JPH03168847A - High-speed data base control system - Google Patents

Abstract

PURPOSE:To prevent the generation of exclusive queuing or dead lock by grouping updating requests having the same exclusive section as a block and serially activating respective blocks having respective identical exclusive sections. CONSTITUTION:Records having the same exclusive section in a data base are grouped as a block and respective blocks having respectively identical exclusive sections are controlled so as to be serially processed at every block in updating processing. For instance, a transaction record group having the same exclusive section are grouped as a block on the data base updating information forming side 5 and sent to the data base updating information reflecting side 6, which serially processes respective blocks by respective data base reflecting processing 15 to 17. Consequently, the generation of exclusive queuing and dead lock between respective blocks can be prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a high-speed database control method in a computer application system having a database.

[Conventional technology]

In the conventional technology, as shown in FIG. 10, transaction records generated on the database update information creation side 5 (A-1, A-2, B-1, etc. shown in FIG. 10), where A-1 is for database A This update information is sent to the side 6 which reflects it on the database in units of (first update information, A-2 also indicates the second update information). Transmission of transaction records from the database information creation side 5 to the reflection side 6 is generally performed through multiple processing to speed up the processing, and Figure 10 shows a case where the multiplicity of transmission and reflection processing is the same. . On the side that reflects updated information to the database, processes 7 and 8.9 for reflecting the updated information in the database are activated for each transaction record sent. Transaction records are generated in databases A, B, and C.
, and for this reason, there is a possibility that an exclusive conflict between transaction records may occur on the database update information reflecting side when updating the database. Database A
, B, and C have different exclusive control classifications, and in FIG. 10, the transaction records A-1 and A-2 have the same exclusive unit, so w4 match occurs.

Incidentally, related techniques include Japanese Patent Application Laid-open No. 191249/1983.

[Problem to be solved by the invention]

The above-mentioned conventional technology has the following problem in that it cannot perform database update processing at high speed.

(1) The transaction record, which is the unit of a database update request, and the startup unit of the update process are the same, and each transaction record requires a program processing step to start and end the database update, which reduces the performance of the entire system. descend. , (2) Since database update requests are originally generated regardless of the range or unit of exclusion determined by the database design, if random update requests are started as they are, mutual exclusion waits will occur for database reflection processing. As the number of update requests increases, the amount of time waiting for exclusive use becomes enormous and the performance of the entire system deteriorates.

(3) As a countermeasure for the program processing steps for starting and terminating the database update process mentioned in issue (1), among the transaction records generated on the database update information creation side, n records are combined into one. The number of program processing steps for startup and termination can be reduced to 1 7 n by executing processing and activating and terminating blocked block units on the database update information reflecting side, but in this case, the following problems occur.

(a) Occurrence of deadlock As shown in Figure 12, databases A and B (A and B
(indicates the classification of exclusive control) are respectively updated records A-1, A-2, B-1, and B-2. In database reflection processing 10, an exclusive request is made to database A in order to reflect A-1 in database A. In database reflection processing 11, in order to reflect B-1 to database B,
Make an exclusive request to. After this, database reflection processing 1
0 makes an exclusive request to B in order to reflect B-2 in database B, but since the database reflection process l1 has already made an exclusive request to B due to the reflection process by B-1,
Database reflection process 10 is database reflection process}1
Waiting for completion. On the other hand, the first step of database reflection processing is
Although A-2 makes an exclusive request for reflection processing, A-2 similarly waits for the database reflection processing 10 to end. In this way, a deadlock in which the database reflection process 10 and the database reflection process 11 each wait for the other to finish is likely to occur.

(b) Increase in exclusive waiting time (see Figure 4) When transaction records, which are database update information that occur randomly, are simply blocked (for example, in the order of occurrence) and reflected in the database, a series of transaction records in the activated block Because the database exclusion units of transaction records are complicated, exclusion waits occur, the degree of multiplicity cannot be increased, and the system performance of database reflection processing deteriorates. For example, as shown in FIG. 13, in the database reflection process 12, an exclusive request is made to databases A and H in order to simultaneously update transaction records A-1, B-1, and B-3.

On the other hand, in database reflection process 13, transaction record A-2
, B-3, and C-2, but since an exclusive request has already been made to A in database reflection processing 12, database reflection processing 13 waits for the completion of database reflection processing 12. Become. In addition, in the database reflection process 14, transaction records C-1, B-2,
In order to update C-3, an exclusive request is made to database C and then to database B, but since the database reflection process 12 has already made an exclusive request for B, the process waits for the database reflection process 12 to finish. As a result, only the database reflection process 12 progresses.

