JPH04302020A - Storage control system - Google Patents

Abstract

PURPOSE:To constitute a disk array by which each disk device can be operated in parallel by preparing a parity code for the set of records on a partial area on the disk device. CONSTITUTION:Even at the time of storing the variable length type records on disk devices a803 and b803, the partial areas 104 of the disk devices a803 and b803 are equal when the device types of the disk devices a803 and b803 are the same. Therefore, parity data (d) are prepared for a data record 800 of the partial area 104 of each disk device a803 and b803, so that the parity code can be prepared from the records whose length is different. Thus the disk array by which each of disk device a803 and b803 can be operated in parallel, even when each of disk device a803 and b803 stores the record having the variable length.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a storage control system suitable for disk arrays.

0002

2. Description of the Related Art The following article by Patterson is known as the closest known example to the invention. A. C. M. Sigmod Conference Proceedings, June 1988, pages 109-116 (D.
Patterson, et al: A Case for
r Redundant Array of Inex
Pensive Disks (RAID), ACM S
IGMOD conference procedure
ng, Chicago, IL, June 1-3, 198
8, pp. 109-116) Patterson's paper discloses a technology for arranging data on a disk array.A disk array physically uses multiple small-capacity disk devices, while a single disk device serves as a processing unit. It is a mechanism designed to increase performance/reliability. Patterson's paper proposes several data arrangement methods, and one typical data arrangement method is the following arrangement method.

In this arrangement method, records, which are units of read/write with the processing device, are arranged as they are on the disk device, and are hereinafter referred to as record unit arrangement. (Among the data arrangement methods proposed in Patterson's paper, a method of dividing one record and placing it on multiple disk devices is also proposed.) In Patterson's paper, record The unit arrangement is called level 4 and level 5 data arrangement. A feature of the record unit arrangement is that read/write processing can be executed for each disk device that makes up the disk array.

(If one record is divided into multiple disk devices, it is necessary to exclusively use multiple disk devices to read/write data for one record.) Therefore, if the record unit arrangement method is used, , it becomes possible to increase the multiplicity of read/write processing that can be executed within the disk array, and performance improvement can be achieved. For this reason, disk arrays with record unit arrangement are considered to be a data arrangement suitable for transaction processing in which relatively short records are randomly accessed.

On the other hand, high reliability of disk arrays is achieved by storing redundant data called parity data in disk devices. In the record unit arrangement, one piece of parity data with an amount of data equivalent to one record is created from each record on a fixed number of different disk devices, and is stored on the disk device as one record. This is called a parity record. on the other hand,
A set of parity records and the records that generated them is called a parity group. Furthermore, records other than parity records that are directly read/written by a processing device are called data records. Usually, each record of the same parity group is stored in a separate disk device. However, the number of parity records in a parity group is
It is not limited to one in particular.

[0006] Even if a failure occurs in one of the disk units that stored the data record that generated the parity record, the contents of the parity record can be determined based on the contents of the parity record and other data records. The content of the stored data record is such that it can be restored. Therefore, even if a failure occurs in any disk device among the set of disk devices in which a certain parity group is stored, data can be recovered. Normally, if the number of parity records in a parity group is n, the data in that parity group can be recovered even if a failure occurs in up to n disk devices.

Furthermore, when the contents of the data record are changed by a write request from the processing device, the contents of the parity record must also reflect the changed value of the data record, and therefore an update process is required.

[0008]

[Problem to be solved by the invention] Patterson
In the record unit arrangement method proposed in the paper, the length of the record on the disk device was always fixed. However, some disk devices allow not only fixed-length record formats but also variable-length record formats. In particular, variable length record formats are often allowed in large magnetic disk drives. That is, in this method,
Records are taken out one by one from each of a plurality of disk devices, and a parity record is created from a set of these records. Therefore, it has been impossible to create a parity record unless the lengths of records retrieved from each disk device are equal. From the above, Patters
The record-by-record arrangement method proposed in the paper by On, was difficult to apply to disk devices that allow variable-length records to be stored.

An object of the present invention is to provide a storage control system having a storage control device for expanding the record unit arrangement proposed in Patterson's paper and allocating data to a plurality of disk devices that allow variable length record formats. Our goal is to provide the following.

0010

[Means for Solving the Problems] Hereinafter, it will be described how the present invention solves the problems described above. In the following explanation, it is assumed that parity records are stored in n disk devices from data records stored in m disk devices.

[0011] In a disk device that allows variable length record format, in order to create a parity record, for a set of records of different lengths on the disk device,
It is necessary to decide how to create parity records. In order to solve this problem, the present invention focuses on the following points. In other words, even in disk devices that allow variable-length record formats, tracks,
The physical areas on the disk device, such as cylinders, have the same size. Therefore, the idea is to create a parity record for a set of data records stored in a physical area on each disk device. In response to the above concept, the present invention provides the following two functions.

[0012] Integrated parity function: This function integrates a set of data records on partial areas in m disk devices in a manner corresponding to the disk devices, and considers them as m pieces of data. Therefore, n pieces of parity data are created for integrated data that is a set of data records on each partial area of m disk devices. Furthermore, each of these n pieces of parity data is stored as one parity record in each of the n disk devices.

[0013] Integrated division type parity function...This function also has the following functions:
A set of data records on partial areas in m disk devices is integrated in correspondence with the disk devices, and is considered as m pieces of data. Therefore, for integrated data that is a set of records on each partial area of m disk devices, n pieces of parity data are created and each is stored in n disk devices. This is similar to the conversion function. The reason for providing the integrated split parity function in addition to the integrated parity function is as follows. When a certain data record is updated using the integrated parity function, it is necessary to update the entire corresponding long parity record on the disk device. Therefore, there is a problem in that the overhead during update processing is large. To solve this problem, the integrated split parity function divides parity data into multiple records and stores them on each disk device. In this case, since it is only necessary to update some of the divided parity records in response to the update of the data record, there is an advantage that the overhead of update processing is small. However, it is generally necessary to provide an interval called a gap between records stored on a disk device, and if the number of divisions is increased, a large area is required to store parity data.

As described above, in order to create a parity record corresponding to a set of records of different lengths on a disk device, a set of data records is first integrated, and then a parity record is created. therefore,
Overhead and the like occur when integrating a set of data records. On the other hand, even in a disk device that allows a variable-length record format, when considering partial areas, there are areas where data records have the same length. For such areas, it is more efficient to use the record unit arrangement proposed in Patterson's paper.

Therefore, the second problem to be solved by the present invention is to
The purpose is to create an efficient parity record. To achieve this, a function is required to create efficient parity records by recognizing partial areas on the disk device where records of different lengths actually exist and partial areas on the disk device where each record has the same length. is necessary. In order to solve the above problems, the present invention provides the following two functions.

Control unit side record length characteristic recognition function: When creating parity data from a set of records on each disk unit, the control unit of the disk unit is provided with a function to recognize the length of the record on each disk unit. establish. As a result, the control device performs the integrated parity function, the integrated split parity function, or the Patt
Decide which method of record unit arrangement proposed by John erson will be used.

Record length characteristic addition function on the processing device side: Normally, the processing device side manages information representing characteristics such as the length of records on partial areas on the disk device. Therefore, when the processing unit issues a read/write request to the control unit or instructs the creation of a parity record, this function reads information indicating the record length characteristics of each partial area of the disk unit. Add to write request. According to this additional information, the control device determines which of the above-mentioned integrated parity function, integrated divided parity function, or record unit arrangement proposed by Patterson is to be used.

[0018]

[Operation] The operation of the present invention will be described. The present invention relates to the creation of parity data for a disk device that allows variable length record formats.

First, the operation when a parity record is created for a set of records of different lengths on a disk device will be explained. In this case, the control device creates n pieces of parity data for a set of data records stored in physical areas on m disk devices. Normally, in a disk device, if the device type is the same, the physical areas such as tracks and cylinders on the disk device are the same size, so even if the length of each record is different, the parity data corresponding to each physical area is It is possible to create

[0020] The unique functions of the integrated parity forming function and the integrated dividing parity forming function will be explained below.

The operation corresponding to the integrated parity function will be explained below. As already mentioned, the control device creates n pieces of parity data. In the integrated parity function, each of the n pieces of parity data is stored as one parity record in each of the n disk devices.

Furthermore, when a certain data record is designated as an update target by the processing device, the control device updates the parity records on n disk devices corresponding to the partial area on the disk device in which this data record is stored. Also updated. As described above, parity records can be created and maintained even for records of different lengths.

