JP2000163294A - Method and device for managing database and machine- readable recording medium with program recorded thereon - Google Patents

Abstract

PROBLEM TO BE SOLVED: To unnecessitate a data buffer for delay update in a computer which performs the reference and update of a database on secondary storage on a data buffer provided in a main storage and also reflects an updated page in the database asynchronously with processing by an application program. SOLUTION: When a commit request is outputted from an application program with which a page constituting a database 1 is referred to and updated on a data buffer 2, a commit processing means 7 records information specifying the updated page as unupdated page information in a recording part 4. A delay updation processing means 8 reads only pages that are not currently used by any of application programs among pages shown by unupdated page information recorded in the part 4 from the buffer 2 and reflects them in the database 1 asynchronously with the processing of an application program.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to database management in a computer, and more particularly to database management using a delayed update method.

[0002]

2. Description of the Related Art Since a database is generally constructed on a secondary storage of a computer, a data area called a data buffer is secured on a main storage, and the contents of the data buffer are stored when a data access request is made from an application program. If the corresponding page is not found in the data buffer, the corresponding page is read from the database on the secondary storage into the data buffer.If there is, the number of secondary storage accesses is determined by using the page on the data buffer. Has been reduced. Since the data buffer is secured in volatile main storage, any data updated in the data buffer must be written back to the original page of the database.

[0005] The method of writing back the data updated on the data buffer by the application program to the database on the secondary storage is roughly classified into a synchronous type and an asynchronous type. The synchronous method is a method in which the update data in the data buffer is written back to the secondary storage when a commit request is made from the application program. In this method, it is ensured that the data has been updated on the secondary storage by establishing the stationary point, so that data integrity can be maintained. However, the application program needs to wait until the data is reflected in the secondary storage, which makes it difficult to execute the application program at high speed. In addition, since data is output to the secondary storage at one time, the secondary storage I / O is momentarily concentrated, and there is a problem that the throughput of database access is reduced due to the processing performance of hardware. A method devised to solve the problem of such a method is an asynchronous method. The asynchronous method updates a database on secondary storage asynchronously with a commit request from an application program, and is called a delayed update method. The present invention relates to database management employing this delayed update method.

[0004] An example of a database management system employing a delayed update system is described in Japanese Patent Application Laid-Open No. 3-288244. In this publication, data updated by an application program is stored in a delay update log on a secondary storage, and then the database is asynchronously updated based on the data stored in the delay update log. However, in this method, it is necessary to access the secondary storage once when storing the information in the delayed update log and twice when reflecting it in the database.

The following method has been considered as a delayed update method for further reducing the number of accesses to the secondary storage. As shown in FIG. 10, two sets of data buffers 101 and 102 of the same size are provided on the main memory 100,
One is referred from the application program 103,
It is used for updating, and the other is used for delayed updating of the database 104. Certain pages 105, 106, 1 in the database 104 from the application program 103
07, the buffer management means 10
8 reads the corresponding page into a free area of the data buffer 101. The application program 103 stores pages 105, 106, and 107 on the data buffer 101.
See and update. Now, pages 105, 106, 10
7 is a, b, and c, and the updated data is A, B, and C.

When the application program 103 issues a commit request, the commit processing means 109 copies the update data A, B, and C of the pages 105, 106, and 107 on the data buffer 101 to the data buffer 102. In addition, the update image 111 of the pages 105, 106, and 107 is output to the journal file 110 in preparation for recovery of the database in the event of a system failure.

After that, the delayed update processing means 112 reads the update data A, B, and C of the pages 105, 106, and 107 from the data buffer 102 and updates the database 104 asynchronously with the processing of the commit processing means 109.

[0008]

In the conventional delay update method shown in FIG. 10, data updated by an application program is copied to another data buffer on a main memory, and then stored in that data buffer asynchronously. Since the database on the secondary storage is updated based on the data that has been read, access to the secondary storage is not required during the commit process, and access to the secondary storage for reading is also unnecessary when reflecting on the database. JP-A-3-288
The number of accesses to the secondary storage is smaller than that of the method described in Japanese Patent No. 244.