As mentioned above, simple blocking has a negative impact on system performance. The purpose of the present invention is to solve this problem. [Means for solving the problem] In order to achieve the above purpose, the following can be done.

(1) The starting unit of the database update information reflection process has been changed from the conventional transaction record unit to a block start method that blocks multiple records, reducing the program processing steps required to start and end the database reflection process to 1 /
Reduce to n.

(2) A group of transaction records (A-1, A-2, A-3 or B-1, B-2,
B-3, etc.) are collectively blocked before performing database update processing.

(3) Blocks consisting of transaction records that have the same exclusive classification are processed serially in database update processing to avoid exclusion conflicts and also prevent reversal of data reflection order.

For example, what is shown in the first construction drawing is an example of this, where a group of transaction records that have the same exclusion are database update information creation or side 5
The data is blocked and sent to the database update information reflecting side 6. On the database update information reflection side 6, each database reflection process 15, 16, and 17 is processed in series. [Operation] According to the present invention, the first means reduces the program processing steps involved in starting and terminating the database reflection process to 1/1.
The second and third means eliminate exclusive waits and deadlocks, and also protect the database update order. [Example] An example of the present invention will be described below with reference to the drawings. As an example, a case of a bank online system will be explained.

This invention applies even when the database update request generating side and the reflecting side are located in the same center (or the same computer) (hereinafter referred to as a local system), or when they are located in different centers and connected by a line, etc. (hereinafter referred to as a remote system) is also applicable.

(.) Example of local system In the bank online night center input processing, database update request data 18 such as salary transfers and utility bill withdrawals brought in by companies and other contractors is reflected in the database through online processing using file input. let The present invention speeds up this processing. A bank's core business is accounting processing, and each branch of a bank serves as a point of contact with policyholders, so databases are created for each branch (hereinafter referred to as "branch"). The unit of exclusion in a database may be a subscriber or a group of multiple subscribers within the same store, but since accounting processing needs to be understood on a store-by-store basis, there are many databases that require exclusion on a store-by-store basis. Therefore, in the updating process of the bank online database, it is most appropriate to use store as the classification key for the same exclusive classification according to the present invention.

The transfer data 18 shown in FIG. 1 that has been brought in for the night center input processing is not in the input data file format for online processing as it is, so the computer 2
0, it is necessary to perform format conversion processing once. At this time, as part of the format conversion, store numbers are assigned to each store that has an account to be updated (■). 100
, 200, and 300 indicate that the stores correspond to store numbers 100, 200, and 300, respectively.

Transfer data to the account for store number 100 is store number 100.
It is stored in the human data file 19 (■). On the database update side 6, when the computer 21 reads the input data for store number 100 (■), it blocks multiple pieces of update information for store number 100 together for reflection processing (■), and the database reflection process 1l! Start ll22 (■). When ending the reflection process, perform the termination process (■).

Processing for the block with store number 100 and the block with store number 200 is started at the same time, but a serial schedule method is used in which only one block of the same tOO is started at the same time. On the database reflection processing side 6, a control table corresponding to store numbers is required so that blocks consisting of transaction records of the same store number are started serially, and this table contains information indicating whether or not the database reflection processing for the store number is being executed. Contains a status management flag. FIG. 1 shows that reflection processes 22 and 23 with flag 1 are being executed, and reflection processes 24 with flag O have not been executed.

FIG. 3 shows the processing flow indicated by ■ in FIG. 1. When the input data includes an input store number, the data is distributed according to the input store number, and when there is no input store number, the data is distributed according to the account store number.

FIG. 4 shows the processing flow of ``■'' in FIG. 1. If there is space to accommodate a transaction record in the block that has already been created (step 41 YES). Blocking processing is performed (step 42), and when the block becomes full (step 43 YES), a new data block is created and prepared (step 44). If there is no space in the block (step 41 NO). A new data block is created (step 45), and blocking processing is performed (step 46).

FIG. 5 shows the processing flow of ■ in FIG.
S), and sets the flag to "1" (step 53).
), a reflection process is performed (step 54). If the flag is already "1" (step 52 NO), the data is placed in a waiting state (step 55).