The operation corresponding to the integrated and divided parity function will be explained. The integrated split parity function differs from the integrated parity function in that one parity data is divided into a plurality of parity records and stored in corresponding disk devices. Therefore, when a certain data record is designated as an update target by a processing device, the control device divides it into multiple sections on each of n disk devices corresponding to the partial area on the disk device where this record is stored. All you have to do is change the contents of some of the parity records that are currently available. As described above, when the integrated and divided parity conversion function is used, more efficient processing can be achieved than when the integrated parity conversion function is used.

Next, the operation of creating an efficient parity record according to the record length characteristics of each partial area of the disk device will be explained. In this case, the above integrated parity function or integrated split parity function is applied to partial areas on the disk device where data records of different lengths actually exist, and if the record lengths are equal, applies Patterson's record unit arrangement. This allows efficient Pa
tterson's record-by-record arrangement can be applied to areas where it can be applied, making it possible to create parity records efficiently as a whole. The unique functions of the record length characteristic recognition function on the control device side and the record length characteristic addition function on the processing device side will be explained below.

Using the record length characteristic recognition function on the control device side, the control device identifies whether the partial areas on the disk device include data records of different lengths or the partial areas where the record lengths are the same. According to this result, it is determined whether to use the integrated parity function, the integrated split parity function, or the record unit arrangement proposed by Patterson.

On the other hand, by using the record length characteristic addition function on the processing device side, the processing device obtains information indicating whether there are partial areas on the disk device where data records of different lengths exist or partial areas where the record lengths are the same. Notify the control device. In accordance with this notification information, the control device determines which method to use: the above-mentioned integrated parity function, integrated split parity function, or record unit arrangement proposed by Patterson. As described above, by using either the record length characteristic recognition function on the control device side or the record length characteristic addition function on the processing device side,
Parity records can be created efficiently.

[0027]

[Examples] Examples of the present invention will be described below. First, the premise of the present invention will be explained.

FIG. 2 shows the configuration of a computer system to which the present invention is applied. The computer system includes a processing device 200,
It consists of a control device 204 and one or more disk devices 205. The processing device 200 includes a CPU 201 and a main memory 20.
2, channel 203. The control device 204 includes a control memory 207 and a work memory 208.
, a parity generation device 209. The control device 204 is
Data transfer between the processing device 200 and the disk device 205 is performed. Alternatively, the controller 204 includes one or more directors 206, each director 206
00 and the disk device 205 may be transferred.

The control memory 207 stores control information. The control memory 207 may be non-volatile.

The working memory 208 is a memory that temporarily stores data read out from the disk device 205 by the control device 204.

The parity generation device 209 will be explained later.

FIG. 3 shows the configuration of the disk device 205. The disk 308 is a medium for recording data, and a plurality of disks 308 are rotating bodies that exist in one disk device 205. A read/write head 309 is a device that reads and writes data to and from the disk 308, and is provided corresponding to the disk 308.

A circular recording unit that can be accessed by the read/write head 309 during one revolution of the disk 308 is called a track 300. A plurality of tracks 300 exist on the disk 303. When a certain track 300 becomes an input/output target, the read/write head 309 is moved to a position where this track 300 can be read/written.

FIG. 4 shows an example of the structure of data in the track 300. On track 300, one or more records 301 exist. The record 301 is the control device 20
4 and the processing device 200. The track 300 has a track head 302 and a track tail 303. The position of the record 301 on the track 300 is expressed in units of fixed-length bytes called cells 304.

(The record 301 always starts storing from the beginning of the cell 304, and does not start storing from the middle of the cell 304.) The number of the cell 304 is the first 3 of the track.
The cell 304 on 02 is numbered 0, and the numbers are added in ascending order of 1 up to the end of the track 303. Immediately after the track head 302, HA (Home Address)
An area called ss) 306 in which control information regarding the track 300 is recorded is provided. HA 306 is an area with a certain fixed length.

There are intervals between each record 301 and between each field within each record. This interval is called a gap 307. In particular, the gap between records 301 is called gap a305. The length of each record 301 may be different. However, the lengths of the gaps a305 are generally the same.

The record 301 immediately after the HA 306 is usually not a record 301 for storing general data, but often stores control data used on the processing device 200 side. Hereinafter, this record 301 will be changed to the control record 31.
Call it 0. The control record 310 has record number 602
is 0, and the lengths of the key section 607 and data section 601 contained therein often take standard values. Hereinafter, the control record 310 in which the record number 602 is 0 and the lengths of the key part 607 and data part 601 are standard values will be referred to as a standard control record.

FIG. 5 shows an example of the structure of data in the record 301. The configuration example in FIG. 5 is an example in which a record 301 is composed of three fields 606 called a control section 600, a key section 607, and a data section 601. Further, as shown in FIG. 6, the record 301 is
and a data section 601 in some cases. Control unit 60
0 stores control information, and the data section 601 stores information processed by a program or the like executed by the processing device 200. The key section 607 stores key information for checking the authority to access the contents of the data section 602.

A gap 307 also exists between the control section 600 and the data section 601. This gap 307 is called gap b605. Each gap b605 has the same length.

The control unit 600 has record numbers 602,
Key section 608, data section 603, defective position information 60
4th prize is included. Record number 602 is an identifier of this record 301.

The key section 608 and data section 603 have the same length as the key section 607 and data section 601. key manager 6
08 and the data length 603 are collectively referred to as a field length 609. (The length of the control unit 600 is usually a fixed length.)
If there is no key part 607 as shown in FIG.
09 takes 0. Therefore, on the disk device 205 according to the present invention, there are three records of different lengths.
01 is allowed to be stored, that is, record 3 in variable length format.
This allows storage of 01.

Normal write processing includes partial write processing that does not include rewriting the control section 600 and control section 600.
There is a format write process that includes rewriting. The data section 603 representing the length of the data section 601 is
Exists within 00. Therefore, in the case of partial write, although the contents of the data section 601 etc. can be rewritten, the length cannot be changed. on the other hand,
In the case of format writing, the lengths of the key section 607 and data section 601 of each record 301 can be changed, that is, the length of the record 301 can be changed.

When executing write processing, the processing device 200 specifies a record identifier in order to cause the control device 204 to identify the record 301 to be subjected to the write processing. The control device 204 compares the specified record identifier and the record number 602 in the control unit 600. In the case of partial write, records 301 for which a match result is obtained in this comparison are updated. On the other hand, in the case of format writing, the record 301 after the record 301 for which a match is obtained in this comparison is updated.

Next, the defective position information 604 will be described. Usually, there are places on the track 300 where data cannot be written, as shown in FIGS. 6 and 7. This location is called a defect 700. The defect position information 604 stores the positions of all defects 700 existing on the track 300. Above the defect 700 are fields 606 (control section 600, key section 607).
, data section 601) cannot be written. From the above,
In order to write the field 606 while avoiding the defect 700, the write position is calculated as follows according to the defect position information 604.

(1) As shown in FIG. 6, defect 5
The writing position of field 606 is shifted after 00.

(2) As shown in FIG.
6 (data section 601) and defect 500.
Write before and after.

As shown in FIGS. 6 and 7, a gap 307 (gap c
701) is recorded.

FIG. 26 shows the configuration of the control memory 207. The control memory 207 stores track unit information 2600
Consisted of. The track unit information 2600 is information provided for each track 300 on all disk devices 205 connected to the control device 204. However, if the capacity of the control memory 207 is limited, track unit information 2600 may be provided for each of a plurality of tracks 300.

FIG. 27 shows the structure of track unit information 2600. The track unit information 2700 is composed of track information a2700 and track information b2701. However, the specific contents of each will be described later.

FIG. 8 shows a storage format of records 301 on each disk device 205 in a disk array arranged in units of records for fixed-length records 301, which has been conventionally proposed in Patterson's paper.

As shown in FIG. 8, in the disk array, records 301 are classified into data records 800 and parity records 801. data record 80
0 is a record that stores data that is directly read/written by the processing device 200. (In other words, the data record 800 is a normal record 301 in a general disk device 205 other than a disk array.) On the other hand, the parity record 801 is a normal record 301 in a general disk device 205 other than a disk array.
This is a record 301 used for recovery processing when a failure occurs in 05.

From disk device a803 to disk device d
Corresponding data records 800 are stored on m disk devices 205 up to 806, respectively. Data records 80 on these m disk devices 205
Parity records 801 to be stored on n disk devices are created from 0 to the corresponding disk devices e807 to f808, respectively. Parity creation device 209 shown in FIG. 2 is used when creating parity record 801. Therefore, in FIG. 8, m data records 800 and n
parity records 801, parity group 8
02 is configured. Generally, n disk devices 2
If the parity record 801 is stored in the parity group 802, even if n disk devices 205 out of the m+n disk devices storing the record 301 in the parity group 802 fail, all records in the parity group 802 will be saved. The contents of 301 can be recovered.