However, the conventional delayed update method shown in FIG. 10 requires two sets of data buffers of the same size on the main memory, and has a disadvantage that the required capacity of the main memory increases.

An object of the present invention is to make it possible to perform delayed update processing using only one set of data buffers, and to reduce the required capacity of main storage.

[0011]

According to the present invention, in order to achieve the above-mentioned object, reference and update to a database on a secondary storage are performed on a data buffer provided in a main storage and updated on the data buffer. When a commit request is made from the application program, information specifying the page updated on the data buffer by the application program is used as unupdated page information in a computer that reflects the updated page in the database asynchronously with the processing by the application program. It is recorded in the non-updated page information recording unit, and is currently used by any application program among the pages indicated by the non-updated page information recorded in the non-updated page information recording unit asynchronously with the processing by the application program. Only pages that are not To reflect on the base.

When there is only one set of data buffers, when a commit request is issued from a certain application program, a page on the data buffer updated by the application program is copied to another data buffer as in the related art. Therefore, in the present invention, the commit processing means records which page has been subjected to the delay processing in the non-updated page information recording unit during the commit processing. The delayed update processing unit refers to the unupdated page information recording unit asynchronously with the application program,
Reflect the target page of the delay processing in the database. However, since the page on the data buffer that has been referenced and updated by the application program that has issued the commit request can be subsequently referenced and updated by another application program, the delayed update processing unit stores the page in the unupdated page information recording unit. Even if the recorded pages are subject to delay processing, pages currently used by any application program are not flushed to the database. That is, at that time, it is not targeted for delayed update processing. The reason is that the database cannot be updated with the update page at the time when the commit request has not been issued yet.

Further, according to the present invention, the information of the page to be subjected to the delay processing, which is recorded in the non-updated page information recording section without being delayedly updated, is periodically inspected, and if there is any such information, if there is, the information is checked. Checkpoint processing means for outputting checkpoint information including information to a journal file is provided. Since the commit processing means outputs an update image to the journal file at the time of a commit request as in the past, it is possible to recover the database in the event of a system failure from this update image and the checkpoint information. Nature is secured.

[0014]

FIG. 1 is a block diagram showing an example of an embodiment of a database management apparatus according to the present invention. In FIG. 1, 1 is a database, 2 is a data buffer, 3 is a journal file, 4 is an unupdated page information recording unit, 1
1 is a journal stack, 12 is a buffer management table, and 13 is an exclusion management table. The database 1 and the journal file 3 are provided on a magnetic disk device or a secondary storage equivalent thereto, and include a data buffer 2, an unupdated page information recording unit 4, and a journal stack 1.
1, buffer management table 12, exclusive management table 13
Is provided on the main memory. Reference numerals 5 to 10 denote functional units realized by the CPU and the control program of the computer, 5 denotes a database access processing unit, 6 denotes a buffer management unit, 7 denotes a commit processing unit, 8 denotes a delay update processing unit, and 9 denotes a check. Point processing means 10 is recovery processing means.

The database 1 is composed of a plurality of files, and each file is composed of a plurality of pages. Input / output to / from the data buffer 2 is performed in page units.

The data buffer 2 has a capacity to store a plurality of pages. Unlike the prior art shown in FIG.
Only one data buffer 2 is provided. The data buffer 2 is used by being logically divided into a number of buffers each storing one page.

The journal stack 11 stores an update journal including a pre-update image and a post-update image of a page referenced and updated by a transaction of an application program.

The non-updated page information recording section 4 is information for specifying a page that has been updated in the data buffer 2 but has not yet been reflected in the database 1 (unupdated page information).
Record FIG. 2 shows a configuration example of the non-updated page information storage unit 4. Each entry has a valid bit 41 indicating whether the entry is valid or not.
And an identifier 42 of a page not yet reflected in the database 1 (for uniquely identifying the page on the database 1).

