JP2853624B2 - Data storage system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data storage system for storing various data using a storage device such as a disk device, and more particularly, to copying a data stored in one storage device to another storage device. The present invention relates to a data storage system that stores data to prepare for a system failure.

[0002]

2. Description of the Related Art In a system for storing data in a storage device such as a disk device, it is necessary to store the data in a redundant configuration so that the stored data is not lost even if a failure occurs in the storage device. Will be Conventionally, as a measure against a failure of a disk device, a method of storing data in a plurality of disk devices in a redundant configuration includes a RAID (redundant configuration).
and Array of Inexpensive Disks). RAID is generally known from level 1 to level 5.

[0003] Level 1 is to store the same contents as those stored in a normally used disk device in a failure disk device, and is called a mirror system. The mirror type data storage system includes a normal disk device for normal operation and a mirror disk device for failure countermeasures. When writing data from the host computer, the same data as in the normal disk device is also stored in the mirror disk device. When a failure occurs in the normal disk device, reading from the mirror disk device prevents a situation in which data cannot be read.

In levels 2 to 5, parity information is generated, and when a failure occurs in one disk device, the contents stored in the failed disk device are restored using the parity information with another disk device. Is what you do.

[0005] in Japanese Unexamined Patent Publication No. 4 -318 640, the data storage system to restore the stored contents of a disk device the failed the RAID system using the parity information is disclosed. When using parity information, usually
For example, data is stored in three of the four disk devices, and parity information corresponding to data stored in the other three disk devices is stored in the remaining disk devices. A disk device storing data and its parity information in this manner is called an ECC group. When one of the three disks storing data fails, the data of the failed disk device is restored based on the data of the remaining two normal disk devices and the parity information.

When the number of disk devices is eight, if the disk devices are divided into four ECC groups, the load for restoring data of a failed disk device becomes E.
They cannot concentrate on the CC group and perform restoration work efficiently. Therefore, in the prior art disclosed in Japanese Patent Laid-Open No. 4 -318 640, a storage area of each disk device is divided into a plurality of blocks, different for each block 4
An ECC group is formed by one disk device. As a result, the load of restoring the data of one disk device can be distributed to the remaining seven disk devices.

In a data storage system using parity information, it is necessary to calculate parity information every time the storage contents of the disk device are updated, and it is difficult to perform high-speed write processing. Further, when a failure occurs in the disk device, it is necessary to restore the data, it takes a long time until the data is read, and it may not be possible to cope with high-speed reading. On the other hand, in the mirror system, the capacity of a disk device to be prepared for a failure countermeasure is increased due to the high degree of redundancy, but there is an advantage that high-speed access is possible.

FIG. 8 shows an outline of the configuration of a data storage system using a mirror system which has been conventionally used. The disk device 301 normally operated and the disk device 302 for troubleshooting are respectively connected to a disk control device 304 via a bus 303. The disk control device 304 includes an interface circuit (not shown) for inputting and outputting data to and from each disk device via a bus, and a CPU (central processing unit) for controlling reading and writing of data. When a write command is issued from the disk control device 304 to the normal disk device 301, the data to be written is stored in the normal disk device 301 and simultaneously written into the failure disk device 302.

By writing the same data to the two disk devices, the normal disk device 301
And the contents stored in the failure disk device 302 are the same. Since the failure disk device 302 holds the same data as the normal disk device 301,
The disk device 302 is called a mirror disk.
When a failure occurs in the normal disk device 301, the disk control device 304 responds to the failure by reading data from the failure disk device 302.

[0010] The data storage system shown in FIG. 8 is intended for a failure occurring in a disk device. However, when a failure occurs in a disk control device which controls reading / writing of data from / to the disk device, the data storage system is normally used. In some cases, the contents of both disk devices for failure and failure cannot be read.

FIG. 9 shows an outline of the configuration of a data storage system capable of coping with a failure of a disk control device. In this system, a first normal disk device 311 and a first failure disk device 312
Is connected to the first disk controller 314 by a bus 313. The disk devices 311 and 312 are also connected to a second disk control device 316 via a bus 315. Similarly, the second normal disk device 3
21 and the second failure disk device 322
3 is connected to the second disk controller 316 and the bus 324 is connected to the first disk controller 3
14 is also connected. The first and second disk controllers 314 and 316 are connected to each other via a network 325. Further network 325
Is connected to a host computer 326 that issues a data read request or a data write request. The disk control device functions as a server, and the host computer functions as a client.

