JPH0934770A - Distributed data base system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recover data even when any fault is generated at a data server device by reproducing a file by merging a recovered preserved block with the respective preserved blocks of the other respective data server devices. SOLUTION: When any one data server device S4 disables reading among respective data server devices S1-S5 to which the respective preserved blocks are transferred, a client device 11 recovers the unreadable preserved block by performing parity arithmetic to the respective preserved blocks and redundant blocks read out of the other respective data server devices S1-S3 and S5. The file is reproduced by merging that recovered preserved block with the respective preserved blocks of the other respective data server devices S1-S3 and S5. Thus, even when any fault is generated at the data server device, the data can be recovered and as a result, responsiveness to a transfer request and reliability can be improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed database system having a mass storage device, and more particularly to a database system capable of improving responsiveness and reliability to a transfer request.

[0002]

2. Description of the Related Art For example, a centralized database system having one server device having a large-capacity storage device is widely used in routine operations such as a bank accounting system and a seat reservation system.

In this centralized database system, the system is operated in a form of centrally managing data, and data can be individually read / written from a plurality of client devices by networking.

However, when a certain number or more of transfer requests are received from each client device, the response to these transfer requests becomes insufficient, and there is a problem that responsiveness deteriorates.

In addition, when the server device has some kind of failure and stops functioning, the entire database system will stop functioning, which causes a problem of lowering reliability. On the other hand, from the viewpoint of solving this kind of problem, a distributed database system including a plurality of server devices is expected.

Since the distributed database system manages data by distributing it to each server device, it can sufficiently cope with a considerable number of transfer requests. Further, even when a certain server device has stopped functioning, the database system as a whole can continue processing.

[0007]

However, in the distributed database system as described above, when a certain server device stops functioning, the data in the server device is used although the entire system can continue processing. Since it becomes impossible, there is a problem that the responsiveness to the transfer request and the reliability are lowered.

The present invention has been made in consideration of the above circumstances, and provides a distributed database system capable of restoring data even when a failure occurs in a data server device and thus improving the responsiveness and reliability to a transfer request. The purpose is to provide.

[0009]

According to a first aspect of the present invention, in a distributed database system in which a plurality of client devices and a plurality of data server devices are connected to each other through a network, each client device is Is a storage block generation unit that divides a file to be stored to generate a plurality of storage blocks; and a redundant block generation unit that performs a parity operation on each storage block generated by the storage block generation unit to generate a redundant block. A block transfer means for distributing and transferring each storage block generated by the storage block generation means and a redundant block generated by the redundancy block generation means to each of the data server devices, and each storage by the block transfer means. One of each data server device to which the block is transferred When one of them is not readable, each storage block and redundant block read from each of the other data server devices are subjected to the parity operation to restore the unreadable storage block, and the block restoration means restores the unreadable storage block. And a file reproduction means for reproducing the file by integrating the storage block with each storage block of each of the other data server devices.

According to a second aspect of the present invention, in the distributed database system according to the first aspect, the file is connected to the network and the file name, each storage block and redundant block are distributed and stored for each file. Storage server information storage means for storing each data server device name and each storage start block address in each data server device; and a plurality of block storage areas provided in each data server device for each file. The file arrangement information storage means in which the block addresses indicating the addresses of the corresponding storage blocks and redundant blocks are stored in a chained manner in the block storage areas, and the read request for a certain file is issued by one of the client devices. Is input to the storage server information in response to this read request. A server search means for searching each data server device and each save start block address corresponding to the file by referring to the storage means, and a save start block address for each data server device based on the search result of the server search means. Read start means for designating and starting the reading of a file, and a read start means provided in each of the data server devices. When a save start block address is designated by the read start means, the save start block is changed from the file arrangement information storage means While searching for each chained block address, the storage block or redundant block designated by the storage start block address or each block address is read, and the storage block or redundant block is transferred to the client device to which the read request is input. With block reading means for It is a database system.

