periphery.

If the sub-blocks are not concentrated on a small number of disk storage devices 6, the j-th disk storage device 6 represented by the above equations (1) and (2) may not be selected. For example, with respect to the disk storage device 6 in which the sub-blocks are arranged at the inner periphery, the disk storage device 6 which uses less the inner periphery is used. Storage device 6 that uses less storage
The method of selecting from is also effective. In any case, the error correction data is arranged from the outer periphery on the disk storage device 6 selected in step S11.

As described above, when the disk storage device 6 is determined, the process returns, and then proceeds to step S2 in FIG.

In step S2, on each of the selected disk storage devices 6, a starting logical sector in which error correction data and sub-blocks are to be arranged is determined.

FIG. 12 is a flowchart showing a processing procedure for determining the start logical sector. First, in step S21, a start logical sector address L _kp is obtained for the disk storage device 6 in which error correction data is arranged from the outer periphery.

FIG. 13 is a flowchart showing details of the processing in step S21 of FIG. First, in step S31, the outermost track is selected as a track for arranging a sub-block from an unarranged area where no sub-block is arranged. As a result, the cylinder number (CYL _ki ) and the media number (MED _ki ) of the physical sector address are determined, and the number of sectors (T _ki ) per track at this arrangement location can be found by referring to the ZBR table 8.

Next, in step S32, based on the cylinder number (CYL _ki ) and the optimum skew (θ _skew ), the head of the track selected in step S31 and the outermost track are determined by the following equation (3). The angle θ _ki between the head and the head is obtained.

Θ _ki = θ _skew × CYL _ki (Equation 3) However, when θ _ki > 360, θ _ki = θ _ki −360 is repeated until θ _ki <360.

Then, the process proceeds to a step S33, wherein the sector number (SE) is obtained by the following equation (4) from the angle θ _ki obtained in the equation (3) and the number of sectors per track (T _ki ).
C _ki ) is required.

SEC _ki = T _ki × θ _ki / 360 (Equation 4)

Next, in step S34, a logical sector address (L _ki ) is determined from the physical sector addresses (CYL _ki , MED _ki , SEC _ki ) obtained in steps S31 to S33, and the routine returns.

Next, the process proceeds to step S22 in FIG. 12, and for all the disk storage devices 6 arranged from the inner periphery,
Each starting logical sector address L _ki (i = 1, 3,
5, ...).

FIG. 14 is a flowchart showing details of the processing in step S22 in FIG. First, in step S41, the innermost track is selected from the unarranged area as the track on which the sub-block is arranged. Thus, the cylinder number (CYL _ki ) and the media number (MED _ki ) of the physical sector address are determined.
By referring to the BR table 8, the number of sectors per track (T _ki ) at this location is known.

Next, in step S42, the angle θ _ki between the head of the track selected in step S41 and the head of the outermost track is determined, and in step S43, the sector number (SEC _ki ) is determined. Then, in step S44, the obtained physical sector address (CY
The logical sector address (L _ki ) is determined from L _ki , MED _ki , and SEC _ki ), and the process returns. Step S described above
The processing in steps S42 to S44 corresponds to step S3 in FIG.
Since the processing is the same as that in steps S2 to S34, a detailed description thereof is omitted here.

Next, the process proceeds to step S23 in FIG. 12, and for all the disk storage devices 6 arranged from the outer periphery,
Each starting logical sector address L _ki (i = 1, 3,
5, ...). The processing procedure here is the same as the case described above with reference to the flowchart in FIG. 13, and a description thereof will be omitted here. Then, when the process ends, the process returns.

In the processing examples shown in FIGS. 12 to 14, when the sub-blocks are arranged on the tracks from the outer periphery, the track on which the sub-blocks are arranged is selected from the outermost track in the non-arranged area, and the sub-block is arranged. When arranging tracks on the inner track, the track on which the sub-block is arranged is selected from the innermost track in the non-arranged area, but it is not always necessary to select the outermost track or the innermost track. Absent. For example, in the ZBR format, if the zone is the same, the number of sectors per track is the same, and the size and gap value of each sub-block do not change.