When data is written to the first normal disk device 311 by the first disk controller 314,
The same data is also written by the first disk controller 314 to the first failure disk device 312.
Similarly, when data is written to the second normal disk device 321 by the second disk control device 316, data of the same content is written to the second disk control device 316 by the second disk control device 316.
Is also written to the failure disk device 323. In this way, the first normal disk device 311 and the first failure disk device 312 hold the same contents.
The second normal disk device 321 and the second failure disk device 322 also hold the same contents.

It is now assumed that a request for reading data stored in the first ordinary disk device 311 has been made from the central management host computer 326 connected to the network 325. When the system is normal, the contents of the first ordinary disk device 311 are read out by the first disk control device 314 via the bus 313 and the host computer 3 is read out via the network 325.
26. When a failure has occurred in the first normal disk device 311, the first disk control device 3
14 reads, via the bus 313, data requested to be read from the first faulty disk device 312, which stores the same contents as the first normal disk device 311. Then, this is transferred to the host computer 326 via the network 325.

On the other hand, when a failure occurs in the first disk controller 314, the contents of the first ordinary disk device 311 are transferred to the second disk controller 31 via the bus 315.
6 and transferred to the host computer 326 via the network 325. In this way, stored data can be read even when a failure occurs in the disk control device. Similarly to this system, there is a data storage system that can cope with both a failure of a disk device and a failure of a disk control device and has a simpler bus configuration.

FIG. 10 shows an outline of a simple data storage system having a bus configuration capable of coping with both a failure of a disk device and a failure of a disk control device. The first disk control device 331 includes a first normal disk device 332 and a first failure disk device 3.
33 are connected via a bus 334. Also, the second
The second normal disk unit 342 and the second failure disk unit 343 are connected to the bus
44 are connected. The first disk controller 331 and the second disk controller 341 are connected to each other via a network 351. Further, the network 351 is connected to a host computer 352 that issues data read and write requests.

In this system, the first normal disk device 3 connected to the first disk controller 331
The second disk device 343 connected to the bus 344 of the second disk control device 341 is used as the mirror disk device corresponding to 32. The mirror disk device corresponding to the second normal disk device 342 connected to the second disk control device 341 is the same as the first disk device 333 connected to the first disk control device 331. Used.

The same data as the data to be written to the first normal disk unit 332 is also written to the second disk unit 343 via the network 351 and the second disk control unit 341. Further, the same data as the data to be written to the second normal disk device 342 is also written to the first failure disk device 333 through the first disk control device 331. When a failure occurs in the first disk control device 331, the contents of the first normal disk 332 are stored in the second disk device 3 for failure.
Since it is also stored in 43, it is read from the second disk device for failure through the second disk controller 341.

[0018]

As described above, the contents of a normally operated disk device connected to one disk controller are copied to a faulty disk device connected to another disk controller. If this is done, data can be read even if a failure occurs in either the disk device or the disk control device. However, when a failure occurs in a disk controller, the load on another disk controller having a mirror disk device corresponding to the disk device connected to the failed disk controller is doubled. There is. In the example of FIG. 10, when a failure occurs in the first disk control device, the contents of the first normal disk device are changed from the second failure disk device holding the same contents to the second disk control device. It will be read through the device. Since the second disk control device also has a role of reading the contents of the second normal disk device, it is necessary to read the data of these two disks, and the load is doubled. As a result, there is a problem that it takes time to read data.

Accordingly, an object of the present invention is to reduce the increase in the load on other disk control devices that perform reading and writing in the event that a failure occurs in the disk control device that controls reading and writing of the disk device. It is an object of the present invention to provide a data storage system capable of performing the above.

[0020]

According to the first aspect of the present invention, a plurality of unit servers connected by a predetermined network include first and second data storage means for storing data, and these data storage means. A disk control device for controlling the means, the disk control device transmits and receives data to be input / output to / from these data storage units to / from the network, and the communication unit transmits / receives data to / from the network. First writing means for writing the data to be stored in the first data storage means to the first data storage means when the data is received, and writing the data to the first data storage means by the first writing means When the same data is sent to a predetermined unit server other than itself through the network described above ,
According to the address written in the first data storage means,
Division sending means for dividing the data into equal parts and sending the divided data; and second writing means for writing the data sent from the division sending means of the other unit server to the second data storage means when the data is received through the network. when this time of reading and failure detection means for detecting the presence of failure of these units server, the data is that there faulty stored in the first data storage means included in the unit server detected by the failure detecting means With this same data instead of
The data after being divided into almost equal parts and transmitted is referred to as the first data.
Address corresponding to the address written to the data storage means
The data storage system is provided with failure readout means for reading out from the transfer destination second data storage means.