Further, the invention according to claim 3 is the distributed database system according to claim 1 or 2, wherein the parity operation is an exclusive OR operation.

[0012]

Therefore, according to the invention corresponding to claim 1, the saving block generating means divides the file to be saved to generate a plurality of saving blocks by taking the above means, and the redundant block generating means A parity operation is performed on each storage block generated by the storage block generation unit to generate a redundant block, and the block transfer unit generates each storage block generated by the storage block generation unit and the redundant block generated by the redundant block generation unit. The data is distributed and transferred to each data server device, and the block restoration means
When any one of the data server devices to which each storage block has been transferred by the block transfer means is unreadable, the storage block and the redundant block read from each of the other data server devices are subjected to a parity operation to be unreadable. Since the save block is restored and the file reproducing means reproduces the file by integrating the save block restored by the block restoring means and each save block of each of the other data server devices, a failure occurs in the data server device. The data can be restored and the responsiveness and reliability to the transfer request can be improved.

Further, the invention according to claim 2 is connected to a network, and for each file, a file name, each storage block, and each data server device name in which redundant blocks are distributed and stored, and each of these data server devices. A storage server information storage unit that stores a storage start block address, and a block that is provided in each data server device and that has a plurality of block storage areas and that indicates the address of each corresponding storage block and redundant block for each file File arrangement information storage means for storing addresses in a distributed manner in each block storage area is provided, and when the server search means inputs a read request for a certain file to any of the client devices, the read request The storage server information storage means is referenced in accordance with the The device and each save start block address are searched, and the read start means specifies the save start block address to each data server device based on the search result of the server search means to start reading the file. , Provided in each data server device, when the storage start block address is designated by the read start means,
Each block address linked to the save start block is searched from the file arrangement information storage means, and the save block or redundant block designated by the save start block address or each block address is read out, and each save block or redundant block is searched. Since the data is transferred to the client device to which the read request is input, in addition to the operation corresponding to claim 1, it can be easily realized with a simple configuration.

Further, in the invention according to claim 3, since the parity operation corresponding to claim 1 or claim 2 is an exclusive OR operation, it can be operated easily and surely.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of a distributed database system according to an embodiment of the present invention. This distributed database system includes four client devices 11 to 1
4, one management server device 21 and five data server devices S1 to S5 are connected to each other via a network 31. The client devices 11 to 14, the management server device 21, and the data server devices S1 to S5 each have a hard disk (hereinafter referred to as HD).

Here, each client device 11-14
In accordance with the procedure shown in FIG. 2, for example, as shown in FIGS. 3 and 4, the transfer target file to be stored is divided into 15 storage blocks, and an exclusive logic as a parity operation is performed for every 5 storage blocks. It has a function of calculating the sum and generating one redundant block based on the calculation result, and a function of sequentially transferring these storage blocks and redundant blocks individually to the data server devices S1 to S5.

The parity operation is performed by "1" of each bit of addresses corresponding to each other in each storage block and redundant block.
Is an even parity operation in which the total number of the storage blocks is an even number. Specifically, as shown in the general truth table of FIG. 5, “1” is set for each bit of the addresses corresponding to each storage block. When the number of bits of is an odd number, the exclusive OR that sets the operation result to "1" is used. The parity calculation may be a calculation for odd parity, if desired.

Further, each of the client devices 11-14 is
According to the procedure shown in FIG. 6, the function of reading out each storage block and each redundant block from each of the data server devices S1 to S5 via the management server device 21 and the normal operation of integrating each read storage block in the original file And a certain data server device Si has a failure, the exclusive OR as a parity operation is calculated from each read storage block and each redundant block in case of a failure of the data server apparatus Si, and based on the result of the operation, it is lost due to the failure. It has a function of restoring the data of the save block and a function of integrating the read save blocks and the restored save block to reproduce the original file.

