JP3468531B2 - Database management device and record storage method thereof - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention is an online database.
A method for storing a data base record in a data block of a data system, particularly when a new data base record is requested to be added, many data base records are stored.
The present invention relates to a database-based record storage system suitable for storing data.

[0002]

2. Description of the Related Art Conventionally, a hash method, an index file method and the like have been known as a method of storing a database record.

When a new record is added to a storage device by the hash method, if there is no free area in the target block in which the new record is to be stored, the new block is allocated and connected by a pointer. ing.

Regarding the record storing method by the hash method, the reference "Principle of Database System" (J.
EFFREY D. ULLMAN, Toshiyasu Kunii, Nobuo Ohbo, 1985 Japan Computer Association), p. 48-56.

In the index file system, the index of the record key information is created by the balance tree system. Further, a group of a plurality of blocks is used as a group, and an empty block for storing data is provided in the group when a file is created. When a new record is added and there is no free area in the block to which the new record is to be added, the existing block is divided into about two, and the free block of the data in the group is acquired and stored. To do. At this time, the lowest index is updated, and the block storing the target record is directly pointed to from the lowest index.

Further, as a method for adding a record of an index block of an index file, for example, Japanese Patent Laid-Open No. 2-67648.
Although there is a publication, it is a method for efficiently storing the index block of a record forming a tree structure, and no reference is made to the data block.

[0007]

In the above-mentioned hash method of the prior art, new records are frequently added to the same hash value, and there is no empty area in the target block to which the new record should be added. In this case, it is necessary to allocate a new block and connect it with a pointer. Therefore, the target record
I need to read multiple blocks when searching for
There is a problem that the reading speed becomes slow.

Further, the above-mentioned conventional index file system has a problem that a new record cannot be added when there are no free blocks in the group and there is no free block to be added (theoretically). An existing group can be divided into two, and an empty group can be allocated and stored, but the number of I / Os and the amount of history information acquired at that time increase, which causes a decrease in online performance. Group division is not done.).

The object of the present invention is to suppress the generation of synonyms in the case of the hash method when a new record is added when there is no free area in the data block to which the new record should be added. In the case of the index file method, however, it is an object of the present invention to provide a data base record storage method capable of suppressing the occurrence of block division and additionally storing a large number of records.

[0010]

In order to achieve the above-mentioned object, a storage device comprising a main area and an overflow area, each of which is divided into a plurality of blocks, is provided with a recording information portion and a segment portion. -When adding a new record in the database system that stores the record, it is sufficient to store the management information part of the new record in the block of the main area to which this new record should be added. If there is a free area, only the management information part is stored in the main area block and a new record is created.
The segment part of the code is stored in the overflow area.
On the other hand, if the block of the main area to which the new record is to be added does not have enough free area to store the management information portion of the new record, among the records stored in this block, A record having a segment part whose area including the segment part and the insufficient empty area is sufficient to store the management information part of the new record is divided into the management information part and the segment part. After the segment portion is stored in the overflow area, the management information portion of the new record is stored in the main area and the segment portion is stored in the overflow area.

[0011]

The online database system performs processing in units called transactions.

The above transaction calls the database manager when accessing the database.

Before accessing the physical block of the database, the database manager performs lock control for one block to be accessed or a group unit of the block, and the block or group is controlled by other units. Access the target block after performing exclusive control to prevent simultaneous access from other transactions. Further, the database manager is a means for storing, in the overflow area, a record which cannot be stored in the main area among the main area and the overflow area as the storage area for the data base record. have. Further, the index file system has means for creating index information by the balanced tree system or the like and storing it in the index area in order to search the records stored in the main area with as few I / Os as possible.

The data managed by the database manager
In the tabular record, a plurality of records are stored in a data set (for example, a logically divided area in a magnetic disk device), and a plurality of records are stored in one physical block in the main area. Are stored. The above record is
For example, in the case of a ledger database of a bank, it is composed of a plurality of segments such as a segment for storing balances, a segment for storing end book information, a segment for storing address information, etc. Each segment is connected by a pointer. A tree structure is formed.