The journal file 3 records information on the start and end of a transaction, images before and after an update, unupdated page information, and the like.

The buffer management table 12 stores management information for each buffer in the data buffer 2. An example of the configuration is shown in FIG. Each buffer is made up of an entry corresponding to each buffer, and each entry has a buffer number (for example, a head address of the buffer) 121 for uniquely identifying the buffer on the data buffer 2 and stored in the buffer. An identifier 122 of the page that has been set and an update presence / absence 123 indicating whether the page has been updated in the data buffer 2 are set. In the initial state, only the buffer number 121 of each entry is set, the page identifier 122 is NULL, and the update status 123 is set.
Is not updated.

The exclusion management table 13 stores the database 1
For exclusive control of each page. FIG. 4 shows an example of the configuration. Each page is composed of an entry corresponding to each page, and in each entry, an identifier 131 of the page and an identifier (AP in use) 132 of an application program using the page are set. In the initial state, only the page identifier 131 of each entry is set, and the in-use AP 132
It is ULL.

The buffer management means 6 manages the data buffer 2 and performs exclusive control so that the database access processing means 5 can refer to and update each page of the database 1 on the data buffer 2. FIG. 5 shows an example of the processing.

When there is a data buffer request specifying the identifier of the page to be accessed and the application program identifier of the access source from the database access processing means 5, first, the exclusive management table 13 of FIG. Check the value of medium AP132,
It is determined whether or not it can be used (step S1). If another application program is already in use, it is determined that the application is unusable, and the fact is notified to the database access processing means 5 (step S2), and the processing in FIG. 5 is terminated.

If the page is not used by any application program, it is determined that the page is usable, and the identifier of the application program to be accessed this time is set in the entry of the exclusive management table 13 corresponding to the page (step S3). ). Then, the buffer management table 12 is searched with the page identifier, and it is checked whether the page exists in any of the data buffers 2 (step S4). If present,
The number of the buffer is stored in the database access processing means 5
(Step S14), and the processing of FIG. 5 ends.

If the requested page does not exist in the data buffer 2, it is checked with reference to the buffer management table 12 whether there is an empty buffer for reading the corresponding page of the database 1 into the data buffer 2 (step S5). ). If there is a free buffer, the corresponding page is read from the database 1 and stored in the free buffer (step S12).
Is set to the page identifier 122 of the buffer stored this time (step S13). And
The number of the buffer is stored in the database access processing means 5
(Step S14), and the processing of FIG. 5 ends.

If there is no empty buffer in the data buffer 2, a buffer to be flushed is selected by an arbitrary algorithm such as LRU (step S6). Next, if the update presence / absence 123 of the selected buffer is not updated (NO in step S7), the page identifier 122 of the buffer in the buffer management table 12 is temporarily set to NULL to make it an empty buffer (step S11). And
The corresponding page is read from the database 1 and stored in the empty buffer (step S12), and the page identifier 1 of the buffer stored this time in the buffer management table 12 is read.
22 is set to the identifier of the page (step S1).
3). Next, the number of the buffer is notified to the database access processing means 5 (step S14), and the processing in FIG. 5 ends.

If the selected buffer has been updated (YES in step S7), the page stored in the buffer is written to the original page of the database 1 (step S8). Next, it is checked whether or not unupdated page information including the page identifier is recorded in the unupdated page information recording unit 4 (step S9). The updated page information is deleted (step S10). Then, the page identifier 122 of the buffer in the buffer management table 12 is temporarily set to NULL to make it an empty buffer (step S1).
1) The corresponding page is read from the database 1 and stored in the empty buffer (step S12), and the page identifier 122 of the buffer stored this time is set in the buffer management table 12 (step S12).
13), the number of the buffer is notified to the database access processing means 5 (step S14).