In the parity group 802 in FIG. 8, a data record 800 is stored in a disk device a 803 to a disk device d 806, and a parity record 801 is stored in a disk device e 807 to a disk device f 808. However, it is not necessary that all records 301 stored on the disk devices 205 from disk device a 803 to disk device d 806 be data records 800. Similarly, not all records 301 on the disk devices 205 from disk device e807 to disk device f808 are parity records 801.

In addition, in FIG. 8, one parity group 802 is created from disk device a 803 to disk device f 808, but parity group 8
The set of disk devices 205 for which 02 is created may be different for each parity group 802. For example, another parity group 802 is disk device b804.
It may also be created in the disk device g809. Similarly, the number of disk devices 205 in which one parity group 802 is stored is not limited to m+n.

However, for the sake of simplicity, the following embodiment will be described assuming that one parity group 802 is stored in m+n disk devices 205, as shown in FIG.

In the configuration shown in FIG. 8, data records 800 are selected one by one from the m disk devices 205, n parity records 801 are created from these data records 800, and the n parity records 801 are created from the n disk devices 205. are stored one by one.

The above configuration is possible because the records 301 forming the parity group 802 have the same length. However, if the variable length format records 301 shown in FIGS. 4, 5, and 6 are allowed to be stored,
It is not possible to use the storage format of the parity record 801 shown in FIG.

[0058] The present invention provides a method for determining the parity of each disk device 205 when each disk device 205 constitutes a disk array that can operate independently in a disk device 205 that is permitted to store variable-length records 301. This relates to the storage format of the record 801.

Each embodiment will be explained below. If storage of variable-length records 301 is permitted, it is necessary to create parity records 801 from records 301 of different lengths. The first example and the second example are examples related to this creation. FIG. 1 is a diagram summarizing contents common to the two embodiments. On the disk device 205,
Even when storage of variable-length records 301 is permitted, if the disk devices 205 have the same device type, the capacities of the partial areas 104 of the disk devices 205, for example, the tracks 300, are the same. Therefore, as shown in FIG. 1, the concept of creating parity data d105 for a set of data records 800 on the partial area 104 of each disk device 205 is used. Thereby, the parity record 801 can be created from the records 301 having different lengths. The first example and the second example will be described below.

FIG. 36 shows an outline of the first embodiment. 1st
In this embodiment, the control device 204 selects the partial area 104 of each disk device 205 from m disk devices 205 that are permitted to store records 301 in variable length format.
One or more data records 800 stored in are extracted and n pieces of parity data d105 are created.

(106) Furthermore, the control device 204
One parity data d105 is defined as one parity record a101, and one
Store them one by one. (107) That is, a set of one or more data records 800 stored in the partial area of each disk device 205 is regarded as integrated data, and a parity record a101 is created for this integrated data. Therefore, hereinafter, this function will be referred to as the integrated parity function, and the parity group 8 created by this function will be referred to as the integrated parity function.
02 is called an integrated parity group 103.

In the integrated parity group 103 in FIG. 36, the disk devices a803 to d80
A data record 800 is stored in the disk device e
Parity record a from 807 to disk device f808
101 is stored. However, disk device a80
Disk devices 205 from 3 to disk device d806
All records 301 stored above do not need to be data records 800. Similarly, disk device e8
Disk device 20 from 07 to disk device f808
All records 301 on 5 are parity records a
It's not 101.

Furthermore, in FIG. 36, one integrated parity group 103 is created from disk device a 803 to disk device f 808, but the set of disk devices 205 on which parity group 802 is created is as follows:
It may be different for each parity group 802. Similarly, the number of disk devices 205 in which one integrated parity group 103 is stored is not limited to m+n.

The above processing is executed by the parity creation unit a100 in the control device 204 using the parity creation device 209. On the other hand, when processing a write request from the processing device 200, the record write unit a102
This is the processing unit used by 04.

First, when specifically describing the contents of the partial area 104 shown in FIG. 36, there are the following two ways of thinking.

(1) Consider the partial area 104 as a track 300.

(2) The partial area 104 is considered to be the area after the control record 310 in the area on the track 300. However, the concept (2) can be used on the premise that all control records 310 on the tracks 300 included in the parity group 802 are standard control records.

FIG. 9 shows the partial area 104 as track 300.
The storage format of the parity record 801 is shown below. Records 301 (data records 800
), n parity records aa900 to be stored on n disk devices are created respectively, and the corresponding disk devices e807 to f8 are respectively created.
08 are stored one by one. Disk devices 205 generally have the same physical structure if they have the same device type. Normally, when configuring a disk array using a plurality of disk devices 205, disk devices 205 of the same device type are selected. Therefore, the lengths of the tracks 300 of the disk devices 205 that constitute the disk array are equal. As already mentioned, since the length of the HA 306 is fixed, the length of the area in which the record 301 can be stored in each track 300 is equal. From the above, even if the lengths of the respective records 301 are different, the track 301
By considering the set of records 301 defined above as one piece of data, the storage format of the parity record aa 900 as shown in FIG. 9 can be realized.

FIG. 10 shows the partial area 104 on the track 30.
The storage format of the parity record 801 when considered as the area after the control record 310 in the area above 0 is shown. A control parity record that stores parity data to be stored on n disk devices from the control record (data record 800) on each corresponding track 300 of m disk devices 205 from disk device a 803 to disk device d 806. n pieces of 1000 are created, and each corresponding disk device e807
The files are stored one by one in the disk device f808. That is, for the control record 310, the same storage format as the parity record 801 shown in FIG. 8 is used. This is possible only with m disk devices 205.
This is because it is assumed that all control records 310 on the upper track 300 are standard control records. Furthermore, from the record 301 (data record 800) stored after the control record 310, a parity record a to be stored on n disk devices is added.
n b1001 are created and stored one by one from the corresponding disk device e807 to disk device f808. From the above, even if the lengths of the records 301 stored after the control record 310 are different, the set of records 301 (data records 800) stored after the control record 310 can be considered as one data. , the storage format of the parity record ab1001 as shown in FIG. 10 can be realized.

Next, data used in creating parity record aa900, parity record ab1001, ie, parity record a101 in the first embodiment will be explained. In this embodiment, the following two types of data are considered.

(1) Records 30 on each partial area 103
A parity record a101 is created using only the data on each field 606 in 1 (data record 801).

(2) Records 30 on each partial area 103
A parity record a101 is created by combining the data on each field 606 in 1 (data record 801) and the gap 307 stored on its partial area 103.

FIG. 11 shows the internal structure of parity record a101 when parity record a (parity record aa900 or parity record ab1001) is created without including gap 307 as shown in (1). In this case, as shown in FIG. 11, each field 606 of each data record 800 is filled with n
parity data a1100 are created. The created parity data a1100 is parity record a10.
1 data section 601 in the disk device 205. In this case, the control unit 60 of the parity record a101
The data section 603 in 0 contains parity data a110.
A length of 0 is stored. (The key length 608 becomes 0.) The defective position information 604 of the track 300 in which the parity record a101 is stored is stored as is. The record number 602 is not particularly specified. However, parity record ab1001
, it is preferable to use a number other than 0 in order to avoid duplication of control parity record 1000 and record number 602. In the storage format shown in FIG.
Parity data a1100 is created in a form padded with 0s. Therefore, the value of each data record 800 is
In order to identify in which position of the parity record a101 the data is reflected, it is necessary to read the data records 800 of the entire track 300. From the above, when the data record 800 on a certain track 300 becomes a write target, in order to determine which position of the parity record a101 should be changed, the data record 800
Data record 8 of the entire track 300 storing 00
It is necessary to read out 00.

FIG. 12 shows the gap 307 as shown in (2).
This shows the internal structure of the parity record a101 when the parity record a101 (parity record aa900 or parity record ab1001) is created. In this case, the gap 307 between each field 606 of each data record 800 and actually recorded on the track 300 is also included.
n pieces of parity data b1200 are created. Therefore, the parity data b1200 is created using the data stored on the track 300 itself. In this case, in order to know in which position in the parity record a101 the value of each data record 800 is reflected, it is only necessary to know the number of the cell 304 on the track 300 where the data record 800 is stored. . Therefore, data record 8 on a certain track 300
When 00 becomes the write target, parity record a1
In order to determine which value in 01 should be changed, read the data record 800,
Cell 304 where the data record 800 is stored
All you have to do is recognize it. How the parity data b1200 is incorporated into the parity record a101 is the same as the case shown in FIG. 11, so a description thereof will be omitted.