After the start logical sector address is obtained as described above, the process proceeds to step S3 in FIG. 10, and the size of each sub-block is determined. The processing procedure for determining the size of the sub-block is shown in FIG.
5 will be described later.

Next, what should be the size (the number of sectors) of each subblock should be considered. First, the sum of the sizes of each sub-block is equal to the size of the original data. That is, the size (the number of sectors) of each sub-block is set as S _ki (i =
1, 2, 3,..., N) and S is the size of the original data (the number of sectors), and is represented by the following equation (5).

S _k1 + S _k2 +,..., + S _kn = S (Equation 5)

On the other hand, in order for the reading time and the writing time of each sub-block to be equal, the value of the gap of each sub-block must be equal. That is, the following equation (6) must be satisfied.

S _k1 / T _k1 = S _k2 / T _k2 =,..., = S _kn / T _kn (Equation 6)

In the above equation (6), T _ki (i = 1,
(2, 3,..., N) is the number of sectors on the track where the i-th sub-block is arranged.

From equations (5) and (6), the number of sectors S _ki of each sub-block is given by the following equation (7).

S _ki = S × T _ki / T (Equation 7) where T = T _k1 + T _k2 +,..., + T _kn .

In practice, since the number of sectors S _ki is given by an integer, each size requires a slight fine adjustment.

On the other hand, the size S _kp of the error correction data is
When the size of each sub-block is determined, it is given by the following equation (8).

S _kp = MAX (S _k1 , S _k2 ,..., S _kn ) (Equation 8)

In equation (8), MAX is an operator for finding the largest one of the sub-block sizes S _{k1 to} S _kn .

The error correction data is stored in step S in FIG.
From step 1 and step S2, since it is guaranteed that the track is arranged on the track from the outer periphery of the disk storage device 6, it can be seen that the following equation (9) holds.

S _k1 / T _k1 = S _k2 / T _k2 =,..., = S _kn / T _kn = S _kp / T _kp (Equation 9) where T _kp is a sub-block S _kp. Is the number of sectors in the track.

FIG. 15 is a flowchart for realizing the above procedure. First, in step S51, the number of sectors T _ki (i = i) per track of the track corresponding to the location where the n sub-blocks are located.
, N) is obtained. Next, the operation represented by the above equation (7) is performed for the variable i from 1 to n, and the size S _ki (i = 1, 2 _,.
···, n)

Then, the process proceeds to a step S52, at which the operation represented by the above equation (5) is performed, and it is determined whether or not the sum of the sizes S _ki of the sub-blocks obtained at the step S51 is equal to the size S of the original data. Is determined. If it is determined that the sum of the sizes S _ki of the sub-blocks is not equal to the size S of the original data, the process proceeds to step S53, and the sum of the sizes S _ki of the sub-blocks is equal to the size S of the original data. _So that the size S _ki of each sub-block is slightly increased or decreased. Thereafter, the process returns to step S52, and the processes of steps S52 and S53 are repeated until it is determined that the sum of the sizes S _ki of the sub-blocks is equal to the size S of the original data.

On the other hand, if it is determined in step S52 that the sum of the sizes S _ki of the sub-blocks is equal to the size S of the original data, the flow advances to step S54 to determine the size S _kp of the data for error correction. Can be That is, the largest size in the sub-block S _ki is set as the size S _kp of the error correction data.

Upon completion of the process in step S54, the process returns. Thus, the processing in step S3 of the flowchart shown in FIG. 10 ends, and all the processing ends.

FIG. 16 shows an example of sub-blocks distributed to the disk storage device 6 as described above. Here, the number of disk storage devices 6 is 6, and the number of sub-blocks is 4. That is, as shown in FIG.
Even for sub-blocks having different sizes, it is possible to generate error correction data by the above-described method.

The error correction data (parity in this case) of the first data is allocated to the disk storage device 6-1 and the sub-blocks are allocated to the disk storage devices 6-2 to 6-5. You. Disk storage device 6
1 and the sub-blocks arranged in the disk storage devices 6-3 and 6-5 are placed in the outermost track and arranged in the disk storage devices 6-2 and 6-4. The sub-block is placed on the innermost track.