Referring again to FIG. 1, the database access processing means 5 refers to and updates the page of the database 1 on the data buffer 2 according to a request from an application program (not shown). When a certain transaction is started by a certain application program, a transaction start journal including the identifier of the transaction is output to the journal file 3. Also, when the identifier of the page to be referred to or updated is notified from the transaction, a data buffer request specifying the identifier of the page to be accessed and the application program identifier of the access source is issued to the buffer management means 6 as described above. When the buffer number is returned, the page is referenced and updated on the buffer identified by the buffer number in the data buffer 2. When the page is updated, the buffer management unit 6 is notified of the updated buffer number, and the buffer management unit 6 changes the update presence / absence 123 of the entry having the corresponding buffer number in the buffer management table 12 to “updated”. Further, each time a page is updated, the database access processing means 5 stores the page identifier, the pre-update image, and the post-update image in the journal stack 11 as one journal.

The commit processing means 7 processes a commit request for a transaction from an application program (not shown). When a commit request is issued, the journal stored in the journal stack 11 for the transaction is read, and the identifiers of the updated pages are all recorded in the unupdated page information recording unit 4. That is, the process of recording the identifier of the updated page in the page identifier 42 of one entry in which the valid bit 41 of FIG. 2 is invalid and validating the valid bit 41 is repeated for all the updated pages. Further, the commit processing means 7 records the transaction end processing start journal, the update journal read from the journal stack 11, and the transaction end processing end journal in the journal file 3. Further, the commit processing unit 7 notifies the buffer management unit 6 of the identifier of the application program that has completed the transaction. The buffer management unit 6 updates the exclusion management table 13 and changes the page being used by the application program to an unused state.

The delayed update processing means 8 performs processing for reflecting the page updated on the data buffer 2 in the database 1 asynchronously with the processing of the commit processing means 7. The delay update processing means 8 is activated by a method in which the timer is activated at regular intervals by a timer. , And an arbitrary method such as a method of starting when the number of updated buffers in the buffer management table 12 becomes equal to or greater than a predetermined value. FIG. 6 shows a processing example of the delay update processing means 8.

First, referring to the unupdated page information recording unit 4, it is checked whether or not there is one or more entries for which the valid bit 41 is valid (step S21). If not, this process ends. If there is, attention is paid to one piece of unupdated page information for which the validity bit 41 is effective (step S22), and the exclusive management table 13 is searched by the page identifier of the unupdated page. It is checked whether or not updating is in progress (step S23). If it is being referred to or being updated, it cannot be reflected in the database 1, so the processing for the unupdated page information is completed, and the process proceeds to step S27. If neither reference nor update is being performed, a buffer number for storing the page is obtained from the buffer management table 12, the updated page is read from the buffer, and written to the original page of the database 1 (step S24). Then, the non-updated page information is deleted from the non-updated page information recording unit 4 (Step S25). That is, the corresponding valid bit 41 is invalidated. Next, the update status 123 of the entry corresponding to the page in the buffer management table 12 is changed to no update (step S2).
6). Then, the process proceeds to step S27.

In step S27, it is checked whether or not unupdated page information remains in the unupdated page information recording unit 4. If it remains, attention is shifted to one of the unupdated page information (step S28), the process returns to step S23, and the above-described processing is repeated. If not, the process ends.

The checkpoint processing means 9 operates periodically and outputs checkpoint information to the journal file 3. The checkpoint processing unit 9 can be activated by any method such as a method of starting at regular intervals by a timer, a method of starting immediately after the commit processing by the commit processing unit 7 is completed, and a method of starting immediately after the operation of the delayed update processing unit 8 Can be adopted. Journal file 3
The checkpoint information output to the server includes an identifier indicating the checkpoint information and all the non-updated page information existing in the non-updated page information recording unit 4 at that time. If there is no unupdated page information at that time, this is included.

The recovery processing means 10 is means for restoring the database 1 based on the contents of the journal file 3 when a host system failure or the like occurs.

Hereinafter, the operation of this embodiment will be described focusing on the operation at the time of delayed update. As an example, assume a situation in which one application program A1 references and updates three pages a, b, and c in the database 1, and another application program A2 references and updates page c. FIG. 7 shows the flow of data and the like at that time.