That is, according to the first aspect of the present invention, the plurality of unit servers connected by a predetermined network include first and second data storage means for storing data,
It has a disk controller for controlling these data storage means. The disk control device includes a communication unit, a first writing unit, a division sending unit, a second writing unit, a failure detection unit, and a failure reading unit. When reading the data stored in the first data storage means, the necessary data is read from the unit server storing the same data in the second data storage means. Since the storage contents of the first data storage means are divided and stored in the second data storage means of the other plurality of unit servers, even if a failure occurs in one unit server, the contents of the other unit servers are stored. The burden does not increase significantly.

According to the second aspect of the present invention, data is written in the first data storage means in file units.
The division sending unit distributes the storage contents of the first data storage unit to a plurality of predetermined destinations in units of a file written in the first data storage unit.

That is, in the second aspect of the present invention, the storage contents of the first data storage means are divided and stored in a plurality of other unit servers in file units.

According to the third aspect of the present invention, the storage area of the first data storage means is divided into a plurality of blocks.
Dividing delivery means partial contents stored in the first data storage means to a plurality of destination predetermined this block units
I divide and memorize it.

That is, according to the third aspect of the present invention, the storage area of the first data storage means is divided into a plurality of blocks,
The storage contents of the first data storage means are divided and stored in a plurality of unit servers in units of this block.

According to the fourth aspect of the present invention, the plurality of unit servers are divided into two or more groups each composed of a plurality of unit servers, and the plurality of predetermined destinations are different from each other in each group. Is set in the unit server.

In other words, in the invention according to the fourth aspect, the same data stored in the first data storage means of each unit server is stored in the first data storage unit of the plurality of other unit servers in the group to which the unit server belongs. And is stored separately in the second data storage means. Thus, since a failure of one unit server can be recovered in each group, failure of two or more unit servers can be dealt with in the entire data storage system.

According to the fifth aspect of the present invention, the plurality of unit servers are divided into two or more groups each composed of a plurality of unit servers, and the plurality of predetermined destinations are respectively assigned to other groups. Set in the unit server.

That is, in the invention described in claim 5, the same data stored in the first data storage means of each unit server is the second data of the unit server of a group other than the unit server. It is stored in the storage means.

[0030]

BEST MODE FOR CARRYING OUT THE INVENTION

[0031]

FIG. 1 shows an outline of the configuration of a data storage system according to an embodiment of the present invention. In this system, first to fourth disk controllers 12 to 15 are connected via a network 11. The first disk control device 12 has a first normal storage device 17 as a disk device used for normal operation via a bus 16 and a first failure storage device as a disk device provided for troubleshooting. 18 are connected. Similarly, the second disk controller 13 receives the second
The normal storage device 22 and the second failure storage device 23 are connected. The third disk controller 14 has a third normal storage device 25 and a third failure storage device 26 via a bus 24, and the fourth disk controller 15 has a fourth normal storage device 26 via a bus 27. The device 28 and the fourth fault storage device 29 are connected to each other.

A disk controller that controls reading and writing to a disk device, and a normal storage device and a failure storage device connected to the disk control device via a bus are collectively called a unit server. A plurality of unit servers connected to a network is called a cluster. Here, the first to fourth unit servers 31 to 34 form a cluster. In the system shown in FIG. 1, a hard disk device is used as a normal storage device and a failure storage device.
In addition, a floppy disk device, a silicon disk device, or the like can be used. If a write operation is not required, a CD-ROM can be used.

As a bus connected to each disk controller, a SCSI (Small Computer System Interface) bus is used. In addition, a serial bus such as a fiber channel or an ATM (Asynchronous Transfer M
It is also possible to use a network such as ode). The number of unit servers constituting the cluster may be three or more, but this embodiment shows a cluster in which four unit servers are connected to a network.

