JPS62147544A - Roll-back processing system - Google Patents

Abstract

PURPOSE:To complete the data base roll-back processing in a short period by using a main memory area where a user program for transaction processing is executed after resetting said memory area as a large-scale buffer which reads a number of blocks of an updating information storing file. CONSTITUTION:The blocks including the head block 1 through the block 100 are read into the 1st buffer 3a from an updating information storing file 2d for restoration of a data base. When this reading operation is through, the updating information on the blocks 1-100 are used successively for execution of the roll-back processing of the data base. At the same time, the next 100 blocks are read into the 2nd buffer 3b. When the roll-back processing is through with the first 100 blocks, the next 100 blocks of the buffer 3b are used for execution of the roll-back processing. Meanwhile other following 100 blocks are read into the buffer 3a with which the processing is through from the file 2d by a request. This action is repeated to carry out the roll-back processing at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a processing method for quickly rolling back a database after a failure occurs in a transaction processing system.

[Conventional technology]

Conventionally, in database rollback processing after a failure occurs in a transaction processing system, each update information is read from the database update information storage file using an update information input buffer set in the control section 7 of the transaction processing system. The method was to perform database rollback processing one block at a time.

[Problem that the invention seeks to solve]

In such a conventional database cool-back method after a failure occurs in a transaction processing system, reading from an information storage file is sequentially performed one block at a time. This information storage file is read the same number of times as the number of blocks in the information storage file because the block size and input buffer size are the same. If the number is large, it will take a long time to read all the blocks in the information storage file. Therefore, there is a drawback that it takes a long time to complete all database rollback processing after a failure occurs in the transaction processing system.

An object of the present invention is to provide a data rollback processing method that eliminates such drawbacks and allows database rollback processing to be completed in a short time.

[Means for solving problems]

The configuration of the present invention is a transaction processing system having an update information storage file on an external storage device consisting of a large number of blocks in which database update image records are stored. In the rollback processing method, the main storage area where the user program for transaction processing is executed under the control of the transaction processing control unit, which is not used during the rollback processing, is used as the main storage area of the update information storage file. By resetting and using it as a large buffer that reads a large number of blocks to perform a database rollback, when the transaction processing is restarted, it is immediately reconfigured as a user program execution area after the rollback is completed. shall be.

[Effect]

According to the configuration of the present invention, after a failure occurs in the transaction processing system, during rollback processing, the main storage area managed by the transaction processing system control unit is not used as a user program execution area for transaction processing. By reconfiguring the user program execution area as a large buffer for reading update information and using this buffer, it is possible to shorten the time for reading from the information storage file and shorten the rollback processing time. . Furthermore, when restarting transaction processing after completing rollback processing, the used large buffer can be reset as the user program execution area and the subsequent transaction processing can be normally completed by directly performing the T-T.

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

FIG. 1 is a block diagram of a system including an embodiment of the present invention, and this embodiment shows the poor management of a database in a database rollback process of a transaction processing system. In this figure, g(+') shows the main storage device 4 in which the update information read buffer necessary for database rollback exists, the database file (magnetic disk drive) 1 to be rolled back, and the update information storage file 2. The database file 1 is in the process of being updated and is rolled back based on the information in the update information storage file 2 based on each processing unit of a plurality of transactions that are being updated.

FIG. 2 is a diagram showing the structure of the update information storage file 2 that is the first cause, and is the update information storage file 2a.

There are two files, update information storage file 2b,
Assume that there are a large number of blocks in one file, for example, 10,000 blocks each for a total of 20,000 blocks, and from block (1) to block (20,000) in chronological order from update information storage file 2a to update information storage file 2b. Suppose that information is stored.

FIG. 3 is a configuration diagram of the first buffer 3a and second buffer 3b in which the user program execution area on the main storage device 4 of FIG. The storage size can be set.〇For database recovery, block (1) from the first block (1) from the update information storage file 2.
00) to the first buffer 3a and the database is rolled back using the update information from block (1) to block (100). During this process, the next 100 blocks are transferred to the second buffer 3.
The process of reading into b is performed in parallel. When the rollback process for the first 100 blocks is completed, the rollback process is performed for the next 100 blocks on the second buffer 3b.

During database rollback processing for 100 blocks of information on the second buffer 3b, the next 100
The first block that has been processed from the information storage file
A read request is made to the buffer 3a. The same process is repeated 200 times up to 20,000 blocks to efficiently read all blocks of the update information storage file 2a and the update information storage file 2b and perform database rollback processing at high speed.

FIG. 4 is a chart of the database rollback process 80 in the online system of this embodiment.

First, in order to detect transactions that are in an incomplete processing state, the update information storage file 2a and the update information storage file 2b are sequentially read from block (1) to block (20,000) using the method described above (step 11)
, T, the transaction numbers of all uncompleted transactions are detected and stored (step 12). Playing card comb 1 with incomplete processing detected in the next step 13
Due to database rollback processing related to
The time-series example is read backwards in the above-described manner, and a rollback is performed on the database related to incomplete transactions recorded in step 8 of the above process.

Next, in step 14, it is determined whether the transaction processing should be restarted, and if the transaction processing is to be restarted, the program is stored in the user's Groderam memory again in step 15, and the program is terminated.If the transaction processing is not restarted, the program is terminated.

In this example, for database role power,
After reading the update information accumulation file 400 times, 4
Although 10,000 blocks will be processed, the number of reads can be reduced to 1/100 compared to reading each block sequentially.

〔Effect of the invention〕

As explained above, according to the present invention, the size of the buffer for reading the update information storage file can be freely set according to the system scale, so the number of times the update information storage file is read during rollback can be reduced. Since the amount of time waiting for read completion is also reduced, the time required for database rollback processing after a failure occurs in the transaction processing system can be significantly reduced.

[Brief explanation of drawings]

FIG. 1 is a block diagram illustrating an embodiment of the present invention.
The figure shows the configuration of the update information storage file in this example, Figure 3 shows the configuration of the reset buffer in this example, and Figure 4 shows the database rollback process used in this example. It's a yard. 1...Magnetic disk device for database (file), 2...Magnetic disk device for 6 update reports, 2a, 2b...Update information storage file, 3a... ...First buffer, 3b...Second buffer 7a, 4...Main storage device. Agent Patent Attorney Shindai Uchihara

Claims (1)

[Claims] In the rollback processing method for post-processing or restart processing after a forced stop due to hardware failure of a transaction processing system that has an update information storage file on an external storage device consisting of a large number of blocks for storing database update image records, this The main storage area in which the user program for transaction processing is executed under the control of the transaction processing control unit, which is not used during rollback processing, is reused as a large buffer for reading a large number of blocks of the update information storage file. A rollback processing method characterized in that by setting and using a database to roll back, when restarting the transaction processing, the area is immediately set as an execution area for a user program after the rollback is completed.