The error correction data of the second data is stored in the disk storage device 6-2, and the sub-blocks are stored in the disk storage devices 6-3 to 6-6. From the inner circumference, put in order from the outer circumference,
The error correction data of the third data is stored in the disk storage device 6-3, and the sub-block is stored in the disk storage device 6-4.
6-6 and the disk storage device 6-1 in the same manner in the order of outer periphery, inner periphery, outer periphery, inner periphery, and outer periphery. In the same manner, data is arranged, and the last data is stored in, for example, the disk storage device 6.
-3 to 6-6 and about the middle of the cylinder of the disk storage device 6-1.

When observing the error correction data and the sub-blocks arranged as described above by focusing on one disk storage device 6, the values of the gaps of the respective sub-blocks are almost the same. The time required for reading and writing is almost constant.

As described above, in the above embodiment, the time for actually reading or writing error-corrected data and sub-blocks is fixed regardless of the data storage location on the disk storage device 6. Can be.

In the above-described embodiment, a case has been described in which the control of reading and writing of data to and from the disk storage device 6 is realized using a dedicated controller. By operating on the CPU, it is possible to control the reading and writing of data from and to the disk storage device 6.

In the above embodiment, when accessing data, scanning is performed from the outer periphery to the inner periphery. However, scanning may be performed from the inner periphery to the outer periphery. In this case, it is possible to set an optimum skew when accessing data while moving the head from the inner circumference to the outer circumference.

Further, in the above embodiment, a hard disk is used as a medium for recording data, but other recording media such as an optical disk and a magneto-optical disk may be used.

[0097]

According to the disk array control method of the first aspect and the disk array control apparatus of the sixteenth aspect, error correction data is generated from the second data having different sizes, and the error correction data is generated in the disk device. On the other hand, the disk device storing the second data, the address on the disk device, and the size of the second data so that the time required for reading and writing the second data and the data for error correction becomes uniform. And the address on the disk device for storing the error correction data and the address on the disk device, so that the time required for reading and writing the error correction data and sub-blocks is fixed regardless of the storage location of the sub block. And real-time performance can be guaranteed while maintaining high reliability.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of a multimedia server to which a disk array device of the present invention is applied.

FIG. 2 is a diagram showing an example of a block map 3 of FIG.

FIG. 3 is a diagram illustrating a method for generating error correction data.

FIG. 4 is a diagram showing a state in which pixels of one frame are divided into a plurality of packets.

FIG. 5 is a diagram showing a configuration of a sub-block.

FIG. 6 is a diagram showing a method for restoring error data.

FIG. 7 is a diagram showing physical sectors of the disk storage device 6;

FIG. 8 is a diagram showing a relationship between a logical sector address and a physical sector address.

FIG. 9 is a diagram showing allocation in which a skew and a gap are made constant.

FIG. 10 is a flowchart illustrating a processing procedure for determining an arrangement location of k-th data.

FIG. 11 is a flowchart illustrating a processing procedure for selecting a disk storage device 6;

FIG. 12 is a flowchart illustrating a processing procedure for determining a start logical sector.

FIG. 13 is a flowchart illustrating a processing procedure for determining a start logical sector from the outer periphery.

FIG. 14 is a flowchart illustrating a processing procedure for determining a start logical sector from the inner circumference.

FIG. 15 is a flowchart illustrating a processing procedure for determining a size.

FIG. 16 is a diagram showing how error correction data and sub-blocks are distributed.

FIG. 17 is a diagram showing a state where sub-blocks are distributed.

FIG. 18 is a diagram for explaining a RAID-3 system.

FIG. 19 is a diagram for explaining a RAID-5 system.

FIG. 20 is a diagram showing physical sectors of a ZBR disk.

FIG. 21 is a graph showing a relationship between the number of sectors per track of a ZBR disk and a cylinder address.

FIG. 22 is a diagram showing a conventional method for distributing data while keeping the size of a sub-block constant.