The data having the same contents as the contents stored in the first ordinary storage device 19 is stored in the first unit server 31 to which the data belongs.
And stored in the failure storage devices 23, 26, and 29 belonging to the second to fourth unit servers 32-34. The copy of the storage contents of the second normal storage device 22 is performed on the unit servers 3 other than the second unit server 32.
1, 33, and 34 are stored in a distributed manner in the failure storage devices 18, 26, and 29. Similarly, copies of the storage contents of the third normal storage device 25 are stored in the first, second, and fourth failure storage devices 18, 23, and 29, and the storage contents of the fourth normal storage device 28. The copies are distributed and stored in the first to third failure storage devices 18, 23, 26, respectively. In this way, the contents of each normal storage device are distributed and stored in the failure storage devices belonging to unit servers other than the unit server to which the normal storage device belongs.

FIG. 2 shows an outline of the configuration of the disk control device of the unit server. The disk controller 41 has a CP which performs a central function of read / write control.
U (central processing unit) 42 is provided. Various circuit devices are connected to the CPU 42 via a bus 43. Among them, a ROM (read only memory) 44 is a read-only memory in which various programs and fixed data are stored. RAM (random access memory) 45
Is a memory for storing various data temporarily required for executing the program. The network control device 46 is a circuit device for inputting and outputting various data and commands to and from a network. SCSI
The controller 47 is a control circuit for transferring data between the normal storage device 48 and the failure storage device 49. The normal storage device 48 and the fault storage device 49 are connected on the SCSI bus output from the SCSI controller 47.

FIG. 3 schematically shows an example of the storage contents of each storage device of the data storage system shown in FIG. The same parts as those in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted as appropriate. Here, a copy of the data stored in the normal storage device is distributed and stored in the fault storage devices of the other three unit servers. Is divided into three according to the number of storage devices for use. In other words, it is divided into divided storage areas that are one less than the number of unit servers (the number of disk control devices). Then, the identification names are assigned to the divided areas for each of the divided areas.

The first divided area 51 of the first ordinary storage device 17 has "D1-1", the second divided area 52 has "D1-1", and the third divided area 53 has "D1-1". Is "D1-
3 ”. The first to third divided areas 54 to 56 of the second normal storage device 22 are assigned“ D ”.
2-1 "," D2-2 ", and" D2-3 "are assigned to the first to third divided areas 57 to 59 of the third ordinary storage device 25. , “D3-2” “D3
-3 "in the first to third divided areas 61 to 63 of the fourth normal storage device 28," D3-1 "," D3-2 "," D
3-3 "are assigned as identification names. Similarly, the first to fourth failure storage devices 18, 23, 26, 2
Each of the nine storage areas is also divided into three.

When writing data to the first normal storage device 17, the divided area of the write destination is "D1-1" (5
If 1), the area "D1-
At the same time as writing "1", the same data is also passed to the second disk controller 13 through the network 11. The second disk controller 13 transfers the received data to the first disk storage device 23 in the second failure storage device 23. Similarly, the data to be written to the divided area “D1-2” (52) of the first normal storage device 17 is transferred to the third disk control device 14, and the third failure is generated. Is written to the first divided area 71 of the first storage device 26. When writing to the divided area "D1-3" (53) of the first ordinary storage device 17, the same data is written to the fourth failure storage device. 2
The data is also transferred and written to the first divided area 74 of No. 9.

Thus, the first ordinary storage device 1
7 is stored in the first normal storage device 1
7 and at the same time are distributed and stored in the second to fourth failure storage devices 23, 26, 29. FIG. 3 shows a case where the storage area of each storage device is divided into logical blocks used in SCSI. The storage capacities of the normal storage device and the fault storage device are the same, and the storage area of each storage device is divided into “N” logical blocks.

At this time, the maximum n that satisfies N ≧ 3n−1 is selected, and numbers “0” to “3n−1” are assigned to the respective logical blocks. Then, the range of the logical block number of each storage device from “0” to “n−1” is set as the first divided area,
The range of the logical block number from “n” to “2n−1” corresponds to the second divided area. Further, the range of the logical block number from "2n" to "3n-1" corresponds to the third divided area.

The same data stored in the first divided area (51) "D1-1" of the first normal storage device 17 is stored in the first divided region 67 of the second fault storage device 23. Is stored in Therefore, the data stored in the logical blocks “0” to “n−1” of the first normal storage device 17 is the logical block “0” to “n−1” of the second fault storage device 23. The block is duplicated. Similarly, the first
The data stored in the logical blocks “n” to “2n−1” as the second divided areas 52 of the normal storage device 17 are “0” to “n−” of the third failure storage device 26. 1 "
Has been duplicated in the logical block. In this manner, the failure storage device at the copy destination and its storage area can be associated with each other by the logical block number. Such a correspondence is registered in the ROM or RAM of the disk control device of each unit server.

