JPS63118931A - Data base managing system - Google Patents

Abstract

PURPOSE:To improve the performance of the parallel processing of a task which accesses to a data base, by storing slave records in every master record unit separately in different files, and managing relation between the master record and the slave record based on file identification information. CONSTITUTION:When the slave record (a) relating to the master record A is added, a relative managing memory 8 is retrieved based on the key value of the master record A, and the file identification information of the slave record 5 on which the slave record (a) is to be added, is taken out, and the slave record (a) a added on a vacant record in the said file base on the above information. Also, when the slave record (a) relating to the master record A is deleted, after the slave record 5 in which the slave record exists, is decided, the slave record (a) to be deleted is retrieved and deleted. When the slave record (a) relating to the master record A is retrieved, the slave record 5 in which the slave record exists is decided, and the slave record (a) is deleted from the said file.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a database management system, and particularly to a database management method and apparatus suitable for processing data having a hierarchical structure.

[Conventional technology]

As stated in the "Method for realizing a database with a hierarchical structure" section of Principles of Document Database Systems (Japan Computer Association), parent-child relationships of records have traditionally been managed using the method shown in Figure 2. It has been known. In other words, there is a method of connecting and managing child records a, b, and c related to parent record A in Figure 2-(1) in a circular chain indicated by pointer 3, or a method of managing child records a, b, and c related to parent record A as shown in Figure 2-(2). This is a method of managing using multiple pointers 3'.

[Problem that the invention seeks to solve]

The above conventional technology has the following problems.

FIG. 3 is an example of a database for managing inventory using the conventional technology. The inventory management database consists of a warehouse file and a parts file, where the warehouse name is set as data in the warehouse file, and the part number and inventory quantity are set in the parts file.

Management is performed by connecting the warehouse and the parts stored in the warehouse with pointers. The above database is also used for tasks such as transporting parts to the warehouse.

It is managed by task Y, which carries out parts removal from the warehouse. Now, when the operator instructs from terminal 1 to carry part #10 into warehouse A, task X is activated and data is written to the parts file. When Task , deletion is performed, but if X and Y operate at the same time, inconsistencies will occur in pointer reconnection and data operations. Therefore, when task X adds or updates the data of part #10, it is necessary to lock the part file. As a result, task Y waits until task X finishes adding and updating the data of part #10, and no data inconsistency occurs. When task Y is activated by an instruction to export part #30 and updates or deletes the part data of part #30, task A may change the part file by task A, even though task Y has no effect on task A. Because it is locked, waiting occurs and parallel processing performance decreases.

Furthermore, as the number of types of parts to be delivered to warehouse A increases, it takes more time to follow the pointer and find the parts for which import/export instructions have been given, and the processing speed of tasks X and Y decreases. On the other hand, if the method shown in Figure 2-(2) is used to reduce the time required to follow the pointer, the time required to follow the pointer will be reduced, but the number of types of parts that can be delivered to one warehouse will need to be limited. Become.

Additionally, if the computer stops processing due to a power outage while Task It may become impossible to do so. Furthermore, when loading and unloading of components into a component file occurs frequently, pointers are also frequently reconnected, which increases the probability of pointer breakage.

As mentioned above, in the conventional method, (1) task parallelism decreases because the occupied range of the file cannot be localized, and (2) processing speed decreases due to increased overhead when searching for pointers. (3) A local failure in the database spread to all records, resulting in a significant decline in reliability.

[Means for solving problems]

In order to solve the above problems, in the present invention, (1)
, child records related to a parent record are stored in separate files for each parent record, and the relationship between the parent record and the child record is managed using file identification information stored after dividing the child record.

(2) Task conflict management is performed in units of files that are divided and stored using the method of (1).

(3) When a failure occurs, failure recovery is performed in units of files that have been divided and stored using the method of (1).

First, the principle of the present invention, which is characterized by the following, will be explained with reference to FIG.

FIG. 1 is an explanatory diagram when the relationship between parent records and child records is managed using file identification information. When adding a child record a related to a parent record A, the association management memory 8 is searched based on the key value of the parent record A, and the file identification information of the child file 5 to which the child record a is to be added is retrieved. Based on this, a child record a is added to the empty record of the file. As shown in FIG. 4, the related management memory 8 is composed of key values of parent records and child file identification information, and when a parent record is added, a corresponding child file is generated and file identification information is added. rear,
These pieces of information are set in the related management memory 8. Additionally, when the parent record is deleted, this information is also deleted. Furthermore, when using a natural number corresponding to 1=1 as the key value (for example, parent record storage record number) for the parent record, use the management information case number instead of the parent record key value on the related management information. is also possible.