The parity operation for restoration is shown in FIG.
As shown in, the operation is an exclusive OR operation similar to the creation of the redundant block, but when the redundant block is created by the operation for the odd parity, the operation for the odd parity is used.

The management server device 21 has registration storage server information for managing files and management information for managing the system in the HD (storage server information storage means), and each client device 11-14 and each data. It is accessible from the server devices S1 to S5. The registered storage server information indicates, for each file, the address of each storage block that is a part of the file and the redundant block that is the calculation result from each storage block. Specifically, as shown in FIG. Further, each file is provided with a set of a file name, a saved data server device name (saved server), and an HD save start block address in the saved data server device. On the other hand, the management information includes items such as a parity generation unit indicating that one redundant block is generated for every five storage blocks, and the number of data server devices.

Each of the data server devices S1 to S5 is provided with a file allocation table (hereinafter, referred to as FAT: file arrangement information storage means) indicating the address of a storage block for each file, and transfers files received from the management server device 21. Based on the instruction, the client device 1
H for each storage block transferred from any of 1 to 14
While writing to D, the written block address is FAT
And a function of sending each storage block of the corresponding file to the client device while referring to the FAT based on the file transfer instruction received from the management server device 21 to the network.

The FAT is individually provided in the data server devices S1 to S5 and indicates a chained block address in the HD of the device itself. For example, as shown in FIG. 9A, in the FAT of the data server device S1, there is "1" in the area of the block address "0", but the next block address of this "1" is "1". When the area of the block address “1” is read, it is “2”, when the area of the next block address “2” is read, it is “3”, and the next block address “3” is read. And "-1" and there is no block address "-1", the chain ends. That is, the desired file is shown in FIG.
As shown in (b), it is stored in the block addresses “0” to “3” in the HD of the data server device S1. The fact that the save start block address of the desired file is “0” is stored in the registered save server information of the management server device 21 described above.

Next, the operation of the distributed database system configured as described above will be described with reference to the flowcharts of FIGS. 2 and 6. (Data Writing) Now, it is assumed that the file F2 of a certain client device 11 is to be registered in the data server devices S1 to S5.

The client device 11 sends a transfer request for the file F2 to the management server device 2 in response to an input operation by the operator.
1 (ST1). The management server device 21 confirms the operating states of the respective transfer destination data server devices S1 to S5, and at the same time, the data server devices S1 to S5 in a transferable state.
To the client device 11 (ST2), and thereafter, the client device 11 is given a transfer permission designating the transfer destination data server devices S1 to S5.

When the client device 11 receives the transfer permission, the client device 11 reads the parity generation unit from the management server device 21 and divides the file F2 as transfer data into storage blocks having a predetermined storage capacity (ST3: storage block). As shown in FIG. 10, the exclusive OR is calculated from each storage block for each parity generation unit to generate a redundant block (ST4: redundant block generating unit).

After that, the client device 11 sequentially cyclically transfers each storage block and redundant block to each of the data server devices S1 to S5 to which the transfer permission is given in a block unit (ST5: block transfer means). Each time each of the data server devices S1 to S5 receives the storage block or the redundant block, as shown in FIG. 11A, the block address of the HD for storing the received block is written in the FAT, and at the same time, as shown in FIG. As shown in b), the received block is designated H by designating the block address.
Write to D (ST6).

When the client device 11 completes the transfer of each storage block and each redundant block of the file F2, the client server 11 obtains the file name of the file F2 and the data server device name to which the first storage block of the file F2 has been transferred. Transfer to device 21.

The management server device 21 retrieves the storage start block address of the HD in which the top storage block is stored from the corresponding data server devices S1 to S5, and stores the storage start block address, the file name and the data server described above. The registered storage server information consisting of a combination with the device name is written in its own HD (ST7).

This completes the writing and registration of the file F2. (Reading of data) Now, each of the data server devices S1 to S
In the case of the file 5, as shown in FIG.
It is assumed that 1 and F2 are saved.