The operation of the data storage system in which data is stored as shown in FIG. 3 when a failure occurs in the normal storage device or the disk control device will be described.

In the system shown in FIG. 3, the correspondence relationship between each divided area of each normal storage device, the faulty storage device to be copied to, and the storage area in the faulty storage device is represented by another unit. This is notified in advance to the disk control device of the server. For example, the first divided area 51 of the first ordinary storage device 17, that is, "0" to "n-
It is determined that the copy destination of the logical block of “1” is the first divided area 67 (the logical block of “0” to “n−1”) of the second failure storage device 23. The host computer 77 that is the data request source is connected to the network 11.

When the disk controller 12 fails, the host computer 77 communicates with the disk controller 13 as an alternative. When the disk controller 13 fails, the host computer 77 provides an alternative. Communication with is set in advance. Furthermore, it is preset that, when a failure occurs in the disk control device 14, the disk controller 15 communicates with the disk controller 15 as an alternative, and when the disk controller 15 fails, the disk controller 12 communicates with the disk controller 12 as an alternative. .

The disk control device of each unit server copies the data stored in each storage area of the failure storage device connected to its own bus based on the information sent from the other unit servers. The original ordinary storage device is stored in its own RAM. In addition, the host computer 77 that is the source of the data read request sends information indicating whether a failure has occurred in the disk control device of each unit server to the RA having the host computer 77.
Store it in M. When a failure occurs, an alternative disk controller is stored in an internal memory.

Now, it is assumed that a failure has occurred in the first disk controller 12. After detecting the occurrence of a failure in the first disk controller 12, the host computer 77, which has requested the data, sends a read request to the first disk controller when reading data from the first ordinary storage device 17. 1
To the second disk controller 13 set as an alternative to the second disk controller 13. The area to be read is represented by the number of a logical block in the first ordinary storage device 17.

The second disk controller 13 calculates a logical block number from a read request sent from the host computer 77. If the calculated logical block number is in the range of “0” to “n−1”, a copy of the data to be read to the second fault storage device 23 connected to the bus of the second disk control device 13 Is determined to be stored. Therefore, the second disk controller 13 connects the second failure storage device 2 connected to its own bus 21.
The corresponding data is read out from 3 and transmitted to the host computer 77 via the network 11.

On the other hand, if the number of the logical block requested to be read is in the range of “n” to “2n−1”, the second disk controller 13 sends a read request to the third disk controller 14 via the network 11. To transfer. The third disk controller 14 is connected to its own bus 24 based on the received read request, and stores the third failure storage device 2.
6 reads the corresponding data from the host computer 7
7 via the network 11. Similarly, if the logical block number is in the range of “2n” to “3n−1”,
The second disk controller 13 transfers the read request to the fourth disk controller 15. As a result, the corresponding data is read from the fourth failure storage device 29 by the fourth disk controller 15 to which the read request has been transferred, and sent to the host computer 77.

The detection of the occurrence of a failure in the normal storage device is performed as follows. Each of the first to fourth disk controllers 12 to 15 issues a data read request to the normal storage device connected to its own bus. Then, when a timeout error occurs in which a response does not arrive within a predetermined time from the normal storage device to which the read request is sent, or when there is no response, it is determined that the normal storage device has failed.

The occurrence of a failure in the disk control unit of the unit server is detected as follows. The disk controller of the unit server constituting the cluster sends an operation confirmation message to the disk controller of the next unit server at regular intervals, indicating that the disk controller is operating normally. If the operation confirmation message has not been received from the immediately preceding disk controller for a certain period of time or more, an inquiry is made to the immediately preceding disk controller to confirm normal operation. If there is no response to the inquiry from the immediately preceding disk control device, an inquiry about operation confirmation is further performed to the immediately preceding disk control device. Until a response is obtained in this way, an inquiry is made to confirm the normal operation one after another. Then, it is determined that a failure has occurred in the disk control device between the disk control device that has started the inquiry and the disk control device that has returned the response.

The method of distributing the data to the failure storage device may be stored in a failure storage device belonging to a unit server other than the unit server to which the normal storage device storing the data belongs. The order and the area are not limited to the example shown in FIG.

As an example, a storage area of a normal storage device can be physically divided, and each area can be allocated to a failure storage device belonging to another unit server. Further, a failure storage device may be assigned in units of a file to be written to the normal storage device.