When deleting a child record a related to a parent record A, the child file 5 in which the child record exists is determined in the same way as in the case of addition, and then the child record a to be deleted is searched from the file and deleted. When searching for child record a related to parent record A, it is possible to determine the child file 5 in which the child record exists in the same way as in the case of addition, and search for child record a from this file.

[Effect]

In the present invention, (1) Child records can be occupied and released in divided file units, and tasks accessing from different parent records can be performed simultaneously.

(2) Since there is no pointer, even if the number of child records related to one parent record becomes large, the impact on the overhead when searching for child records is reduced.6 (3) The file in which child records exist is Since the child records are composed only of child records having the same parent record, the child records are arranged in physically adjacent storages, and the number of times the disk 170 is searched is reduced.

(4) Failures are less likely to occur because dynamically changing related information such as pointers is not included.

(5) Even if a failure occurs, failure recovery can be performed in units of divided files.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 5 shows an example of an inventory management database implemented according to the present invention, with the parent file 1 in FIG. 1 as a warehouse file and the child files 5, 6 and 7 as parts files. Now, according to the instructions to bring part #10 to warehouse JiA,
While task It will be locked.

However, in the present invention, since parts files are divided and managed in parent record units, locking can be localized only to parts file 11, and export task Y is related to parent records other than warehouse A that import task X is accessing. Parts file 12
．． 13 can be accessed simultaneously with import task X.

Additionally, if computer processing is stopped due to a power cut, etc. while accessing a part file, and a failure occurs in the part file, the entire part file shown in Figure 3 becomes inaccessible, and the part file that was evacuated before the failure occurs. It was necessary to restore the entire part file by restoring it to file data or initializing the file and then registering the data again. On the other hand, in the case of the present invention, as shown in FIG. 6, when a failure occurs in the component file 6, old data is loaded from the backup file 10 only for the component file 6 where the failure has occurred, or only for the file 6. By registering the data again after initializing , the fault can be localized and the parts files 5 and 7 other than those in which the fault has occurred can be localized.
is accessible, so the impact on the system can be kept to a minimum. Furthermore, since the amount of data to be recovered is small, the time required for failure recovery can be shortened.

〔Effect of the invention〕

According to the present invention, in a database having a hierarchical structure, (1) file conflict management can be performed in detail, so that the parallel processing performance of tasks accessing the database can be improved.

(2) When a failure occurs, failure recovery can be performed in units of files into which child records are divided, so maintainability can be improved.

(3) Since there are no pointers, highly reliable database management is possible.

(4) Since there is no pointer, there is less impact on overhead due to an increase in child records, and the number of I10s can be reduced, resulting in improved processing speed.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of the configuration of an embodiment of the present invention, Fig. 2 is an explanatory diagram of a case where a database with a hierarchical structure is realized by the conventional technology, Fig. 3 is an explanatory diagram showing the conventional technology, and Fig. 4 is an explanatory diagram of the conventional technology. An explanatory diagram of related management memory. FIG. 5 is an explanatory diagram of an embodiment according to the present invention, and FIG. 6 is an explanatory diagram of failure recovery. 1...Parent file, 2...Child file, A, B, C
... 1 Figure 6--1 Darkness! Metso extension figure 3 (+, figure height 5 figure

Claims (1)

[Claims] 1. In a database management system that manages the relationship between parent records and child records and uniquely identifies a parent record, it is possible to automatically search for a group of child records related to the parent record. When storing a record group by dividing it into multiple files, storing the relationship between parent records and child records using the storage file identification information of the divided child records, and adding child records to the parent record. uniquely determines the file that stores the child record from the storage file identification information of the divided child record related to the parent record, adds the child record to the file, and deletes the child record from the parent record. When doing this, after uniquely determining the file in which the child record is stored based on the storage file identification information of the divided child record related to the parent record, the file is searched and the corresponding child record is searched. When deleting a parent record and searching for a child record, the file in which the child record is stored is uniquely determined from the storage file identification information of the divided child record related to the parent record, and then A database management method characterized by searching within the file to find a corresponding child record.