Here, the client device 11 sends a transfer request for the file F1, for example, to the management server device 21 (ST11). The management server device 21 is the same as that shown in FIG.
File F from the registered storage server information of own HD shown in
The stored data server device name and storage start block address of 1 are searched (ST12: server search means),
While determining the transfer availability status including the operating status and failure status of each of these data server apparatuses S1 to S5 (ST13), each data server apparatus S in the transferable status.
Specify the save start block address in 1 to S5,
Although a transfer instruction is cyclically given (ST14: reading start means), when there is a failure, the name of the data server apparatus with the failure is notified to the client apparatus 11 (ST15) and the processing of step ST14 is executed.

On the other hand, each of the data server devices S1 to S5
When the transfer instruction is received from the management server device 21, the block address stored in the FAT shown in FIG. 11A is designated and the storage block of the file F1 is set to 1 from the HD.
While reading only the block (ST16: block reading means), this storage block is transferred to the client device 11 (ST17: block reading means), and the next storage block or redundant block of the file F1 is similarly referenced with reference to the FAT again. Is transferred by one block. The transfer of the next block is periodically executed up to the final storage block or redundant block of the file F1.

The client device 11 executes step ST1.
It is determined whether the status of each of the data server devices S1 to S5 is normal or faulty by the presence or absence of the notification of No.
8), based on the determination result, each of the data server devices S1 to S
The storage blocks transferred cyclically from 5 are integrated to reproduce the file F1.

At this time, the procedure for the client device 11 to reproduce the file F1 is as follows. (When each of the data server devices S1 to S5 is normal) (1a) As shown in FIG. 12, the storage blocks received from the first to the fifth are integrated, and the redundant block received from the sixth is ignored (ST19). ). (2a) (1a) is repeated by the number of sets of even parity. (3a) The integrated blocks integrated in the procedure of (1a) to (2a) are further integrated to reproduce the file F1 (S).
T20). (When the data server device S4 is damaged) (1b) As shown by the broken line arrow in FIG. 13, the exclusive OR operation is performed on each of the first to fourth received storage blocks and the fifth received redundant block. Applying (ST21: block restoring means), the damaged saved block is restored based on the calculation result (ST22: block restoring means). (2b) As shown by the solid line arrow in FIG. 13, the restored storage blocks are interrupted in the order of the first to fourth received storage blocks to integrate the respective storage blocks (ST23). The order of interruption is indicated by the damaged block order Dbl, and the damaged block order Dbl is obtained by the following equation (1). Dbl (n) = (Snu−Bnu) (n−1) + Dsa (1) However, n: the order of the set of even parity (n = 1 to 3 ... In the case of file F1), Snu: the stored blocks Number of servers (Snu = 5 ... S1, S2, S3, S4, S5), Bnu: Number of blocks in one set forming even parity (Bnu = 6 ... n-1, n-2, n-3, n-4) , N-5, n-
P), Dsa: rank of damaged server (Dsa = 4 ... Damage of data server device S4).

Next, the above conditions are substituted into the equation (1) to obtain the equation (2). Dbl (n) = (5-6) (n-1) + 4 = -n + 5 (2) From the equation (2), when n = 1, Dbl = 4th, when n = 2, Dbl = 3rd , N = 3, Dbl = second is obtained.

However, since n = 1 here, Dbl =
4th. In the next repetition of (3b), n = 2, n =
3 is used sequentially. (3b) Here, (1b) to (2b) are repeated. (4b) As shown in FIG. 14, the integrated blocks obtained by integrating the steps (1b) to (3b) are further integrated to reproduce the file F1 (ST20: file reproducing means).

As described above, according to this embodiment, the client device 11 divides the file F2 to be saved to generate a plurality of save blocks, and performs a parity operation on each save block to generate a redundant block. The file F2 is distributed and stored in each of the data server devices S1 to S5 by distributing and transferring each storage block and redundant block to each of the data server devices S1 to S5.