When allocating a fault storage device using a logical block number, as described above, a logical block is allocated so as to be continuous by a predetermined number, and adjacent logical blocks may be allocated to different fault storage devices. it can. That is, when allocating a failure storage device in units of logical blocks used in SCSI, a value obtained by dividing the number of the logical block in which data is written in the normal storage device by the number of distributed failure storage devices is used as a reference. Allocate to.

For example, when the storage contents of one normal storage device are distributed and stored in three failure storage devices, the failure storage device to be stored based on the remainder obtained by dividing the logical block number by "3". Allocate devices. In this case, since the remainder takes three values of "0" to "2", data is distributed and stored in the copy destination in units of logical blocks by assigning a failure storage device to each of these values in advance. Can be.

When there are a plurality of normal storage devices connected to the unit server, each of the normal storage devices can be treated as a divided area of the storage device described above. For example, it is assumed that a cluster is configured by the first to fourth unit servers, and the first to third normal storage devices are connected to each unit server. At this time, the contents of the first normal storage device of the first unit server are copied to the failure storage device of the second unit server. Further, the contents of the second normal storage device of the first unit server are copied to the failure storage device of the third unit server, and the contents of the third normal storage device of the first unit server are copied to the fourth normal storage device. Is copied to the failure storage device of the unit server.

First Modified Example

In the embodiment described so far, the storage contents of the normal storage device of each unit server are distributed and stored in the failure storage devices of all other unit servers. However, in the first modification, , And are distributed and stored in some unit servers.

FIG. 4 shows an outline of the configuration of a data storage system according to a first modification of the present invention.
In this system, the first to sixth disk controllers 81
86 are connected to the network 87. The normal storage devices 101 to 106 and the failure storage devices 111 to 116 are connected to the respective disk control devices 81 to 86 via buses 91 to 96. In the data storage system in the first modified example, these are divided into a first partial cluster 121 and a second partial cluster 122. Data is distributed and stored in each partial cluster as in the embodiment. That is, in the first partial cluster 121, the copy of the contents of the normal storage device 101 connected to the first disk control device 81 is copied by the second and third disk control devices 82 belonging to the same partial cluster 121. , 83 are distributed and stored in the failure storage devices 112, 113 connected to the failure storage devices 112, 113.

As a result, in the embodiment, 1 is set for the entire cluster.
While only one unit server is allowed to fail, the first modification can handle up to one unit server in each partial cluster. Of course, the number of partial clusters is not limited to two, and a similar effect can be obtained with more than two.

Second Modification

FIG. 5 shows the outline of the configuration of the data storage system in the second modification. The same parts as those in FIG. 4 are denoted by the same reference numerals, and description thereof will be omitted as appropriate. The second modification is also similar to the first modification in the first modification.
And the second partial clusters 121 and 122 are divided into clusters. In the first modification, copies of the storage contents of the normal storage devices in each partial cluster are distributed and stored in the failure storage devices in the same partial cluster. On the other hand, in the second modification, a copy of the storage content of the normal storage device belonging to each partial cluster is created in the failure storage device belonging to another partial cluster.

The first to first groups belonging to the first partial cluster 121
A first storage device including third normal storage devices 101 to 103;
The storage contents of the normal storage device cluster 131 are copied to a second fault storage device cluster 132 composed of fourth to sixth fault storage devices 114 to 116 belonging to the second partial cluster 122. . Similarly, the fourth to sixth ordinary storage devices 104 belonging to the second partial clusters 122
Second ordinary storage device cluster 1 composed of
The storage contents of 33 are copied to a first failure storage device cluster 134 composed of first to third failure storage devices 111 to 113 belonging to the first partial cluster 121.

As described above, the storage contents of the normal storage device belonging to one partial cluster are distributed and stored in the failure storage device belonging to another partial cluster, so that a cluster composed of a plurality of partial clusters is obtained. It is possible to cope with a failure of a plurality of unit servers within the system. As shown in FIG. 5, when the number of partial clusters is three or less, recovery cannot be performed if failures occur simultaneously in a plurality of clusters. However, when the number of partial clusters is four or more, Can handle obstacles.

FIG. 6 shows an outline of a data storage system composed of four partial clusters. In this system, the first to twelfth disk control devices 14
1 to 153 are connected to the network 154. Each of the disk controllers 141 to 153 has a bus 1
The normal storage devices 181 to 193 and the failure storage devices 201 to 213 are connected via 61 to 173. The clusters configured via the network 154 include first to first clusters.
It is divided into fourth partial clusters 221-225.