[Explanation of symbols]

1 placement unit (decision means), 2 scheduler, 3 block map, 4 data buffer, 5 SCSI driver, 6 disk storage device, 7 access request buffer, 8 ZBR table, 9 error correction controller (generation means)

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code FI G11B 20/12 G11B 20/12

Claims (16)

[Claims] 1. A disk array control method for dividing first data into a plurality of second data and storing the divided data in a plurality of disk devices, comprising: generating error correction data from the second data having different sizes; The disk device that stores the second data and the disk device and the disk device so that the time required to read and write the second data and the error correction data to and from the disk device are uniform. A disk array control method comprising: determining an address, a size of the second data, and an address on the disk device that stores the error correction data and an address on the disk device. 2. The number of the second data constituting the first data is represented by n, and the j-th one of the third data constituting the i-th second data is represented by A _ij . When
When the error correction data is generated based on A _ij selected from n or less different second data, and when each A _ij is generated at least once, any error correction data is generated. 2. The disk array control method according to claim 1, wherein the method is used. 3. The number of the second data constituting the first data is n, the size of the i-th second data is S _i (1 ≦ i ≦ n), and the i-th second data is _Aj (1 ≦ i ≦ n, 1 ≦ j ≦ S
when a _i), the size S _p of the error correction data,
The j-th data P _j (1 ≦ j ≦ _Sp ) of the error correction data is the maximum value of S _i and the j-th data P _j of all the second data satisfying j ≦ S _i 3. The disk array control method according to claim 2, wherein the data is generated from the third data _Aij . 4. The size of the first data is S, the size of each second data obtained by dividing the first data is S _i (1 ≦ i ≦ n), and the size of the second data is The size of the generated error correction data is S _p , the number of sectors in which each second data is included in a track arranged on the disk device is T _i (1 ≦ i ≦ n), and the error correction data is Assuming that the number of sectors included in a track arranged on the disk device is T _p , S ₁ + S ₂ +,..., + S _n = S S ₁ / T ₁ = S ₂ / T ₂ =,. 4. The disk array control method according to claim 3, wherein said first data is divided into respective second data so that S _n / T _n = S _p / T _p . 5. When the size of the second data is not constant, the size of the second data is the same as the size of the second data having the largest size of the second data. 2. The disk array control method according to claim 1, wherein predetermined data is added to the other second data. 6. The number of the second data constituting the first data is n, the size of the i-th second data is S _i (1 ≦ i ≦ n), and the i-th second data is _Aj (1 ≦ i ≦ n, 1 ≦ j ≦ S _i )
, The size S _p of the error correction data is S _i
When B _ij = A _ij (when 1 ≦ j ≦ S _i ) or B _ij = 0 (when S _i <j ≦ S _p ), the j-th error correction data P _j (1
The disk array control method according to claim 5, wherein? J? _Sp ) is generated from _Bij (1? I? N). 7. The size of the first data is S, the size of each second data obtained by dividing the first data is S _i (1 ≦ i ≦ n), and the size of the second data is The size of the generated error correction data is S _p , the number of sectors in which each second data is included in a track arranged on the disk device is T _i (1 ≦ i ≦ n), and the error correction data is Assuming that the number of sectors included in a track arranged on the disk device is T _p , S ₁ + S ₂ +,..., + S _n = S S ₁ / T ₁ = S ₂ / T ₂ =,. 7. The disk array control method according to claim 6, wherein said first data is divided into respective second data such that S _n / T _n = S _p / T _p . 8. A first group in which each second data obtained by dividing the same first data is arranged on a track from the outer periphery of each disk device, and The cylinder address and the size of the disk device on which the error correction data and the second data constituting the first set are arranged are the same or substantially the same, respectively,
8. The disk array control method according to claim 7, wherein the cylinder addresses and sizes of the disk devices in which the respective second data constituting the second group are arranged are the same or substantially the same. 9. The number n of each second data divided from the same first data, and the number no of the second data arranged on a track from the outer periphery of the disk device is no. The number of tracks arranged on the track from the inner circumference is n