The record includes key information for identifying the record and management information such as overflow management information that points to a segment that cannot be stored in the main area and is stored in the overflow area. Are stored in the data block of the main area. When a record addition request is made in a state where there is no free area for storing only management information on the data block of the main area, among the records stored in the same block, that The area that combines the segment part and the above-mentioned insufficient empty area is a new record
Check if there is a record with enough segment portions to store the management information portion of the record. If the corresponding record exists, the record is divided into a management information part and a segment part, this segment part is stored in the overflow area, and then the management information part is stored in the main area, and then the main area. Create an empty area in the corresponding data block of. After that, the record of only the management information including the key information of the record requested to be added is stored in the data block of the main area, and the remaining segment information is opened.
Store in the block in the buffer area.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing a computer system to which the present invention is applied. FIG. 1 shows an example in which an online database management system 20 operating on the data processor 1 accesses the database file devices 61-63.

A communication control device 2 is connected to the data processing device 1, and a terminal device 4 is connected to the communication control device 2 via a modem 3. A plurality of data base file devices 61 to 63 and a journal file device 5 are connected to the data processing device 1. The data processing device 1 receives a message from the terminal device 4 via the communication control device 2 and the modem 3 and sends a message to the terminal device 4.

Data database connected to the data processing apparatus 1
The file processing devices 61 to 63 are the business processing programs 30.
Online database management system 20
Access through the operating system 10.

The online database management system 20 also performs processing in units called transactions.
As a transaction history, a file journal as an update history of the database file devices 61 to 63, a user arbitrary journal as a business history, etc. are journaled.
It is stored in the null file device 5.

FIG. 2 is a diagram showing the storage state of the database records in the index file system.
(A) is a diagram showing the contents of a data block 100 in the main area of the data base file devices 61 to 63 before the data base record C is added as a new record. FIG. 2 (b) shows a data base file device 61 after adding a data base record C as a new record.
~ 63 main area data block 100 ',
It is the figure which showed the contents of the blocks 200 and 300 of a buffer area.

In FIG. 2A, the data block 1 in the main area of the data base file units 61-63.
00 stores a database-based record A 110, a database-based record B 120, and block control information 150, and has an empty area 140. Here, the record A 110 and the record B 120 have record management information and segments A and B, respectively. This record management information comprises the size of the record, key information for identifying the record, and overflow management information for storing pointer information of the overflow area. The segments A and B respectively include management information of the segments A and B (identifier of the segment, status information of whether or not the segment has a child segment), and user information. The control information 150 stores the number of records in the data block 100 and the size of the empty area 140.

The data block 10 shown in FIG.
In FIG. 2 (b) in which the data base record C is added to the data base 0, the data base 100 'in the main area of the data base file units 61 to 63 is connected to the data base record. A111, data base record B120, data base record C130, and block control information 15
0 is stored, and there is an empty area 141 in addition. Here, the record B 120 and the control information 150 have the same data structure as in FIG.
The record C130 has only record management information consisting of record size, key information for identifying the record, and overflow management information for storing pointer information of the overflow area.

Record A11 of the data block 100 '
Reference numeral 1 is a record in which the record A110 is changed and stored when the record C is added. The segments A and B, which depend on the record A110 before adding a new record C, are stored in the overflow area block 200 as the record A112 together with the management information including the record size and the key information. Is stored. Similarly, the segments A and B of the additional record C are also stored in the block 300 of the overflow area as the record C131 together with the management information including the record size and the key information.

FIG. 3 is a flow chart for explaining the process of adding a database record of the online database management system 20. Below, based on FIGS. 1 and 2, the data base file devices 61 to 63 will be described in detail.
The case of adding the record C as a new record to the data block 100 in the main area will be described.

When a message is input from the terminal device 4, the message is passed to the data processing device 1 via the modem 3 and the communication control device 2. This input message is the operation
The data is transferred to the business processing program 30 through the operating system 10 and the online database management system 20.

The business processing program 30 analyzes the message and, if it is necessary to add a database record C, the online database management program 20 sends a database record to the database. Make an additional store request for C.