Here, the storage contents of the first ordinary storage device cluster 231 composed of the ordinary storage devices 181 to 183 belonging to the first partial cluster 221 are stored in the failure storage device belonging to the second partial cluster 222. 204-206
Are stored in a distributed manner in the second failure storage device cluster 232 composed of Also, the second partial cluster 22
The storage contents of the second ordinary storage device cluster 233 composed of the ordinary storage devices 184 to 186 belonging to the second
Storage devices 207 to 207 belonging to the partial cluster 223 of FIG.
209, the third storage cluster for failure 23
4 and stored.

Similarly, the storage contents of the third ordinary storage device cluster 235 belonging to the third partial cluster 223 are distributed and stored in the fourth failure storage device cluster 236 belonging to the fourth partial cluster 224. You. The storage contents of the fourth normal storage device cluster 237 belonging to the fourth partial cluster 224 are distributed and stored in the first failure storage device cluster 238 belonging to the first partial cluster 221. In this way, by setting the copy destination of the storage contents of the normal storage device cluster of each partial cluster to the failure storage device cluster of the next partial cluster, it is possible to cope with a case where a failure occurs in a plurality of partial clusters although there are restrictions. . That is, it is possible to cope with a failure occurring simultaneously in non-adjacent partial clusters.

For example, it is possible to cope with a case where a failure occurs simultaneously in the first partial cluster 221 and the third partial cluster 223. In this case, the first partial cluster 22
1 is read from the failure storage device cluster 232 of the second partial cluster 222 and the third partial cluster 22 is read.
3 can be read from the failure storage device cluster 236 of the fourth partial cluster 224. Similarly the second
If a failure occurs simultaneously in the partial cluster 222 and the fourth partial cluster 224, the failure storage device cluster 238 of the first and third partial clusters 221 and 223,
The fault can be recovered using 234.

Third Modification

FIG. 7 shows the outline of the configuration of the data storage system in the third modification. In the third modification, the contents of the normal storage device of each unit server are distributed and copied to the failure storage devices of any other number of unit servers without performing grouping like a partial cluster. ing. In this system, first to ninth disk controllers 241 to 249 are connected to a network 251. Each disk controller 241-2
49 is connected to normal storage devices 271 to 279 and fault storage devices 281 to 289 via buses 261 to 269.

A failure storage device for storing a copy of the storage contents of the normal storage device of each unit server is set for each normal storage device of each unit server. For example, failure storage devices 282 to 284 are assigned as copy destinations of data stored in the normal storage device 271. If the setting is made different for each normal storage device, the management becomes complicated. Therefore, here, three failure storage devices continuous from the unit server following the unit server to which the normal storage device belongs are set as copy destinations. Of course, the copy destinations do not need to be continuous, and may be different.

In the above-described embodiment and the first to third modifications, the normal storage device and the fault storage device are physically separate devices, but the same storage device is used in the same storage device. And a storage area for a failure storage device. In addition, a workstation or a host computer can be used as the disk control device.

[0072]

As described above, according to the first aspect of the present invention, the copy of the storage contents of the first data storage means of each unit server is copied to the second data storage means of another plurality of unit servers.
Since the data is divided and stored , even if a failure occurs in one unit server, the load on another unit server does not increase significantly. Not only when a failure occurs in the first data storage unit of the unit server, but also when a failure occurs in another circuit part of the unit server such as the first writing unit, the failure of the unit server in which the failure has occurred. The same data stored in the first data storage means can be read from another unit server. Sending Moreover in the first aspect of the present invention, dividing delivery means divides the same data as that written in the first data storage means to a plurality of destination a predetermined non-self substantially equal respectively I decided to do it. For this reason, there is an effect that the amount of data division is made uniform and the load at the time of occurrence of a failure is minimized.

According to the second aspect of the present invention, the data is divided and stored in a plurality of other unit servers for each file, so that data division and reading of the divided data can be easily managed. Can be.

According to the third aspect of the present invention, the first
Since the same data transfer destination is assigned to each block obtained by dividing the storage area of the data storage device into a plurality,
Splitting and data, the reading of the divided data can be easily managed.