FIG. 6 shows the processing flow of ■ in FIG. 62NO). The flag is reset to "O" (step 63).

If there is waiting data (step 62 YES), start the reflection process for the waiting data (step 64).
.

FIG. 7 shows the structure of the control table, in which a flag 72 is provided corresponding to each store number 7l.

Figure 8 shows an example of the transaction record format, (a)
(b) is an example when the transaction record is data brought in by the contractor.
is an example of transaction input data from a branch terminal. In the local system shown in the first @, data (.) is input.

(b) Remote system implementation example As bank online system failures have an increasing impact on society, a backup system in case of disasters has become necessary. As shown in Figure 2, the backup system side 26 has the same database as the primary system side 25, and updates the primary system side database (primary ledger) 27 in real time to the backup side database (secondary ledger).
It is necessary to reflect this on the 28th. The present invention is applied to the main reflection process.

In the computer 30 on the main system side 25 shown in FIG. 2, an online accounting processing system is operating. If the ordinary deposit database 27-1 related to store number 100 is updated by the Otsutsu deposit withdrawal process from the business store with store number 100 (■), it is necessary to reflect this on the backup system side 26.

The regular ledger database update information is extracted (■), blocked for writing to the transmission queue along with data for other store numbers 100 by sorting by store number (■), and written to the transmission queue 32 (■). The data in the transmission queue 32 is sent to the backup system reception queue 33 via the line 29 (
■). At this time, serialization processing is realized by providing a transmission queue 32, a reception queue 33, and a transmission route for each store number. The computer 31 blocks the database update information taken out from the reception queue 33 for reflection processing (■), and starts database reflection processing (■).
When ending the reflection process, perform the termination process (■). In the database reflection processes 34, 35, and 36, a control table is provided for each store number so that the same store number block is started serially (its structure is the same as in the case of the local system). Note that the processing flow of (2) is shown in FIG. 3, the processing flow of (2) is shown in FIG. 4, the processing flow of (2) is shown in FIG. 5, and the processing flow of (2) is shown in FIG. 6. Further, in the remote system shown in FIG. 2, the branch terminal transaction data shown in FIG. 8(b) is input.

It goes without saying that the transaction record blocking processing indicated by ■ in FIG. 1 and ■ in FIG. 2 may be performed by the database update information creation side 5 and the primary system side 25, respectively.

Figure 9 shows the conventional method that does not block transaction records (
The processing time is compared between a) and the blocking method (b) of the present invention. In the case of the present invention, the startup processing and termination processing are reduced by blocking. Processing time is reduced.

[Effects of the Invention] According to the present invention, update requests having the same exclusive classification are grouped together into blocks, and blocks having the same exclusive classification are activated serially, so that no exclusive wait or deadlock occurs. This enables high-speed database reflection processing.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing an embodiment of the present invention in a local system, Fig. 2 is a diagram showing an embodiment of the present invention in a remote system, Fig. 3 is a flowchart of store number distribution processing in the embodiment, and Fig. 4 is a diagram showing an embodiment of the present invention in a remote system. 5 is a processing flowchart of blocking for database update information reflection processing of the embodiment, FIG. 5 is a flowchart of database update information reflection processing startup processing of the embodiment, and FIG.
7 is a diagram showing the configuration of the control table, FIG. 8 is a diagram showing the input data format of the embodiment, and FIG. 9 is a diagram comparing the processing time between the conventional method and the present method. FIG. 10 is a diagram showing the conventional database reflection process, the first drawing is a diagram showing the database reflection process of the present invention, and the first drawing is a diagram showing the database reflection process of the present invention.
FIG. 2 is a diagram showing an example in which a deadlock occurs due to simple blocking, and FIG. 13 is a diagram showing an example in which an exclusive wait occurs due to simple blocking. 5... Database update information creation side, 6... Database update information reflection side, 15, 16.17... Database reflection processing, A. B, C... database, A-1, A-2,... transaction record for database A. Atsushi 1 ■ Figure 2 Figure 3 Figure 4 (2) Figure 5 Herkutsu Otsu Figure 7 Flash Figure g Funeral 9 Figure (α) (b) Figure 10 Flash 5 Otsu

Claims (1)

[Claims] 1. In the process of updating the database using database update records, the records with the same exclusive classification for the database are blocked together, and in the update process, the blocks with the same exclusive classification are serially processed one block at a time. A high-speed database control method that is characterized by controlling the database to