Online database management system 20
When receiving a request to add the database base record C, the index block which is the block storing the position information of the data block in which the target record should be stored is read, and the additional record is read. The data block 100 in the main area in which the data C is to be stored is read (step 301).

The key of the record C which receives the additional request
Whether or not a record having the same key information as the information exists in the data block 100, that is, the data block 1
It is checked whether or not there is the same key information as the key information of record C in 00 (step 302), and if there is the same key information as the key information of record C, an error return is performed. .

If the same key information as the key information of the record C is not present in the data block 100, the empty area 140 in the data block 100 and the key information of the record C are overlapped. It is checked whether or not there is a size for storing a record composed of buffer management information (step 303).

If the vacant area 140 is not large enough to store a record consisting of the key information of the record C and the overflow management information, the record is placed in the data block 100.
Record consisting of key information of C and overflow management information
It is checked whether or not there is a record having a size longer than twice the record size (step 304). If there is a long-sized record (record A110 in this embodiment is applicable), a block 200 of the overflow area is newly allocated, and segment information (segment) of the long-sized record A110 is allocated. A and B) are edited into the block 200 of the overflow area (step 305). Then, the long-sized record A110 in the main area is edited into a record consisting of key information and overflow management information, and then overwritten.
The pointer information of the buffer area is stored in the overflow management information portion (step 306).

Thereafter, in order to store the record C for which an addition request has been made, a block 3 of a new overflow area is newly added.
00 is assigned and the information of the segment portion of the record C for which the addition request is made is edited (step 309). Then, in the data block 100 'of the main area, a record composed of the key information of the record C and the overflow management information which is requested to be added is edited, and the overflow management information portion is overwritten.
The pointer information of the block 300 in the buffer area is stored (step 310).

Then, the journal information is acquired in the journal file device 5 as the update history of the updated block 100 'and the newly allocated blocks 200 and 300 (step 311), and the overflow area is updated. The added blocks 200 and 300 are written to the database file devices 61 to 63 (step 312), and the updated / added block 100 'of the main area is written to the database file devices 61 to 63 (step 313). ), The processing ends.

In step 303, the empty area 140 of the data block 100 in the main area is recorded in the record C.
If there is only a size for storing a record consisting of the key information and the overflow management information, the data block 100 'in the main area has the key information of the record C and the overflow management information. Is stored, and the processing shown in steps 309 to 313 is performed to store the segment information in the overflow area 300.

In step 304, if the data block 100 in the main area does not have a record of a long size, an empty data block in the main area group, which is a set of a certain amount of data blocks. It is checked whether there is any (step 307). If there is no empty data block, error return is performed, and if there is an empty data block,
An empty data block in the main area is secured and the data block in the main area is divided (step 3).
08).

The data area division processing of the main area performed here includes a plurality of records stored in the data block 100 and a record C for which an addition request has been made, and several records each. The first group is stored in the data block 100, and the remaining one group is stored in the reserved empty data block.

After the division processing of the data blocks in the main area is performed, the procedure is the same as that when there is an empty area in the data blocks in the main area (steps 309 to 31).
3) Store the record C for which addition has been requested.

In the present embodiment, the record C, which has been requested to be added, is added after the record B 120 in the ascending order of the key information, but the present invention is not limited to this and there is a request for addition. The record C can be added between the record A110 and the record B120. In this case, if there is sufficient free area in step 303, the record C to be stored between the record A 110 and the record B 120.
It is possible to move the record B120 by an amount corresponding to the record size, to create an empty area between the record A110 and the record B120, and then execute the processing from step 309. If there is not enough free area in step 303, it is possible to carry out the processing from step 304 to 308 and then move the record B 120 in the same manner as described above and then execute the processing from step 309. .

In this embodiment, step 305,
At 306, the record C key is placed in the data block 100.
It is checked whether or not there is a record having a length that is twice or more the record size consisting of information and overflow management information, and the corresponding record (record A110 in this embodiment) is record managed. Although the information is divided into the information part and the segment part, the present invention is not limited to this. Of the records stored in the data block, the segment part is stored in the overflow area. The area formed by the main area block and the empty area which is not enough to store the record management information portion of the existing record C in the data block are combined into a new record. Other records (for example, record B120) may be divided and edited as long as the record has a segment part sufficient to store the management information part of the above. In this case, as a method of determining which record is to be divided and edited, for example, in step 304, the records in the data block 100 are sequentially viewed from the front and the first record satisfying the above conditions is divided. There are ways to edit.