When referring to and updating three pages a, b, and c of the database 1 in a certain transaction of the application program A1, the database access processing means 5 sends a request to the buffer management means 6 in accordance with a request from the application program A1. Page a,
Requests a buffer for referencing and updating b and c. The buffer management unit 6 executes the processing shown in FIG. 5 to notify the buffer numbers if the pages a, b, and c already exist in the data buffer 2, and if not, the database 1 to the data buffer 2. Pages a, b, and c are read, and their buffer numbers are notified. at this point,
The exclusion management table 13 records that pages a, b, and c are being used by the application program A1. Thereafter, the database access processing means 5 refers to and updates the pages a, b, and c on the buffer of the notified number. When the pages a, b, and c are updated, the update status 123 of the buffer storing the pages a, b, and c in the buffer management table 12 is changed to updated. Further, a journal including images before and after updating the pages a, b, and c is output to the journal stack 11.

Thereafter, the application program A1
Issues a commit request for the transaction, the commit processing means 7
1, the non-updated page information for each of the pages a, b, and c is recorded in the non-updated page information recording unit 4.
The journals for pages a, b, and c are output to journal file 3. Further, the buffer management unit 6 is notified of the identifier of the application program A1.
The buffer management unit 6 updates the exclusion management table 13 so that the pages a, b, and c are not used by the application program A1.

Thereafter, the application program A2
In a certain transaction, an attempt was made to refer to and update page c, so that database access processing means 5
Requests the buffer of the page c from the buffer management means 6, it judges that the page c is usable because it is not used by any application program, and since the page c already exists in the data buffer 2, the buffer number is immediately To the database access processing means 5. At this time, the buffer management unit 6 updates the exclusion management table 13 and manages that the page c is being used by the application program A2.

As described above, it is assumed that the delayed update processing means 8 is activated when the pages a, b, and c are updated by the application program A1 and the page c is referenced and updated by the application program A2. The non-updated page information 4 records three pieces of non-updated page information of pages a, b, and c. Among them, the exclusive management table 13 indicates that the page c is currently being used by the application program A2. Can be recognized.
For this reason, the delay update processing means 8 sets only the pages a and b as the delay update targets. That is, the pages a and b are read from the data buffer 2 and reflected in the database 1, and the unupdated page information of the pages a and b is deleted from the unupdated page information recording unit 4. As a result, only the unupdated page information of the page c remains in the unupdated page information recording unit 4.

Therefore, if the checkpoint processing means 9 is activated at this time, the checkpoint processing means 9
Outputs the checkpoint information including the unupdated page information of the page c to the journal file 3.

As described above, in this embodiment, the delay update is realized using only one data buffer 2.

Next, an example of using the checkpoint information output to the journal file 3 will be described with reference to a simple example.

FIG. 8 is a diagram showing how two transactions update page a of the initial image ABC. Transaction 1 starts after (TS1)
Referring to page a (ABC), the part of C is changed to D (ABD), the part of D is changed to E (ABE), and the processing is completed (TE1). Accordingly, the images (ABC, D) before and after the update, (A
BD, E) are stored in the journal file 3, and the start journal (TS1) of the transaction 1, the end processing start journal (TE1S) of the transaction 1,
Images before and after updating (ABC, D), (ABD, E),
Transaction 1 end processing end journal (TE1
E) is stored. Transaction 2 is
Start after the end of transaction 1 (TS2), AB
E is changed to ACE, and the transaction is still ongoing. For this reason, the image (ABE, C) before and after the update is stored in the journal stack 11, and the start journal (TS
Only 2) is stored. Then, since the delayed update processing has not been performed yet, the page a on the database 1 remains as the initial image ABC.

It is assumed that a host system failure has occurred at the timing of * 1 under the above operating conditions. At this timing, since transaction 1 has been completed, it is desired to restore page a of database 1 to ABE. The recovery processing means 10 searches the journal file 3 in time order from the latest journal to the past journal, and finds the most recently output checkpoint information CP1 (empty). Then, since the found checkpoint information CP1 (empty) indicates that there is no unupdated page information, the latest (newer) journal is input from that time, and the updated image is used.
Transaction 1 end processing end journal (TE1
Until E), the processing for reflecting the updated image is executed.
That is, the updated images D and E are sequentially reflected on the portion C of the content ABC of the page a at the time of the checkpoint CP1, and the ABE is restored.

On the other hand, suppose that a host system failure has occurred at the timing of * 2 or * 3. At this timing, since the transaction 2 has started but not ended, it is desired to restore the database 1 to the time of the ended transaction 1. That is, ☆ 1
The page a is restored up to the ABE in the same manner as the above. The recovery processing means 10 searches the journal file 3 in time order from the latest journal to the past journal, and finds the checkpoint information CP2 (a) output most recently. And this found checkpoint information C
Since P2 (a) indicates that there is unupdated page information,
The search is continued until the checkpoint information in which the unupdated information of the page a described in the checkpoint information CP2 (a) is not recorded is found, and CP1 (empty) is obtained. Then, a journal that is more recent than that point in time is input, a process of reflecting the updated image is executed using the updated image up to the end processing end journal (TE1E) of transaction 1, and the contents of page a are restored to ABE. I do.

As described above, the checkpoint information output by the checkpoint processing means 9 gives information as to how far back in the database recovery, the contents of the database 1 are guaranteed. Therefore, if the activation cycle of the checkpoint processing unit 9 is too long, the period going back to the past becomes longer. On the other hand, if the period is shortened, the overhead due to the operation of the checkpoint processing means 9 increases. Therefore, in an actual system, the activation cycle of the checkpoint processing means 9 is determined in consideration of both.

FIG. 9 is a block diagram showing an example of a computer to which the present invention is applied. The central processing unit 1000 includes a central processing unit 1000, a secondary storage device 1001, and a recording medium 1002.
Contains. The secondary storage device 1001 is composed of a magnetic disk device or the like, and is used for storing the database 1 and the journal file 3 shown in FIG. Recording medium 10
Reference numeral 02 denotes a machine-readable recording medium such as a CD-ROM, a semiconductor memory, and a magnetic disk, which records a database management program. The database management program recorded on the recording medium 1002 is
When the CPU 10 starts up or when the power is turned on, the CPU 10
03 and stored in the main storage device 1004 in FIG.
, The unupdated page information recording unit 4, the buffer management table 12, the exclusion management table 1
3, the storage unit such as the journal stack 11 is initially generated, and the operation of the CPU 1003 is controlled so that the database access processing unit 5, the buffer management unit 6, the commit processing unit 7, the delay update unit shown in FIG. The processing unit 8, the checkpoint processing unit 9, and the recovery processing unit 10 are realized.

[0048]

As described above, according to the present invention, information for specifying a page updated on the data buffer at the time of a commit request from an application program is recorded as unupdated page information in the unupdated page information recording unit. After that, out of pages indicated by the unupdated page information recorded in the unupdated page information recording unit, only pages not currently used by any application program are reflected in the database asynchronously with the processing by the application program. Thus, since the delay update is realized, it is not necessary to provide two data buffers as in the related art, and the required capacity of the main storage can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of an embodiment of a database management device of the present invention.

FIG. 2 is a diagram illustrating a configuration example of a non-updated page information recording unit.

FIG. 3 is a diagram illustrating a configuration example of a buffer management table.

FIG. 4 is a diagram illustrating a configuration example of an exclusion management table.

FIG. 5 is a flowchart illustrating an example of a process of a buffer management unit.

FIG. 6 is a flowchart illustrating a processing example of a delay update processing unit.

FIG. 7 is an operation explanatory diagram of the embodiment of the present invention.

FIG. 8 shows two transactions as an initial image ABC
It is a figure which shows a mode that the page a of this is updated.

FIG. 9 is a block diagram illustrating an example of a computer to which the present invention has been applied.