Furthermore, regarding the gap 307, it is not necessary to incorporate the pattern itself recorded on the track 300. Although the length must be maintained, the value itself is not particularly specified. However, a desirable form is to use a value that is distinguishable from the record 301.

Furthermore, the parity data b shown in FIG.
If the creation format of 1200 is converted, the format shown in FIG. 13 can be considered. In the format shown in FIG. 12, track 3 is used as data for creating parity record b1200.
The very gap 307 recorded on 00 was used. On the other hand, in the format shown in FIG.
00, a virtual storage format 1301 is set that is composed of each field 606 and gap 305 recorded on the track 300, assuming that there is no defect 700 at all. Then, parity data c1300 is created from this virtual storage format. In this case as well, in order to know in which position of the parity record a101 the value of each data record 800 is reflected, it is necessary to separate the cell 304 on the track 300 where the data record 800 is stored and the defective position information 640. Bye. Figure 11 shows how the parity data c1300 is incorporated into the parity record a101.
Since this is the same as the case shown in FIG. 13, detailed description of FIG. 13 will be omitted.

On the other hand, FIG. 14 shows the data used when creating the control parity record 1000 in the first embodiment and the structure of the control parity record 1000. When creating the control parity record 1000, m from disk device a803 to disk device d806
Since the control record 310 on the track 300 corresponding to each of the three disk devices 205 is a standard control record, the record number 602 of each record 301 is 0,
The key length 608 is 0, and the data length 603 is a constant value. Therefore, the record number 602, key section 608, and data section 603 of the control section 600 of the control parity record 1000 inherit the values of the control section 600 of each control record 300 as they are. The defect location information 604 indicates the track 30 in which the control parity record 1000 is stored.
0 defective position information 604 is stored. On the other hand, n pieces of control parity data 1400 are created from the data section 601 of each control record 310, and each control parity record 1
It is stored on the disk device 205 as a data section 601 of 000.

FIG. 28 shows the structure of track information a2700. The track information a2700 is composed of the information shown below.

Field configuration flag 2800...The parity record a101 created on the relevant track 300 is
As shown in FIG. 11, it shows that it is created using the data on field 606.

Field/gap configuration flag 2801
...Parity record a created on the relevant track 300
101 is created with data on field 606 and gap 307, as shown in FIG.

Virtual storage format flag 2802: Indicates that the parity record a101 created on the track 300 is created using a virtual storage format, as shown in FIG.

Next, the processing contents of the parity record creation section a100 and the record write section a102 will be explained.

FIGS. 15, 16, and 17 show the processing flow of the parity record creation unit a100, which is executed by the control device 204 when creating the parity record a101. The parity record creation unit a100 creates a parity record a10 of one integrated parity group 103.
1 creation function. Therefore, when creating parity records a101 for a plurality of integrated parity groups 103, the parity record creation unit a101 repeatedly
Execute.

FIG. 15 shows that when issuing a positioning request to the disk device 205 storing the data record 800,
This is the processing flow to be executed.

At step 1500, controller 204:
A positioning request is issued to the disk device 205 that stores the data record 800. After this, the process is once terminated.

FIG. 16 shows a processing flow executed when a positioning request to the disk device 205 is completed.

In step 1600, the control device 204 checks whether the loading process of the data records 800 on the tracks 300 of the m disk devices 205 has been completed. If the load process has been completed, the positioning process to the disk device has been executed to write the parity record a101 to the disk device 205. Therefore, the process jumps to step 1605.

If the loading process has not been completed, the positioning process to the disk device has been executed to load the data record 800 on the track 300 into the working memory 208. Therefore, step 1601
Now, controller 204 loads data record 800 on track 300 into working memory 208 . Next, in step 1602, the control device 204 checks whether the data records 800 on the track 300 have been loaded into the working memory 208 from all m disk devices 205. If not, step 160
In step 3, the control device 204 issues a positioning request to the disk device 205, and ends the process once.

When the loading process of the data records 800 on the track 300 from all m disk devices 205 is completed, in step 1604, the control device 204
In order to create parity record a101,
Call the processing flow shown in 7.

In step 1605, the control device 204 stores the parity record aa900 created corresponding to the track 300 for which the positioning process has been completed, or the control parity record 1000 and the parity record ab100.
1 is written to the disk device 205. Step 160
6, the control device 204 controls n disk devices 205.
Check whether all writing is completed. If not completed, jump to step 1603. If completed, complete the process.

FIG. 17 is executed to create parity record a101.

First, in step 1700, the control device 2
04 determines whether to create a control parity record 1000. If not created, jump to step 1703.

When creating control parity records 1000, the control device creates n control parity records 1200 from control records 310 on m tracks 300 in step 1701. In step 1702, the controller controls record 3 after control record 310.
01 (data record 800) as parity record a
101 as the record 301 to create,
Jump to step 1704.

In step 1703, the control device selects all records 301 (data records 800) on track 300 as records 301 for creating parity record a101.

In step 1704, the control device 204 creates the parity record a101 by checking the gap 3.
Check whether 07 is used. If gap 307 is used, jump to step 1706. gap 3
07 is not used, in step 1705, step 1
702 or the data on the field 606 in the record 301 selected in step 1703, n pieces of parity data a1100 are created. Then, the control device 204 provides track information a270 corresponding to the track 300 storing each parity record 801.
Turn on the field configuration flag 2800 in 0 and turn off the other flags. After this, the process jumps to step 1709.

In step 1706, the control device 204 determines whether to perform conversion to the virtual recording format 1300 when the defect 700 does not exist when using the gap 307. If conversion is to be performed, the process jumps to step 1708.

[0097] If no conversion is to be performed, in step 1707, the control device 204 performs step 1702 or
n pieces of parity data b1200 are created using the data on the field 606 in the record 301 selected in step 1704 and the gap 307 actually stored on the track 300. Additionally, a track 300 stores each parity record a101.
The field/gap configuration flag 2801 in the track information a 2700 corresponding to is turned on, and the other flags are turned off. After this, the process jumps to step 1709.

When converting to the virtual recording format 1300, in step 1708, the control device 204 converts the record 301 and gap 307 selected in step 1702 or step 1704 to defect 700.
n parity data c1300 for the result of conversion to the virtual recording format 1300 when there is no parity data c1300
Create. Furthermore, each parity record a1
Turn on the virtual storage format flag 2802 in the track information a 2700 corresponding to the track 300 storing 01,
Turn off other flags.

In step 1709, the control device 204 creates n parity records a using n parity data a 1100, n parity data b 1200, or n parity data c 1300.
a900 or n parity records ab10
01 (n parity records a101) is created.

FIGS. 18, 19, and 20 show the processing flow of the record write section a102. FIG. 18 is a processing flow in the record write unit a102 that is executed when a write request is received from the processing device 200. In step 1800, the control device 204 issues a positioning request to the disk device 205 that stores the data record 800 to be written. In step 1801, the control device 204 once disconnects from the processing device 200. This post-processing is completed.

FIG. 19 shows a processing flow executed when a positioning request to the disk device 205 is completed.

At step 1900, the control device 204 checks whether the working memory 208 stores the value before updating of the data record 800 to be written. If it is stored, jump to 1909. If it is stored, in step 1901, the field configuration flag of the track information a2801 is referred to, and it is checked whether the gap 307 is included in the corresponding parity record a101.

If no gaps 307 are included, then in step 1902 controller 204 loads all data records 800 on track 300 into working memory 208 . If gap 307 is included, step 190
3, the data record 800 to be written is loaded into the working memory 208.

In step 1904, the control device 204 reconnects with the processing device 200 and performs the write processing requested by the processing device 200 between the processing device 200 and the disk device 20.
Execute between 5 and 5. However, at this time, the data written to the data record 800, that is, the updated value of the data record 800, is stored in the working memory 208. After this, in step 1905 again, the processing device 200
Separate.

[0105] In step 1906, the received write request is a format write and the control record 31
Check whether 0 is targeted for writing. If so, jump to step 1916.

[0106] In step 1907, the control device 204:
For each of the n disk devices 205, it is determined which part of the parity record aa900, or the control parity record 1000 and parity record ab1001 will be the updated part.

[0107] The control device 204 stores the parity record aa.
900, or in control parity record 1000 and parity record ab1001, step 1906
In order to load the record 301 that actually includes the updated part recognized in step 190 into the working memory 208,
At step 8, a positioning request is issued to the disk device 205. After this, the process is terminated once.

[0108] In step 1909, the control device 204 creates parity record a801 (parity record aa9).
00, or the updated values of control parity record 1000 and parity record ab1001) are stored in working memory 2.
Check if it was created in 08. If it has been created, positioning processing to the disk device has been executed in order to write this updated value to the disk device 205. Therefore, the process jumps to step 1913.