When the client device 11 cannot read any one of the data server devices S1 to S5 to which each storage block is transferred,
A parity operation is performed on each storage block and redundant block read from each of the other data server devices S1 to S3 and S5 to restore the unreadable storage block, and the restored storage block and each of the other data server devices S1 to S1. Since the file F1 is reproduced by integrating the storage blocks of S3 and S5, the data can be restored even if a failure occurs in the data server device, and thus the responsiveness to the transfer request and the reliability can be improved.

Further, according to this embodiment, the management server device 21 causes the file name, each save block, and each data server device name in which redundant blocks are distributed and stored, and each save start in each data server device, for each file. The registered storage server information including block addresses is stored, each data server device S1 to S5 has a plurality of block storage areas, and each file has a block address indicating the address of the corresponding storage block and redundant block. When the management server device 21 has a transfer request (read request) for a certain file input to any of the client devices, it has FATs that are distributed and stored in chain in each block storage area, and responds to this transfer request. The registered storage server information is referred to, and each data server device name and each storage start block corresponding to the file are referenced. When a save start block address is specified, each data server device S1 to S5 is caused to specify the save start block address to start reading the file. Each block address linked to the save start block is searched, the save block or redundant block designated by the save start block address or each block address is read, and a client to which a transfer request is input for each save block or redundant block is input. Since it is transferred to the device 11, it can be easily realized with a simple configuration.

Further, according to this embodiment, since the parity operation is the exclusive OR operation, it can be operated easily and surely. It should be noted that, in the above-described embodiment, each of the data server devices S1 to S5 is sequentially cyclic when data is read,
In addition, although the case where the storage block or the redundant block is periodically transferred to the client device 11 has been described, the present invention is not limited to this, and the client device 11 sequentially cyclically and sequentially transfers the data server devices S1 to S5. Even if the transfer instruction is sent to the device, the present invention can be implemented in the same manner and the reliability of the operation can be improved.

Further, in the above embodiment, the case where the redundant block is transferred at the time of data reading even in the normal time has been described, but the present invention is not limited to this, and, for example, the FAT indicates a storage block or a redundant block. Even if a flag is added and the redundant block transfer is omitted in normal operation by referring to this flag, the present invention can be implemented in the same manner and the transfer efficiency can be further improved.

Further, although the contents of the data are not described in the above embodiment, any data can be used in the present invention. For example, in the case of a printing database,
It is possible to use image data, character data, layout data for designating the arrangement of images and characters, and the like.
In addition, the present invention can be modified in various ways without departing from the scope of the invention.

[0042]

As described above, according to the first aspect of the present invention, the storage block generation means divides the file to be stored to generate a plurality of storage blocks, and the redundant block generation means uses the storage block generation means. A parity operation is performed on each generated storage block to generate a redundant block,
The block transfer means disperses and transfers the respective storage blocks generated by the storage block generation means and the redundant blocks generated by the redundant block generation means to the respective data server devices, and the block restoration means performs the respective storage blocks by the block transfer means. When any one of the transferred data server devices is not readable, each storage block and redundant block read from each other data server device is subjected to a parity operation to restore the unreadable storage block, Since the reproducing means reproduces the file by integrating the storage block restored by the block restoring means and the respective storage blocks of the other data server devices, the data can be restored even if the data server device fails. , A distributed database system that can improve responsiveness and reliability to transfer requests It can be provided.

According to the second aspect of the present invention, each data server device connected to the network and having the file name, each storage block, and the redundant block distributedly stored for each file, and each of these data server devices. A storage server information storage unit that stores a storage start block address, and a block that is provided in each data server device and that has a plurality of block storage areas and that indicates the address of each corresponding storage block and redundant block for each file File arrangement information storage means for storing addresses in a distributed manner in each block storage area is provided, and when the server search means inputs a read request for a certain file to any one of the client devices, the read request is added to the read request. Correspondingly, referring to the storage server information storage means, each data server device corresponding to the file And each save start block address are searched, and the read start means specifies the save start block address to each data server device based on the search result of the server search means to start reading the file. When the save start block address is designated by the read start means provided in each data server device, each file address information storage means is searched for each block address linked to the save start block, and the save start block address or each save start block address is searched. Since the storage block or redundant block designated by the block address is read and these storage blocks or redundant blocks are transferred to the client device to which a read request is input, in addition to the effect of claim 1, a simple configuration facilitates A distributed database system that can be realized can be provided.