According to the fourth aspect of the present invention, the same data stored in the first data storage means of each unit server is stored in a plurality of other unit servers in the group to which the unit server belongs. Is distributed and stored in the second data storage means. Thus, a failure of one unit server can be recovered in each group, and failures of two or more unit servers can be dealt with in the entire data storage system.

Further, according to the fifth aspect of the present invention, the same data stored in the first data storage means of each unit server is stored in the second data of the group server other than the unit server. Is stored in the data storage means. For example, if the server is divided into four or more groups, it is possible to cope with a failure of a unit server in two or more groups.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram showing an outline of a configuration of a data storage system according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an outline of a circuit configuration of the disk control device shown in FIG.

3 is an explanatory diagram schematically showing an example of storage contents of each storage device of the data storage system shown in FIG.

FIG. 4 is a system configuration diagram showing an outline of a configuration of a data storage system according to a first modified example of the present invention.

FIG. 5 is a system configuration diagram showing an outline of a configuration of a data storage system according to a second modified example of the present invention.

FIG. 6 is a system configuration diagram showing an outline of a data storage system composed of four partial clusters.

FIG. 7 is a system configuration diagram showing an outline of a configuration of a data storage system according to a third modification of the present invention.

FIG. 8 is a system configuration diagram showing an outline of a configuration of a data storage system using a mirror system which has been conventionally used.

FIG. 9 is a system configuration diagram showing an outline of a configuration of a data storage system capable of coping with a failure of a conventionally used disk control device.

FIG. 10 is a system configuration diagram showing an outline of a simple data storage system having a bus configuration capable of coping with both a failure of a disk device and a failure of a disk control device conventionally used.

[Explanation of symbols]

11, 87, 154, 251 Network 12-15, 41, 81-86, 141-153, 24
1 to 249 Disk control device 16, 21, 24, 27, 91 to 96, 261-269
Buses 17, 22, 25, 28, 48, 101-106, 18
1 to 193, 271 to 279 Normal storage device 18, 23, 26, 29, 49, 111 to 116, 20
1-213, 281-289 Failure storage device 31-34 Unit server 42 CPU 43 CPU bus 44 ROM 45 RAM 46 Network controller 47 SCSI controller 51-59, 61-69, 71-76 Partition area 77 Host computer 121, 122, 221-224 Partial cluster 131, 133, 231, 233, 235, 235, 237 Normal storage cluster 132, 134, 232, 234, 236, 238 Fault storage cluster

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G06F 12/00 G06F 12/16 G06F 3/06 G06F 11/16-11/20 G06F 15/16

Claims (5)

(57) [Claims] 1. A plurality of unit servers connected via a predetermined network include first and second units for storing data.
Data storage means, and a disk control device for controlling these data storage means, wherein the disk control device transmits and receives data to be input / output to / from these data storage means to / from the network; and When data to be stored in the first data storage means is received through the network by the communication means, the first data is written into the first data storage means.
And when the first writing means writes data to the first data storage means, sends the same data to a predetermined unit server other than itself through the network. Storage means
Almost a division transmitting means for transmitting divided equally, the second data storage means which upon receiving the data sent from the dividing transmitting unit of another unit server via said network in response to the written address Second to write to
Writing means, a failure detecting means for detecting the presence or absence of a failure in the unit server, and data stored in the first data storage means of the unit server which is detected to have a failure by the failure detecting means. When reading
Instead of this, the same data is used to divide the above-mentioned approximately equally.
The data after being divided and transmitted is stored in the first data storage means.
Based on the written address and the corresponding address,
A data storage system comprising: a failure reading unit that reads data from each of the transfer destination second data storage units. 2. The data is written to the first data storage unit in units of files, and the divided sending unit is configured to store the plurality of predetermined data in units of files written to the first data storage unit. 2. The data storage system according to claim 1, wherein the storage contents of the first data storage unit are distributed to the transfer destination. 3. The storage area of the first data storage unit is divided into a plurality of blocks, and the divided transmission unit stores the first data in the plurality of predetermined destinations in units of the blocks. 2. The data storage system according to claim 1, wherein the storage contents of the means are divided . 4. The plurality of unit servers are divided into two or more groups each composed of a plurality of unit servers, and the predetermined plurality of transfer destinations are other unit servers in each group. 2. The data storage system according to claim 1, wherein: 5. The plurality of unit servers are each divided into two or more groups each composed of a plurality of unit servers, and the predetermined plurality of transfer destinations are each a unit server of another group. 2. The data storage system according to claim 1, wherein: