JPH02105950A - Recovery system for data base - Google Patents

Abstract

PURPOSE:To shorten the recovery time of a data base by providing a journal selecting means which reads a journal regarding the same data base page out of a loop continuously in time-series order. CONSTITUTION:The journal regarding the same data base page is taken out of the pool by the journal selecting means 9 continuously in the time-series order. The journal regarding the same data base page is sent to an input/output buffer 3 in the time-series order (successively from the head) to make a recovery regarding the data base page, so the data base is never misrecovered. Further, input/output operation of the data base input/output buffer 3 is performed only once for each page. Consequently, the number of times of the input and output of data base pages is decreased to shorten the recovery time.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a database recovery method in a database system, and in particular, the following steps are taken when recovering a database after a physical media failure or system shutdown. The present invention relates to a database recovery method suitable for reducing the number of database page inputs and outputs while suppressing the amount of space occupied on virtual memory.

[Conventional technology]

Generally, in an f-tapace system, a physical media failure or system shutdown occurs during transaction processing that involves updating a large number of databases.

The amount of journals to be handled during subsequent database recovery will be enormous.

Conventional as this kind of database recovery method.

For example, there is a method described in Japanese Unexamined Patent Publication No. 151754/1983. In this method, for each journal entry,
Enter the appropriate page from the database into the buffer,
After each journal is covered one by one, it is output to the database.

[Problem to be solved by the invention]

In the conventional method described above, the number of input/outputs for database pages increases in proportion to the amount of journals, so if the number of journals handled increases as described above, the rate of input/output errors will increase as the number of input/outputs increases. This caused problems in terms of both performance and reliability. Additionally, this increased number of I/Os increases database recovery time.

During this time, there were problems such as the inability to execute online transactions.

However, the above-mentioned conventional technology does not take into consideration the problems of an increase in the incidence of input/output errors and an increase in database recovery time due to an increase in the number of input/outputs.

Therefore, an object of the present invention is to solve the problems of the prior art described above, while suppressing the amount of virtual memory occupied.

A database recovery method that shortens database recovery time by reducing the number of inputs and outputs of database pages during database recovery, and improves the reliability of database recovery by reducing the incidence of input/output errors. Busy to provide.

[Means to solve the problem]

In order to achieve the above objects, the database recovery method of the present invention includes a means for extracting a journal from a journal file, a means for inputting and outputting database pages (input/output buffer), and an input/output buffer (input/output buffer) for retrieving a journal from a journal file. In addition to the means for recovery using journals, there is also a pool (journal Boolean) in virtual memory that stores journals retrieved from journal files, and 1. For recovery processing, journals related to the same database page chain are retrieved from the pool. A journal selection means for successively extracting journals in series order is provided.

Also, journals regarding database pages currently being recovered and raw database pages /L/&! , followed by I/O for recovery processing (directly to ZFA), but once the journals for different database pages are
It is stored in the dual Boolean and is configured in 5 formats.

In this case, if there is no free area (unused area) for the journal to be stored in the journal pool, the journal stored in the leading area of the journal pool and the free area of the pool created by it. Repack the journal so that everything is in one place.

The structure is such that the journal that was intended to be stored is stored in the free area (terminal area) that is generated (collected) as a result.

[Effect]

The effect based on the above configuration will be explained.

Provide the journal selection method mentioned above? , : From pressure.

Journals about the same database page.

Since the data is continuously sent from the pool to the input/output buffer in chronological order (starting from the beginning) and recovery related to the database page is performed, there is no risk of erroneously recovering the database. In addition, input/output to the database input/output buffer only needs to be performed once for the same page.
This can be achieved by reducing the number of database page input/outputs and shortening the recovery time.

Journals are stored in the pool in response to journal buttons related to database pages different from the database page currently being recovered. This journal is stored in the free area of the pool, but if there is not enough free space to accommodate the journal to be stored, the journal in the top area of the pool and the same database page as this journal will be stored. The recovery process for all journals that have the data is performed in chronological order. the result.

Free space is created in the pool, but the journal is refilled so that these free spaces are gathered in one place (at the end of the pool) (the journal is compressed so that there is no space between the beginning of the pool and the middle) ). Then, a new journal from the journal file is stored in the free area at the rear end of this pool. In this manner, the chronological order of the journals stored in the pool is guaranteed, so that there is no possibility of erroneous recovery of the database.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to one drawing.

FIG. 1 is a block diagram of the database recovery mechanism of this embodiment, in which 1 is a journal, 2 is a journal boolean, 3 is an input/output buffer, 4 is a journal file that stores journal l, and 5 is a database.

6 is a database page, 8 is a journal retrieval mechanism for retrieving journal 1 from journal file 4, 9
1 is a journal selection mechanism, 10 is a database page recovery mechanism, 11 is a database page input/output mechanism, and 12 is a database recovery mechanism that centrally controls the journal handling mechanism 8 or the database page input/output mechanism. The journal selection mechanism 9 selectively sends the retrieved t-journal to the journal boolean 2 or the input/output buffer 3 (via the database page recovery mechanism 10) by the journal extraction mechanism 8, or selects a required journal page in the journal pool 2. Take out the crowd and cover it 3
It is possible to perform control such as sending to The database page input/output mechanism 11 can control the sending of pages to be recovered from the database 5 to the input/output buffer 7 and the transfer of one recovered page from the input/output buffer 3 to the database 5.

Next, the more detailed structure and operation of the embodiment of the present invention will be explained with reference to FIGS. 2(a) to (g) and FIG. 3.

FIGS. 2(a) to 2(g) are diagrams for explaining the operations of various functions in this embodiment, and FIG. 3 is a flowchart showing the operations of this embodiment.

In FIGS. 2(a) to (gl), parts having the same configuration as those in FIG.
This is a database page similar to 6.

(1) As shown in FIG. 2 (al), when the first journal 1 is input from journal file 4 during recovery of database 5 (steps 301 and 3 in FIG.
02, 303), inputs the input journal IK corresponding database page 6 from the database 5 to the input/output buffer 3 (step 304), and then inputs the input journal 1 to the database page 6 input to the input/output buffer 3. Cover.

(2) As shown in FIG. 2 fbl, if the next input journal l from the journal file 4 is a journal related to the same database page as the database page 6 currently being recovered (steps 301 to 303, 305, 30
6”), overlay input journal 1 over database page 6 input to input/output buffer 3 (30
7).

(3) As shown in FIG. 2 (cl), the next input journal l from journal file 4 is a journal relating to a database page different from the database page 6 currently being recovered (steps 301 to 303305, 306
), and there is enough free space to store the input journal 1 at the end of journal Boolean 2 (step 308), the input journal 1 is stored in journal Boolean 2.
(step 312).

+4] JR2 diagram (as shown in dl, the input journal 1 from journal file 4 contains database pages 6 and 14 that are currently being recovered).
r −+ A/ (su-y-tsubu 301~3(')3
, 305.

306), and there is not enough free space to store the input journal 1 at the end of the journal pool ((step 309), and the total free space in the journal Boolean 2 is larger than the size of the input journal 1). ,
And if the usage rate of journal pool 2 (total size of stored journals/size of journal pool) is less than 1/2 (step 310), refill the journals stored in journal boolean 2. After creating a free space a at the end of the journal pool 2 as shown in FIG. 2(C) (step 311), the input journal 1 is stored in the journal pool 2 (step 312).

(5) As shown in Figure 2(e), click button 7 to indicate that the next input journal 1 from journal file 4 is currently being recovered.
'-p Journal for database pages different from base page 6 (steps 301 to 303, 305°3
06) and there is not enough free space at the end of journal pool 2 to store the input journal 1 (
Step 308), free area a in journal Boolean 2
is smaller than the size of the input journal 1 or the usage rate of the journal pool 2 exceeds 1/2 (step 309 or 310), the input journal 1
As shown in FIG. 2(f), the first journal among the journals stored in the journal Boolean 2 is taken out (step 313).
14).

After outputting the database page 6 currently being recovered that has been input to the input/output buffer 3 to the database 5 (step 315), the database page 7 corresponding to the journal taken out from the journal pool 2 is transferred from the database 5 to the input/output buffer 3. (Step 316)
, the relevant journal (the first journal taken out from journal pool 2) is placed over database page 7 input to input/output buffer 3 (step 317).
. Steps 318, 319) in which all journals related to the same database page 7 that is currently being recovered are retrieved from the journal pool.