If this is not the case, the purpose of the positioning process to the disk device is to load the data into the working memory 208. Therefore, in step 1910, controller 204 controls record 3 on track 300.
01 (Parity record aa900, or control parity record 1000 and parity record ab10
01), the record 301 on the track 300 that actually includes the updated portion recognized in step 1907 is loaded into the working memory 208. Next, in step 1911, the control device 204 checks whether the loading process of all n disk devices 205 into the working memory 208 has been completed. If not completed,
Jump to step 1908.

[0110] When the loading process for all n disk devices 205 is completed, in step 1912, the control device 2
04, to create parity record a101,
The processing flow shown in FIG. 20 is called. Thereafter, the process jumps to step 1908 in order to write the created parity record a101 to the disk device 205.

In step 1913, the parity record aa900 created corresponding to the track 300 for which the positioning process has been completed or the control parity record 10 is
00 and parity record ab900, the record 301 that actually incorporates the updated value is stored in the disk device 205.
write to.

In step 1914, the control device 204 checks whether writing has been completed to all n disk devices 205. If not, step 19
Jump to 06. If completed, step 19
15 to report the completion of the write request to the processing device 200,
Finish the process.

In step 1916, the parity generator a
Transfer control to 100. This is because when the control record 310 is no longer a standard control record due to execution of the format write, the storage format of the parity record a101 must be changed. Therefore, the parity creation unit a100 is called and the parity record a101 is re-created. When the re-creation process is completed, the process jumps to step 1915.

FIG. 20 is executed to create an updated value for parity record a101. In step 2000, the control device 204 selects the value before and after the update of the data record 800 to be written stored in the working memory 208, the parity record aa900 determined in step 1907, or the control parity record. 1000 and the pre-update value of the updated portion of parity record ab1001, an updated value of the updated portion is created.

In step 2001, the control device 204 stores the created update value in the parity record aa900 in the working memory 206 or in the control parity record 1.
000 and the parity record ab1001. After this, the process ends.

FIG. 21 shows an outline of the second embodiment. Second
In this embodiment, from m disk devices 205 that are permitted to store variable-length records 301, one file stored in the partial area 104 of each disk device 205 is
Three or more data records 800 are taken out and n pieces of parity data d105 are created. (106) Furthermore,
Each is divided into p parity records b2100, and a set of n sets of p parity records b2100 is
One set is stored in each of the n disk devices 205. (21
04) That is, the difference from the first embodiment is that the parity data d105 is divided into p parity records b2100 and stored in the disk device 205. The above processing is performed by the control device 204 by the parity creation unit b21.
01 and the parity generation device 209. On the other hand, the record write unit b2102
This is a processing unit used by the control device 204 when processing a write request from.

[0117] The difference between the second embodiment and the first embodiment is as follows.
parity data d105 into p parity records b2
The data is divided into 100 parts and stored in the disk device 205. In the first embodiment, one data record 8
00 was designated as an update target by the processing device 200, it was necessary to rewrite the entire parity record a101, which occupies almost the entire data amount of the track 300. However, in the second embodiment, by dividing into p parity records b2100, when one data record 800 is designated as an update target by the processing device 200, the amount of data to be rewritten is reduced to approximately 1/p. can be reduced.

The concept of the partial area 104 is the same as that of the first embodiment, as shown below.

(1) Consider the partial area 104 as the track 300.

(2) The partial area 104 is considered to be the area after the control record 310 in the area on the track 300. However, the concept (2) can be used on the premise that all control records 310 on the tracks 300 included in the parity group 802 are standard control records.

FIG. 22 shows the partial area 104 on the track 30.
The storage format of the parity record 801 in the second embodiment when it is considered as 0 is shown. FIG. 9 in the first embodiment
The difference from the storage format shown in is that m disk devices 20 from disk device a803 to disk device d806
5 on respective corresponding tracks 300
1 (data record 800), parity record ba22 for storing p pieces on n disk devices each.
The point is that n×p pieces of 00 are created, and p pieces are stored from the corresponding disk device e807 to the disk device f808. Other than that, the storage format shown in FIG. 22 is the same as the storage format shown in FIG. 9, so the following explanation will be omitted.

FIG. 23 shows the partial area 104 on the track 30.
Parity record 8 in the second embodiment when considered as the area after the control record 310 in the area above 0
01 storage format is shown. Furthermore, from the record 301 (data record 800) stored after the control record 310, parity records bb2300 for storing p pieces each on n disk devices are created respectively by n×
p disk devices are created, each with a corresponding disk device e807
p pieces are stored in the disk device f808. Other than that, the storage format shown in FIG. 22 is the same as the storage format shown in FIG. 9, so the following explanation will be omitted.

Next, in the second embodiment, parity record ba2200, parity record bb2300,
That is, the data used when creating the parity record b2100 will be explained. In the second embodiment, as in the first embodiment, three types of data corresponding to FIGS. 11, 12, and 13 are considered.

FIG. 24 does not include the gap 307 like FIG. 11 in the first embodiment, and the parity record b2
100 (parity record ba2200, parity record bb2300) is created. In this case, the process is the same as the first embodiment until the parity data a1100 is created. However, in FIG. 24, parity data a1
Divide 100 into p pieces and create p parity records b2
Create 100. Each parity record b2100
The data length 603 in the control unit 600 stores the length of the parity record b2100. (key manager 60
8 becomes 0. ) In the defective position information 604, the defective position information 604 of the track 300 in which the parity record b2100 is stored is stored as is. Record number 6
02 is not particularly specified. However, it is desirable that the record numbers 602 of the p parity records b2100 do not overlap. Also, parity record bb230
Regarding the record number 602 of 0, in order to avoid duplication of the record number 602 with the control parity record 1000, it is desirable that the record number 602 is other than 0 so that there is no duplication of each other. Other than that, the structure shown in FIG. 24 is the same as the structure shown in FIG. 11, so the following description will be omitted.

Next, the parity record creation section b2101
The processing contents of the record write unit b2102 will be explained.

FIG. 25 shows the parity record creation section b21.
01 processing flow. Parity record creation part b2
101 is executed by the control device 204 when creating the parity record b2100. However, the processing flow of FIG. 15, which is the processing flow of the parity creation unit a100 in the first embodiment, is different from that of the parity record creation unit b2.
It can be used as is in the processing flow of 101.

On the other hand, the parity record creation unit b2101
When using , the processing flow executed by the control device 204 when the positioning request to the disk device 205 is completed is as follows.
This corresponds to the processing flow of the parity creation unit a100. The differences between the processing flow when using the parity record creation unit b2101 and the processing flow in FIG. 16 are shown below.

Difference 1 between the processing flow when using the parity record creation unit b2101 and FIG. 16...The record write unit a102 in FIG. 36 creates the parity record a101
(parity record aa900 or parity record ab1001), whereas the record write unit b2102 operates on parity record b2100 (parity record ba2200 or parity record bb2300) which is divided into p pieces.

Difference 2 between the processing flow when using the parity record creation unit b2101 and FIG. 16: The processing flow of FIG. 25 is called instead of the processing flow of FIG. 17.

[0130] Only the above points are different. Furthermore, the configuration of the processing flow is also the same. Therefore, the processing flow when using the parity record creation unit b2101 is as follows:
Parity record b2100 (parity record ba2
200 or parity record bb2300), and the processing flow in FIG. 25 is called instead of the processing flow in FIG. 17. From the above,
Here, explanation using the processing flow will be omitted.

Therefore, the processing flow of FIG. 25, which is the processing flow of the parity creation unit a100 in the first embodiment corresponding to FIG. 17, will be described below.

Steps 2500 to 2508 in FIG. 25 are equivalent to step 170 in FIG.
This corresponds to the steps from step 0 to step 1708.

Steps 2500 to 2508 in FIG. 25 and step 1700 in FIG.
The difference between the steps from to step 1708 is the same as the difference 1 between the processing flow when using the parity record creation unit b2101 described above and FIG. 16. Therefore, description of steps from step 2500 to step 2508 will be omitted.

In step 2509, the control device 204 divides the n parity data a 1100, the n parity data b 1200, or the n parity data c 1300 into p pieces, and generates n×p parity data. A record ba2200 or n×p parity records bb2300 (n×p parity records b2100) are created.

The differences between the processing contents of the record write unit b2102 and the record write unit a102 are as follows: Difference 1 between the process flow when using the parity record creation unit b2101 described above and FIG. The flow structure is also the same. Record light part b210
Another difference between the record write section a102 and the record write section a102 is that the record write section b is used instead of the record write section a100.
Control is transferred to 2100.