According to the present embodiment, for example, the block size is 8,192 bytes, and four 2,000-byte database records are stored, and the key information and the overflow of the main area are stored. -Record including management information When the record size of the management information portion is 50 bytes, about 160 records can be added to one block.

Further, according to this embodiment, since only the record management information portion can be stored in the data block of the main area, the empty area can be enlarged as much as possible.
The hash method is effective in reducing the occurrence of synonyms.

Further, according to this embodiment, there is an effect that the number of times of input for searching the target block can be reduced. For example, if the block size is 8,192 bytes,
If there are 160 records of 000-byte data base records, and 160 records have the same hash value,
In the conventional hash method, the average number of inputs is 160 ÷ (81
92/2000) ÷ 2≈20 times, but the total number of inputs according to the present invention is 2 in the main area and the overflow area.
It only needs to be entered once.

[0043]

According to the present invention, the record management information portion of the data base record is stored in the data block of the main area, and the other portion is stored in the data block of the overflow area. By storing the data in the main area, the empty area of the data block in the main area can be made as large as possible. Therefore, when a request for adding a new data base record is made, many data base records are required. Can be added efficiently.

[Brief description of drawings]

FIG. 1 is a block diagram of a computer system to which the present invention is applied.

FIG. 2 is a diagram showing a storage state of a data base record in a data block in the case of the index file system.

FIG. 3 is a flow chart showing an embodiment of the processing of the data base record addition processing system of the present invention.

[Explanation of symbols]

1 ... Data processing device, 2 ... communication control device, 3 ... modem, 4 ... terminal device, 5 ... Journal file device, 10 ... Operating system, 20 ... Online database management system, 30 ... Business processing program, 61-63 ... a data base file device,

Continuation of front page (56) Reference JP-A-2-105237 (JP, A) M. et al. M. ASTRAHAN, System R: Relational Approach to Data base Management, AC M Transactions in Database Systems, USA, ACM, June 1976, Vol. 1, No. 2, p. 97-137 (especially 117-122) Masami Yamatani, Introduction to File Organization, Nihon, Ohmsha, July 25, 1980, 1st edition, p. 18 −− 23, 102 −− 113 (58) Fields investigated (Int.Cl. ⁷ , DB name) G06F 12/00 G06F 17/30

Claims (2)

(57) [Claims] 1. A record storage method in a database device for storing a plurality of records having management information and a segment in a storage device having a main area and an overflow area, wherein a record including the management information part and the segment part is provided.
The segment part of the
The first part of the split record containing at least the report part
The minute record is stored in the main entry, and the management information is stored.
The second split, which is the other split record that does not contain
Store a record in the overflow area and store the first partial record in the main area
In this case, the first partial record is stored in the main area.
If there is not enough free space to store, store it in the main area
The segment part of the first partial record
Segment part in one or more first partial records containing
The total capacity of minutes is larger than the first partial record stored above.
If not, the first partial record is repartitioned and
The segment part included in the record
First partial record to be moved to the low area and stored
Is stored in the main area.
Storage method. 2. A main area and an overflow area are provided.
A storage device that stores management information and segments
In a database device for storing a plurality of records, a recorder including the management information part and the segment part
The segment part of the
The first part of the split record containing at least the report part
The minute record is stored in the main entry, and the management information is stored.
The second split, which is the other split record that does not contain
A means for storing a record in an overflow area, and a means for storing the first partial record in the main area
In this case, the first partial record is stored in the main area.
If there is not enough free space to store, store it in the main area
The segment part of the first partial record
Segment part in one or more first partial records containing
The total capacity of minutes is larger than the first partial record stored above.
If not, the first partial record is repartitioned and
The segment part included in the record
First partial record to be moved to the low area and stored
Is provided in the main area.
Database management device that has been designated.