Further, according to the invention of claim 3, since the parity operation of claim 1 or claim 2 is an exclusive OR operation, it is possible to provide a distributed database system which can operate easily and surely.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a distributed database system according to an embodiment of the present invention,

FIG. 2 is a flowchart for explaining a write operation in the same embodiment,

FIG. 3 is a schematic diagram for explaining file division in the same embodiment;

FIG. 4 is a schematic diagram for explaining generation of redundant blocks in the same embodiment;

FIG. 5 is a diagram showing a general truth table of exclusive OR.

FIG. 6 is a flowchart for explaining a read operation in the embodiment.

FIG. 7 is a diagram showing a truth table of general exclusive OR,

FIG. 8 is a schematic diagram showing registration storage information in the embodiment.

FIG. 9 is a schematic diagram showing a block storage method in the embodiment.

FIG. 10 is a schematic diagram showing generation of redundant blocks in the same embodiment;

FIG. 11 is a schematic diagram showing the configuration of FAT and HD in the same embodiment;

FIG. 12 is a schematic diagram showing the integration of storage blocks under normal conditions in the same embodiment;

FIG. 13 is a schematic diagram showing integration of storage blocks at the time of failure in the same embodiment;

FIG. 14 is a schematic diagram showing reproduction of a file in the example.

[Explanation of symbols]

11 to 14 ... Client device, 21 ... Management server device, 31 ... Network, S1 to S5 ... Data server device, HD ... Hard disk.

Claims (3)

[Claims] 1. A distributed database system in which a plurality of client devices and a plurality of data server devices are connected to each other via a network, wherein each client device divides a file to be saved into a plurality of save blocks. A storage block generation unit that generates a redundant block by performing a parity operation on each storage block generated by the storage block generation unit; and a storage block generated by the storage block generation unit. And one of a block transfer unit for distributing and transferring the redundant blocks generated by the redundant block generation unit to each of the data server devices, and each of the data server devices to which each of the storage blocks has been transferred by the block transfer unit. When data cannot be read, other data server devices Block restoring means for performing the parity operation on each of the storage blocks and the redundant blocks read from the storage block to restore the unreadable storage block; and the storage block restored by the block restoring means and each of the other data server devices. A distributed database system comprising: a file reproducing unit that reproduces the file by integrating each storage block. 2. The distributed database system according to claim 1, wherein a file name, each storage block, and each data server device name in which distributed blocks of redundant blocks are connected and connected to the network and each of these files are stored. Storage server information storage means for storing each storage start block address in the data server device, and a plurality of block storage areas provided in each of the data server devices, and an address of each corresponding storage block for each file And file arrangement information storage means in which each block address indicating the address of the redundant block is distributed and stored in chain in each of the block storage areas, and when a read request for a certain file is input to any of the client devices, In response to this read request, the storage server information storage means is referred to, A server search unit that searches each data server device name and each save start block address corresponding to the file, and specifies a save start block address for each data server device based on the search result of the server search unit. Read start means for starting read, and each block address provided in each of the data server devices and linked to the save start block from the file arrangement information storage means when a save start block address is designated by the read start means Block reading means for searching the storage start block address or the storage block or redundant block designated by each block address, and transferring the storage block or redundant block to the client device to which the read request is input. Distributed type characterized by having Over database system. 3. The distributed database system according to claim 1 or 2, wherein the parity operation is an exclusive OR operation.