[0136] Therefore, the record write section b2102
The processing contents are shown in FIGS. 18, 19, and 20 in the first embodiment.
The word parity record a101 (or parity record aa900, or parity record ab1001) that appears in the processing content explained using parity record b2100 (parity record ba220)
0 or parity record bb2300). Based on the above, explanation using the processing flow will be omitted here.

Next, each partial area 1 of the disk device 205
An example of efficiently creating a parity record according to the length characteristics of the record 301 of 04 will be described. This basic idea is shown in FIG. 37. As shown in the first embodiment and the second embodiment, the disk device 205
When creating a parity record 801 for the above set of data records 800 of different lengths, overhead and the like occur when integrating the set of data records 800. On the other hand, considering the partial area even in the disk device 205 that allows variable length record format, data record 8
There are also areas where the length of 00 is the same. Therefore, for such areas, it is more efficient to use the record unit arrangement proposed in Patterson's paper. According to the above idea, as shown in Figure 37, there are areas where the record unit arrangement proposed in Patterson's paper is applicable.
It would be more efficient to apply the record unit arrangement proposed in the paper by n. In this case, Patterson
Several embodiments are possible depending on how the control device 204 recognizes whether or not the record unit arrangement proposed in the paper of 2003 is applicable. In the present invention, two embodiments, a third embodiment and a fourth embodiment, will be described.

FIG. 29 shows an outline of the third embodiment. Third
In the embodiment, the controller 204 controls the disk device 205.
recognize the length of the data record 800 in the partial area 104 of Decide whether to use the integrated split parity function explained in the example.

In the third embodiment, the control device 204
uses the parity creation unit c2900 to execute the following processing.

[0140] In step 2902, the control device 204
Tracks 300 of each of n disk devices 205
The upper record 301 (data record 800) is read into the working memory 208.

[0141] In step 2903, the control device 204 reads the data record 800 on the read track 300.
The characteristics are stored in the control memory 208 for each track 300. Furthermore, based on the characteristics of the data record 800 on the read track 300, the Patte shown in FIG.
Determine whether the record unit arrangement proposed by rson can be realized. If possible, step 2
At 904, the control device 204 controls the Patte shown in FIG.
A parity record 801 is created using the record unit arrangement proposed by Rson. If this is not possible, in step 2905, the control device 204 uses the integrated parity function described in the first embodiment or the integrated divided parity function described in the second embodiment.

On the other hand, the record write unit c2901 is a processing unit used by the control device 204 when processing a write request from the processing device 200.

In this embodiment, the Pa shown in FIG.
It is determined whether to use the record-by-record arrangement proposed by Robertson et al., the integrated parity function described in the first embodiment, or the integrated split parity function described in the second embodiment.

[0144] When using the record unit arrangement proposed by Patterson, when a certain data record 800 is updated, it is only necessary to update the parity record 801 with the same length as that data record 800, which makes updating efficient. Processing is possible. Therefore, for areas where an efficient storage format can be used, an efficient storage format is used, and for areas where this is not possible, the integrated parity function or the integrated parity function described in the first embodiment is used. By using the integrated division parity function described in the second embodiment, the overall efficiency is improved.

In this example, Patterson
Specific record storage conditions under which the record-by-record arrangement proposed by the authors can be used will be described below.

Record storage condition a...m tracks 30
For each data record 800 on
m data records 8 with the same storage order starting from 02
00, the following matters hold true. That is, the record number 602 and key length 60 of each data record 800
8, and the data section 603 are equal.

Record storage condition b...m tracks 30
The following matters hold true between each control record 310 on 0. That is, the record number 602, key length 608, and data length 603 of each data record 800 are the same. Further, regarding the data records 800 after the control record 310 on each track 300, the following matters hold true. The record numbers 602 are arranged in ascending order one by one starting from the data record 800 immediately after the control record 310. Furthermore, all data records 80 other than each control record 310 on m tracks 300
The key length 608 between 0 and the data length 603 are equal.

[0148] When record storage condition a or record storage condition b is satisfied, each parity record 801
In the key section 607 and data section 601, parity data created from the data in the key section 607 and data section 601 of the corresponding m data records 800 is stored as is. Control unit 6 of each parity record 801
The record number 602 of 00 stores the same record number 602 as the corresponding data record 800. The lengths of the key part 607 and data part 601 are set as they are in the key part 608 and data part 603. Defect location information 6
04 stores defective position information on the track 300 where the parity record 801 is stored as is. Therefore, each track 3 of m disk devices 205
Data records 800 extracted one by one from 00
Correspondingly, n parity records 801 are created, and one can be stored in each track 300 of n disk devices. As a result of the above, record storage condition a or
If record storage condition b is satisfied, Patterso
It is possible to use the record unit arrangement proposed by n.

FIG. 30 shows the format of track information b2701 in the control memory 207 when the record storage condition a is satisfied and the record unit arrangement proposed by Patterson is used. When record storage condition a is satisfied, if the record unit arrangement proposed by Patterson is used, each record 301
record number 602, key length 608, and
It is necessary to memorize the data section 603. Therefore, when record storage condition a is satisfied, Patter
When using the record-by-record arrangement proposed by J. Son, the control memory 207 needs to have sufficient capacity.

[0150] Storage record number 3000, storage key section 3001, and storage data section 3002 are stored in track 300.
This is information stored in units of records 301 above. Storage record number 3000, storage key section 3001, and storage data section 3002 are track 3 from control record 310.
It is assumed that the records 301 are stored in the order in which they are stored on 00. Also, memory record number 3000,
It is assumed that the number of areas for storing the storage key section 3001 and the storage data section 3002 is equal to the number of records 301 that can be stored on one track 300.

FIG. 31 shows the format of track information b2701 in the control memory 207 when the record storage condition b is satisfied and the record unit arrangement proposed by Patterson is used. When the record storage condition b is satisfied and the record unit arrangement proposed by Patterson is used, the control record 31
For a set of 0 and other records 301, record number 602, key length 608, and data length 60
You need to remember 3. Therefore, when the record storage condition a is satisfied, the capacity of the control memory 207 can be reduced compared to the case of using the record unit arrangement proposed by Patterson. Furthermore, if it is necessary to reduce the capacity of the control memory 207, it may be possible to store this information in multiple tracks. However, in this case, the condition is that the values of the information defined in FIG. 31 are the same for all the tracks 300 that are stored together.

The control storage record number 3100, control storage key section 3101, and control storage data section 3102 are the record number 602, key section 608, and data section 603 of the control record 310 of the track 300.

[0153] When the value of the fixed-length record flag 3103 is on, the control record 31 on the relevant track 300 is
The record number 602 of the data record 800 immediately after 0 is in ascending order one by one from number 1, and the control record 310
data record 800 key section 608 other than, and
This indicates that the data lengths 603 are equal.

[0154] When the fixed length record flag 3103 is on, the fixed storage key section 3104 and the fixed storage data section 3105 store data records 80 other than the control record 310.
It represents the value of the key length 608 and data length 603 of 0.

[0155] Hereinafter, the parity creation unit c2900 in FIG.
, the processing contents of the record write unit c2901 will be explained.

FIG. 32 is a processing flow of the parity creation unit c2900. When the parity creation unit c2900 creates the parity record b2100, the parity creation unit c2900
Executed by However, the processing flow of FIG. 15, which is the processing flow of the parity creation unit a100 in the first embodiment, can be used as is for the processing flow of the parity record creation unit b2101.

Therefore, the processing flow shown in FIG. 32, which corresponds to the processing flow shown in FIG. 16, of the parity creation unit a100 in the first embodiment will be described below.

Similar to FIG. 16, FIG. 32 is also executed to create parity record b2100. The processing flow in FIG. 32 is similar to the processing flow in FIG. 16. Therefore, only the differences between the processing flow in FIG. 32 and the processing flow in FIG. 16 will be described below. Note that in the processing flow of FIG. 32, the steps with the same step numbers as those in the processing flow of FIG. 16 have the same processing contents as those of FIG. 16.

In step 3200, the control device 204 checks whether the capacity of the control memory 207 is large. If not, jump to step 3204.

If it is larger, in step 3201, the characteristics of the read data record 800 on the track 300 are stored in the track information b2701 of the corresponding track 300 in the format shown in FIG.

[0161] In step 3202, the control device 204 reads the data record 801 on the read track 300.
Check whether the storage format satisfies record storage condition a. If not satisfied, jump to step 3208. If satisfied, in step 3203, the characteristics of the parity record 801 to be created are stored in the track information b2701 of the corresponding track 300 in the format shown in FIG. After this, the process jumps to step 3207.

In step 3204, the characteristics of the read data record 800 on the track 300 are stored in the track information b2701 of the corresponding track 300 in the format shown in FIG.

[0163] In step 3205, the control device 204 records the data record 801 on the read track 300.
Check whether the storage format satisfies record storage condition b. If not satisfied, jump to step 3208. If satisfied, in step 3206, the characteristics of the data record 800 on the read track 300 and the characteristics of the parity record 801 to be created are determined as shown in FIG.
It is stored in the track information b2701 of the corresponding track 300 in the format shown in FIG.

[0164] In step 3207, Patterso
Execute the process of performing the record unit arrangement proposed by n. In this embodiment, detailed explanation will be omitted.

[0165] In step 3208, the control device 204 uses the process shown in FIG. 17 or
Call the processing flow shown in FIG.

[0166] In step 3209, when the processing flow of FIG.
001 is written to the disk device 205. On the other hand, when the processing flow of FIG. 28 is called in step 3208, the parity record ba2200, or the control parity record 1000 and the parity record bb2300
is written to the disk device 205.

The other processing flow shown in FIG. 32 is as follows.
Since the contents are the same as those shown in FIG. 16, the following explanation will be omitted.

33 and 34 show the record write section c2
901 is a processing flow. FIG. 33 shows the processing device 200
This is the processing flow that is executed when a write request is received from.

In step 3300, control device 204 checks whether the received write request is a format write. If so, jump to step 3304.

[0170] In step 3301, the track information b2701 of the track 300 in which the parity record 801 to be updated by the write request is stored is checked, and it is determined that the area is arranged in record units as proposed by Patterson. Check the area. Patte
If the area does not have the record unit arrangement proposed by Rson, the process jumps to step 3303.

[0171] In the case of an area in which the record unit arrangement proposed by Patterson is adopted, in step 3302, the corresponding processing is executed. In this embodiment, detailed explanation will be omitted.

At step 3303, the record write section a102 and record write section b2120 are executed.

[0173] In step 3304, a positioning request is issued to the disk device 205 in order to execute format writing. After this, in step 3305, the connection with the processing device 200 is once disconnected.

FIG. 34 is a processing flow that is executed when the positioning process to the disk device 205 for executing format write is completed.

[0175] In step 3400, control device 204 reconnects with processing device 200.

In step 3401, the control device 204 executes the write process requested by the processing device 200. In step 3402, the control device 204 again disconnects from the processing device.

In step 3403, there is a possibility that the characteristics of the data record 800 of the track 300 have changed due to the execution of the format write. Therefore, in order to recreate the parity record 801 corresponding to the data record 800 to be written, the parity generation unit c2900 is executed.

After the rebuild process is complete, step 340
In step 4, the control device 204 reports completion to the processing device 200.

An outline of the fourth embodiment is shown in FIGS. 35 and 38. In the fourth embodiment, the control device 204 controls the processing device 2
00, apply the record unit arrangement proposed in Patterson's paper, or
It is determined whether to use the integrated parity function described in the first embodiment or the integrated divided parity function described in the second embodiment.

FIG. 35 shows the creation process of the parity record 801 in the fourth embodiment. In the fourth embodiment, when the processing device 200 issues a request to create the parity record 801 to the control device 204, step 3600
As shown in FIG. 3, the control device 204 is notified of the characteristics of the record 301 of the track 300 on the disk device 205 in which the parity group 802 is stored. The specific information is track information b shown in FIG. 30 or 31.
This is information stored in 2701. In the third embodiment, the control device 204 reads the record 301 of the track 300 on the disk device 205 in which the parity group 802 is stored, grasps its characteristics, and reads the track information b.
2701 was created. On the other hand, in the fourth embodiment,
Processing device 200 notifies control device 204 of this information.

[0181] The control device 204 uses the parity creation unit d3601 to execute the following processing in accordance with a request from the processing device 200 to create a parity record 801.

[0182] In step 3602, the control device 204 converts the characteristic information of the record 301 of the track 300 notified by the processing device 200 into the corresponding track information b2701.
to be memorized.

In step 3603, it is determined based on the characteristics of the record 301 of the track 300 whether the parity record 801 can be created using the record unit arrangement proposed by Patterson.

If so, in step 3604, the controller 204 arranges the parity record 801 according to the record unit arrangement proposed by Patterson.
Execute the process to create the . After completion, step 3606
Jump to.

[0185] Otherwise, in step 3605, the control device 204 controls the parity generation unit a100 or
The execution of the parity creation unit b2101 begins. After completion, step 3606 is executed.

In step 3606, the control device 204 reports completion to the processing device 200. FIG. 38 shows the operation when the control device 204 receives a write request from the processing device 200 in the fourth embodiment. When the processing 200 issues a write request, as shown in step 3607, the following specification information is added to the write request. In other words, track 3 to be processed by the write request
This is the characteristic of record 301 on 00, and moreover, the characteristic after execution of the process corresponding to the write request.

The control device 204 uses the record write creation unit d3608 to execute the following process in accordance with the request from the processing device 200 to create the parity record 801.

In step 3609, the control device 204 determines based on the additional information from the processing device 200 whether the write request changes the characteristics of the record 301 of the track 300 to be processed by the write request. If so, jump to step 3613.

In step 3610, it is determined whether the specification from the processing device 200 specifies information for which the parity record 801 is created according to the record unit arrangement proposed by Patterson.

[0190] If so, in step 3611, the control device 204 sets the parity record 8 in the record unit arrangement proposed by Patterson.
01 update processing begins.

Otherwise, in step 3612, the control device 204 transfers control to the record write section a100 or the record write section b2100.

In step 3613, the control device 204 stores information regarding the characteristics of the record 301 of the track 300 added from the processing device 200 in the corresponding track information b2701, and executes the write process requested by the processing device 200. . After this, parity record 801
Step 360 shown in FIG.
3 to step 3606 are executed.

[0193]

The conventional disk array using the record-by-record arrangement proposed in Patterson's paper is characterized in that the individual disk devices constituting the disk array can operate in parallel. However, conventionally P
The disk array using the record-by-record arrangement proposed in the paper by Atterson is intended for disk devices that store fixed-length records, but is not suitable for disk devices that allow variable-length records to be stored. Can not. The present invention creates a parity record for a set of records on a partial area on a disk device, so that individual disk devices can operate in parallel even in a disk device that allows variable length records to be stored. It is possible to configure a disk array that can be used.

[Brief explanation of drawings]

FIG. 1 shows a method for creating parity data for records of different lengths.

FIG. 2 shows a computer system targeted by the present invention.

FIG. 3 shows the configuration of a disk device.

FIG. 4: Track configuration.

[Fig. 5] Record structure.

FIG. 6: Storage format with fields shifted to the rear.

FIG. 7 shows a storage format in which fields are divided.

FIG. 8 is a storage format for records arranged in units of records.

FIG. 9 shows a record storage format when a partial area is used as a track.

FIG. 10 shows a record storage format when the partial area is the area after the control record.

FIG. 11 shows a storage format when a parity record is created using data on a field.

FIG. 12 shows a storage format when data on a field and a parity record recorded on a track are created.

FIG. 13 shows a storage format in which parity data is created using a virtual storage format assuming that there are no defects.

FIG. 14 shows data and structures used in creating control parity records.

FIG. 15 is a processing flow for issuing a positioning request to a disk device by the parity creation unit a.

FIG. 16 is an execution processing flow when the parity creation unit a completes a positioning request to the disk device.

FIG. 17 is a process flow for creating parity record a.

FIG. 18 is a processing flow for issuing a positioning request to a disk device in response to a write request.

FIG. 19 is an execution processing flow when positioning is completed when a write request is received.

FIG. 20 is a process flow for creating an update value for parity record a.

FIG. 21 shows an overview of the second embodiment.

FIG. 22 shows a record storage format when partial areas are tracks in the second embodiment.

FIG. 23 shows a record storage format when the partial area is the area after the control record in the second embodiment.

FIG. 24 shows the internal structure of parity record b.

FIG. 25 is a process flow for creating parity record b.

FIG. 26 shows the configuration of control memory.

FIG. 27 shows the structure of track unit information.

FIG. 28 shows the structure of track information a.

FIG. 29 shows an overview of the third embodiment.

FIG. 30 shows the structure of track information b corresponding to record storage condition a.

FIG. 31 shows the structure of track information b corresponding to record storage condition b.

FIG. 32 is an execution processing flow when the parity creation unit c completes a positioning request to the disk device.

FIG. 33 is a processing flow executed when a write request is received in parity creation unit c.

FIG. 34 is a processing flow when format write positioning processing to a disk device is completed.

FIG. 35 is an overview a of the fourth embodiment.

FIG. 36 is an overview of the first embodiment of the present invention.

[Figure 37] How to use record unit placement and integrated parity function.

FIG. 38 is an overview b of the fourth embodiment.

[Explanation of symbols]

100... Parity creation part a, 101... Parity record a, 102... Record write part a, 103... Integrated parity group, 104... Partial area, 105... Parity data d, 2100... Parity record b, 2101... Parity creation part b, 2102...Record write section b, 210
3...Integrated split parity group, 2900...Parity creation section c, 2901...Record write section b, 2900...Parity creation section c, 2901...Record write section c, 36
01...Parity creation section d, 3609...Record write section d.

Claims (22)

[Claims] 1. A storage control system comprising a processing device and a storage control device connected to a plurality of storage devices that read/write data on a record-by-record basis and store the data in a plurality of partial areas. The storage control device includes:
Create n pieces of redundant data from a set of the records stored in the partial areas on the m storage devices, and set one of the redundant data as one redundant record, respectively.
A storage control system comprising means for storing the redundant records one by one in n storage devices other than the m storage devices. 2. The storage device is a disk device, the partial area is a track, and the storage control device is configured to
n pieces of redundant data are created from a set of the records stored in the tracks on the m disk devices, and one of the redundant data is treated as one redundant record. 2. The storage control system according to claim 1, further comprising means for storing said redundant records one by one in another n disk devices. 3. The storage device is a disk device, the area in which records other than the first record in a track are stored is the partial area, and the storage control device comprises:
n pieces of redundant data are created from a set of records other than the first record in the track on the m disk devices, and one of the redundant data is treated as one redundant record, and each of the m disks is A storage control device comprising means for storing the redundant records one by one in n disk devices separate from the device. 4. The storage control device according to claim 1, wherein the write request targets an area in the redundant record that must be updated in response to a write request received from a processing device. A storage control device comprising means for recognizing a record in the partial area including the record by reading it. 5. The storage control device according to claim 1, when the storage device is a storage device storing a gap between the records, the portion on m storage devices Create n pieces of redundant data from a set of the records and the gaps stored in the area, and treat one of the redundant data as one redundant record, respectively.
A storage control device comprising means for storing the redundant records one by one in n storage devices other than the m storage devices. 6. The storage control device according to claim 5, wherein the write request targets an area in the redundant record that must be updated in response to a write request received from a processing device. A storage control device characterized by comprising means for recognizing the record based on a storage position within the partial area. 7. The storage control device according to claim 1, wherein the storage device is a storage device that stores gaps between the records, and when there is a defective part in the partial area, storing the gap in the defective location, moving the record later and storing it in the partial area,
Assuming that the defective location does not exist in the partial areas on the m storage devices, n pieces of redundant data are created from a set of the records stored in the partial areas and the gaps, and the redundant data is A storage control device comprising means for storing one of the redundant records as one redundant record in n storage devices different from the m storage devices, respectively. 8. The storage control device according to claim 7, wherein the write request writes an area in the redundant record to be updated in response to a write request received from a processing device. A storage control device characterized by comprising means for recognizing based on a storage position in the partial area. 9. A storage control device in which a read/write unit is a record and a plurality of storage devices are connected to store the record in a plurality of partial areas, wherein the record is stored in the partial areas on m storage devices. Create n pieces of redundant data from a set of the records, divide each of the redundant data into p redundant records, and divide the divided sets of p redundant records into the m storage devices. A storage control device comprising means for storing one set in each of n other storage devices. 10. The storage control device according to claim 9, wherein when the storage device is a disk device and the partial area is a track, the record stored in the track on m disk devices n pieces of redundant data are created from a set of , each of the redundant data is divided into p redundant records, and the divided set of p redundant records is stored in a separate set of n disk devices from the m disk devices. A storage control device comprising means for storing one set in each of the disk devices. 11. The storage control device according to claim 9, wherein the storage device is a disk device, and when the partial area is an area in which a record at the beginning of a track is stored, m n pieces of redundant data are created from a set of records other than the first record of the track on the disk device, each one of the redundant data is divided into p redundant records, and the divided p pieces of redundant data are divided into p redundant records. A storage control device comprising means for storing one set of redundant records in each of n disk devices other than the m disk devices. 12. The storage control device according to claim 11, wherein in response to a write request received from a processing device,
Recognizing the redundant record to be updated among the p redundant record sets stored on each of the storage devices by reading the record in the partial area that includes the record to be written by the write request. A storage control device comprising means. 13. The storage control device according to claim 9, when the storage device is a storage device storing gaps between the records, the portion on m storage devices Create n pieces of redundant data from a set of the records and the gaps stored in the area, divide each one of the redundant data into p redundant records, and set the divided p redundant records. A storage control device comprising means for storing one set of , in each of n disk devices other than the m disk devices. 14. The storage control device according to claim 13, wherein in response to a write request received from a processing device,
comprising means for recognizing the redundant record to be updated among the p redundant record sets stored on each of the storage devices based on the storage position within the partial area of the record to be written by the write request. A storage control device characterized by: 15. The storage control device according to claim 9, wherein the storage device is a storage device that stores gaps between the records, and there is a defective part in the partial area. , store the gap in the defective area, move the record later and store it in the partial area, and assume that the defective area does not exist in the partial area on m storage devices. Create n redundant data from a set of the stored records and the gaps, divide each one of the redundant data into p redundant records, and set the divided p redundant records, A storage control device comprising means for storing one set in each of n disk devices other than the m disk devices. 16. The storage control device according to claim 15, wherein in response to a write request received from a processing device,
comprising means for recognizing the redundant record to be updated among the p redundant record sets stored on each of the storage devices based on the storage position within the partial area of the record to be written by the write request. A storage control device characterized by: 17. A storage control device in which a read/write unit is a record and a plurality of storage devices are connected to store the record in a plurality of partial areas, wherein the record is stored in the partial areas on m storage devices. From the set of records, n
create p redundant data, divide each redundant data into p redundant records, and store the divided set of p redundant records on n storage devices different from the m storage devices. A first means for storing one set in each of the storage devices, and n pieces of redundant data are created from the records retrieved one by one from the m storage devices, and each of the redundant data is stored in one set. As a redundant record, in each of the n storage devices different from the m storage devices,
second means for storing said redundant records one by one; m
It is determined whether a set of the records stored in the partial area on the storage device of the second unit satisfies a certain storage condition, and according to the determination result, the first means or the second means
A storage control system comprising a third means for selecting one of the means. 18. The storage control device according to claim 17, when the record is composed of a control field and one or more other types of fields, and the control field includes a record number, m units of storage are provided. A set of the records stored in the partial area on the device has the same storage order in the partial area on m storage devices, the record numbers of the m records are the same, and m records The lengths of fields of the same type within are equal.
Determine whether the condition is satisfied for all records in the partial area, and according to the determination result,
A storage control device comprising a fourth means for selecting either the first means or the second means. 19. The storage control device according to claim 17, when the record is composed of a control field and one or more other types of fields, and the control field includes a record number, m units of storage are provided. A set of the records stored in the partial areas on the device is such that the record numbers of the m records stored at the beginning of each of the partial areas on m storage devices are the same, and
fields of the same type in records are equal in length,
The record numbers of records other than the record stored at the beginning of each of the partial areas are stored in ascending order starting from No. 1, and the records are stored at the beginning of each of the partial areas on m storage devices. The first means or the second means determines whether the lengths of fields of the same type in all records other than the m records stored in the file are equal.
A storage control device characterized in that it has a fifth means for selecting one of the means. 20. The storage control device according to claim 17, wherein when receiving a write request from a processing device in which the length of the field in the record may change, the storage control device executes the third means. A storage control device comprising sixth means for activating. 21. A storage control system comprising a storage control device and a processing device connected to a plurality of storage devices that read/write records and store the records in a plurality of partial areas, the storage control device comprising: From the set of records stored in the partial areas on m storage devices, n
create p redundant data, divide each redundant data into p redundant records, and store the divided set of p redundant records on n storage devices different from the m storage devices. A first means for storing one set in each of the storage devices, and n pieces of redundant data are created from the records retrieved one by one from the m storage devices, and each of the redundant data is stored in one set. As a redundant record, in each of the n storage devices different from the m storage devices,
a second means for storing the redundant records one by one, and a third means for activating the first means or the second means according to an instruction from the processing device, the processing device A storage control system comprising a fourth means for instructing whether to activate the first means or the second means. 22. The storage control device according to claim 21,
A storage control system comprising fifth means for activating the third means when a write request in which the length of the field in the record may change is received from a processing device.