FIG. 10 is a block diagram showing a conventional example of a delay update method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Database 2 ... Data buffer 3 ... Journal file 4 ... Unupdated page information recording part 5 ... Database access processing means 6 ... Buffer management means 7 ... Commit processing means 8 ... Delayed update processing means 9 ... Checkpoint processing means 10 ... Recovery Processing means 11: Journal stack 12: Buffer management table 13: Exclusive management table

Claims (8)

[Claims] 1. A computer which performs reference and update to a database on a secondary storage on a data buffer provided in a main storage and reflects updated pages on the database asynchronously with processing by an application program. At the time of a commit request from a program, a step of recording information specifying a page updated on the data buffer by the application program as unupdated page information in an unupdated page information recording unit, asynchronously with processing by the application program, Reflecting only pages not currently used by any application programs among the pages indicated by the unupdated page information recorded in the unupdated page information recording unit in the database. Based management methods. 2. If the non-updated page information exists, checkpoint information including the existing non-updated page information is periodically output to a journal file by referring to the non-updated page information recording unit. 2. The database management method according to claim 1, further comprising a step of outputting checkpoint information indicating that no unupdated page information does not exist to the journal file. 3. A computer which performs reference and update on a database on a secondary storage on a data buffer provided in a main storage and reflects an updated page on the database asynchronously with processing by an application program. At the time of a commit request from the program, commit processing means for recording information identifying the page updated on the data buffer by the application program as unupdated page information in the unupdated page information recording unit, and which page is currently in use And a table holding information indicating that the page is not currently used by any application program among the pages indicated by the non-updated page information recorded in the non-updated page information recording unit asynchronously with the processing by the application program. Is the Database management system which comprises only page shown Le and delay updating processing means for reflecting in the database. 4. If the non-updated page information is present, checkpoint information including the existing non-updated page information is periodically output to a journal file by referring to the non-updated page information recording unit. 4. The database management apparatus according to claim 3, further comprising a checkpoint processing unit that outputs checkpoint information indicating that no unupdated page information does not exist to the journal file. 5. A computer for managing a database on a secondary storage, a data buffer provided on a main storage and having a capacity capable of storing a plurality of pages constituting the database, wherein the data is stored in accordance with a request from an application program. Database access processing means for referring to and updating a page on a buffer; buffer management means for reading a page constituting the database into the data buffer in response to a request from the database access processing means; and indicating which page is currently in use. A table for holding information, a non-updated page information recording unit provided on the main memory, and, when a commit request is made from the application program, information for identifying a page updated on the data buffer by the application program, has not been updated. page A commit processing unit that records information in the unupdated page information recording unit; and any application program among pages indicated by the unupdated page information recorded in the unupdated page information recording unit, asynchronously with processing by the application program. And a delayed update processing means for reflecting only the pages which are not currently used in the table in the table to the database. 6. A journal file which is provided on a secondary storage and holds an update image journal of page update by an application program, and the non-updated page information is periodically referenced by referring to the non-updated page information recording unit. If it exists, checkpoint information including the existing non-updated page information is output to the journal file; otherwise, checkpoint information indicating that no unupdated page information exists is output to the journal file. 6. The database management apparatus according to claim 5, further comprising a point processing unit. 7. The database management apparatus according to claim 6, further comprising recovery processing means for restoring the database based on the contents of the journal file. 8. A computer for managing a database on the secondary storage, a database access processing means for referring to and updating a page on a data buffer of the main storage in accordance with a request from an application program, and in accordance with a request from the database access processing means. Buffer management means for reading a page constituting the database into the data buffer; when a commit request is made from an application program, information specifying a page updated on the data buffer by the application program is stored as unupdated page information in a main storage. A commit processing means for recording in the above-mentioned non-updated page information recording unit; and an application among the pages indicated by the non-updated page information recorded in the non-updated page information recording unit, asynchronously with the processing by the application program. Delaying update processing unit, a machine-readable recording medium recording a program to function as to reflect only the pages that the Activation program is not currently being used in the database.