When i, no + ni = n, and the number of the disk devices for arranging the error correction data and the second data on tracks from the outer periphery is (no + 1) out of a total of m disk devices, Out of the remaining (m-no-1) disk devices, ni disk devices for arranging the second data on tracks starting from the inner circumference are ni outer disks or inner disks of a predetermined disk device. 9. The disk array control method according to claim 8, wherein the selection is made so that the second data is not arranged in a concentrated manner. 10. The total number of cylinders of each disk device is represented by L
When the cylinders are numbered from 1 to L in order from the outer circumference to the inner circumference, among the cylinders in which the second data on the j-th (1 ≦ j ≦ m) -th disk device is not recorded, The cylinder at the outermost position is Mo
(J) When the cylinder located closest to the inner circumference is defined as Mi (j) (1 ≦ Mo (j) ≦ Mi (j) ≦ L), the error correction data and the second data When selecting a disk device to be arranged on a track, only (no + 1) devices with the smallest Mo (j) are selected in order from m disk devices in total, and the second data is transferred from the inner circumference. When selecting a disk device to be arranged on a track, M (no-no-1) disk devices are selected from among the (m-no-1) disk devices.
10. The disk array control method according to claim 9, wherein only ni items having a larger i (j) are selected in order. 11. The number of each second data segmented from the same first data is represented by n, and among the second data,
When the number of tracks arranged on the outer track of the disk device is no and the number of tracks arranged on the inner track is ni, no + ni = n, and the second data is stored on the track from the inner track. The disk devices to be arranged are n out of a total of m disk devices.
The disk devices for arranging i pieces of the error correction data and the second data on tracks from the outer periphery are replaced by (no + n) out of the remaining (m−ni) disk devices.
9. The disk array control method according to claim 8, wherein 1) the number of the second data is selected so as not to be concentrated on the outer circumference or the inner circumference of the predetermined disk device. 12. When the total number of cylinders capable of storing the second data of each disk device is L and the cylinders are numbered from 1 to L in order from the outer circumference to the inner circumference, j (1 ≦ 1) j ≦ m) the second disk device on the
Of the cylinders in which no data is recorded, Mo (j) is the cylinder located at the outermost periphery and Mi (j) is the cylinder located at the innermost periphery (1 ≦ Mo (j) ≦ M
i (j) ≦ L), and when selecting the disk device in which the second data from the inner circumference is to be arranged, the disk devices having the largest Mi (j) are sequentially selected from n disk devices out of m in total.
When selecting only the i disk devices and the disk device in which the error correction data and the second data are arranged from the outer periphery, Mo (Mo-n) disk devices are selected from among the (m-ni) disk devices.
12. The disk array control method according to claim 11, wherein only (no + 1) items having a smaller (j) are selected in order. 13. The first data k (= 1, 2,...)
..) each of the divided second data is SB _i
If (1 ≦ i ≦ n), SB _i is based on the value of i,
It is determined whether the track is located closer to the outer circumference or the inner circumference of the track. When j = MOD (k + i−1, m) +1, SB _i is the j (1 ≦ j ≦ m) -th When j = MOD (k−1, m) +1, the disk device that stores the error correction data is stored in the j (1 ≦ j ≦ m) th disk device. 9. The disk array control method according to claim 8, wherein: 14. The disk array according to claim 13, wherein the second data SB _i is arranged alternately from the outer circumference and the inner circumference of the track based on the value of i in order. Control method. 15. The error correction data and the second data arranged on a track from the outer periphery of the disk device, wherein the error correction data and the second data on each disk device are not recorded. The processing of arranging the track on the outermost track among the tracks and the second data of the track arranged on the inner track of the disk device are performed by using the error correction data and the second data on each disk device. A process of arranging the error correction data and the second data on the disk device in turn is alternately repeated, and the error correction data and the second data are sequentially arranged on the disk device. The disk array control method according to claim 9, wherein 16. A disk array control device that divides first data into a plurality of second data and stores the divided data in a plurality of disk devices, wherein error correction data is generated from the second data having different sizes. Means, the disk device and the disk storing the second data so that the time required for reading and writing the second data and the error correction data from and to the disk device is uniform. A disk array control comprising: an address on a device; a size of the second data; a disk device for storing the error correction data; and a determination unit for determining an address on the disk device. apparatus.