It is placed over the database page 7 input to the input/output buffer 3 (step 317). If there are no journals related to the same page, the journal processing flag in the journal boolean is turned off (step 320). The process of extracting journals from journal pool 2 is continued until the total free space a in journal boolean 2 is larger than the size of input journal l kept in the pending state, and the usage rate of journal pool 2 is 1/2 or less. repeat. As shown in Figure 2 (d) & C, the total amount of free space a in journal Boolean 2 is larger than the size of input journal 1, which has been put on hold, and the usage rate of journal pool 2 has become 172 or less. In this case, after refilling the journal stored in the journal Boolean 2 and creating a large amount of free space a at the end of the VC and journal pool as shown in Figure 2 (C), the corresponding Store input journal 1 in journal pool 2.

(6) As shown in FIG. 2(g), when there is no journal in journal file 4 (step 302).

All the journals remaining in the journal pool 2 are taken out for each journal related to the same database page and are placed over the database page 6 input to the input/output buffer 3. When the database page is switched, the database page 6 currently being recovered is output to the database 5 and then taken out from the journal pool 2.
, inputs the corresponding database page 7 from the database 5 into the input/output buffer 3iC, and overlays the database page 7 input into the input/output buffer 3 with the journal. The extraction of the journal remaining in the journal pool is completed (step 321).
), the database page 7 inputted in the input/output buffer 3 is outputted to the database page 5 (steps 322 and 323).

According to this embodiment, compared to a method of buffering input/output buffers of database pages, the amount of virtual memory occupied can be reduced. Furthermore, according to this embodiment, the number of times database pages are input/output to the database can be reduced compared to a method in which database pages are input/output for each journal, thereby reducing database recovery time. .

〔Effect of the invention〕

As described above in detail, according to the database recovery method of the present invention, a plurality of journals related to the same database page can be processed by one database page input/output, thereby reducing the database recovery time. Furthermore, since the journal is buffered using a journal pool, it has the effect of reducing the amount of virtual memory occupied compared to a method that has multiple input/output buffers for database pages. Furthermore, since the recovery of the same database page by several journals is performed successively in chronological order, the number of input/outputs is reduced and the incidence of input/output errors is reduced.

This has the effect of improving the reliability of database recovery.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a database recovery mechanism according to an embodiment of the present invention, FIGS. 2(a) to (g) are diagrams for explaining the operation of the embodiment of the present invention, and FIG. 3 is a flowchart showing the operation of the embodiment. 1...Journal, 2...Journal Boolean. 3...I/O buffer, 4...Journal file, 5...Database, 6,7...
...Database page, 8...Journal retrieval mechanism, 9...Journal selection mechanism, 10...Database page recovery mechanism, 11
...Database page input/output mechanism, 12...
...Database recovery mechanism. Figure 2 (b) (C) Figure 2 (d) (e) (f) Figure 2

Claims (1)

[Claims] 1. A journal file storing a journal that is an update history of a database, means for retrieving the journal from the journal file, input/output means for the pages of the database, and information input to the input/output buffer. In a database recovery method that includes means for recovering database pages using a journal, a pool on a virtual memory for storing the journal retrieved by the journal retrieving means and a pool for recovery processing by the restoring means are identical. A database recovery method, comprising: journal selection means for successively extracting journals related to database pages from the pool in chronological order. 2. Journals related to the same database page as the database page being recovered by the recovery means:
The journal is sent directly from the journal file to the input/output buffer for continuous recovery processing, and journals related to database pages different from the database page being recovered by the recovery means are temporarily stored in the pool. 2. The database recovery method according to claim 1. 3. If there is no free space in the pool for the journal to be stored, the recovery process is performed consecutively using the journal stored in the top area of the pool and the journal that has the same database page as the stored journal. The method is characterized in that the journals are repacked so that the resulting free space in the pool is gathered in one place, and the journal to be stored is stored in the free space created thereby. The database recovery method according to claim 1 or 2.