JPH10240753A - Database management system and storage medium used therein - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a database management system capable of high-speed processing of an inquiry using a relation between data. SOLUTION: Storage means for storing related data for adding a value capable of identifying related data to related data and related data, and means for combining related data and related data. Means for retrieving associated data from a database from a value identifying the associated data;
Storage means for storing related destination data and related source data in an appropriate node; analyzing means for analyzing an inquiry including a relationship between data and creating an optimum processing procedure according to a storage state of related data; Means for detecting the storage state of the original data and related data, and information for analyzing and storing information for determining an optimal processing procedure for an inquiry using a relation between given data based on the detected storage state It consists of analysis means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a database processing apparatus, and more particularly to a database management system suitable for managing related data when an object-oriented extension of a relational database management system, and a storage medium used for the database management system.

[0002]

2. Description of the Related Art When managing related data in a database, an object database capable of managing the relationship in a database management system is suitable.
For example, one of the object database standards is an object database standard: ODMG-93 (R.
GGCattel, et al. “THE OBJECT DATABASE STANDARD: OD
MG-93 Release 1.1 ”, Morgan Kaufmann Publishers, In
c., 1994), the system is to provide a REF type for handling associations. The REF type is a pointer that points to an object (data). Suitable for navigation that follows an association and references an object. In addition, in order to speed up the related operation, it is possible to specify a clustering for storing related objects at a position (for example, the same page) where the related objects can be acquired by one disk access, at the time of the storing operation. It is possible to speed up the navigation of individual objects.

On the other hand, in a relational database,
There is a technique for parallelizing processing so that it can handle a huge amount of data. A system having a function of selecting the optimal parallel processing according to the contents of an inquiry and the state of data
6-214843). Currently, there is a move to extend relational databases in an object-oriented manner. SQL3 (Database Language SQL, ISO W
orking Draft, July 1996) has proposed a REF that expresses the relationship between data.

[0004]

The prior art has the following problems. (1) An object database association is a pointer to an object. This is useful for processing that is closely related, but when processing large amounts of data,
Speeding up is difficult because navigation is performed for each object. (2) Clustering in an object database is also ineffective when processing a large amount of data because the possibility of successful clustering is reduced. Furthermore, considering the expansion of a relational database,
It is difficult to store different types of data (from different tables) on the same page.

(3) In a relational database,
The relationship between data cannot be managed directly by the system. In general, a database designer defines a key for each related data and implements a relation between the data by sharing the value. (4) The specification for speeding up the related operation between data of the object-oriented extension of the relational database has not been studied. The technology for selecting the optimal parallel processing according to the contents of the inquiry and the stored related state has not been established. An object of the present invention is to solve the above problems (1) to (4) and to process a query including a related operation at high speed without depending on the number of data to be queried.

[0006]

In order to achieve the above object, the present invention comprises a plurality of nodes for performing database processing, and each of the nodes comprises a database management system connected to another node. When there is a relationship between the related data and the related data, a value that can identify the related data is added to the related data and the related data. Means for combining the data of the relation source and the data of the relation destination by matching the value for identifying the data of the relation added; means for extracting the data of the relation destination from the database from the value for identifying the data of the relation destination; Storage means for storing related destination data and related source data in an appropriate node in accordance with the type of the related data to be queried; Receiving a query that contains an association between data, analyzes the query, the storage state of data with associated, and analyzing means for creating an optimum procedure,
Means for detecting the storage state of the data of the relation destination and the data of the relation source, and analyzing and storing information for determining an optimal processing procedure for an inquiry using a relation between given data based on the detected storage state. Information analysis means.

[0007]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a schematic diagram of a parallel processing of an inquiry including a related operation. A node (10) that analyzes a query or distributes and combines data related to a node (101) that outputs a query result and a node (10) that analyzes and manages storage information of related data
The node (106) for performing the sort / merge of the processing result data with the node (5) are interconnected by a network (100). The node (102) that distributes and combines related data sets the data (11
0) is managed. Data is managed in units called tables.
One table can be divided into a plurality of nodes such as 150 and 170 and stored. In this example, both Table A (150) and Table B (170) are divided into three nodes. The association between data is realized by sharing a reference value managed by the system.

Assuming that the data (160 to 168) in Table A is related to the data (180 to 185) in Table B, reference numeral 151 denotes a column for holding a reference value of a related destination.
This is because the system 171 manages its own reference value. Data in which these two reference values are equal is referred to as related data. In this case, the data in Table A is an association source, and the data in Table B is an association destination. A related operation is performed by performing a combination using the reference value as a key. In order to perform the related operation, it is necessary to select an optimal processing means depending on the storage state of the related data.

[0009] For example, FIG. 7 shows a case where data is distributed and stored in each node, and when the data of the relation source and the relation destination are stored in the same node, an inquiry execution 700 including a relation operation is performed. It is a processing outline. Since the relation source and the relation destination are stored in the same node, the relation operation does not straddle the node. Therefore, the related operation processing 71
0, 711, and 712 are asynchronous processes for each node,
It is possible to execute in parallel. The results processed for each node are collected by a sort / merge 713 of the processing results. The related information and the storage state are stored in the storage information (111) of the node 105 and can be accessed at the time of optimization.

FIG. 2 shows a hardware configuration according to an embodiment of the present invention. A data processing device (211) including a central processing unit and a main storage device and a node (200) as a processing device unit including a data storage device (210) such as a hard disk, or a data input / output device such as a keyboard and a display are added. Nodes (201) are connected by a network (100) such as a LAN.
Each node independently functions as a database system and can exchange data through a network. This configuration is a system in which a plurality of database systems are mutually connected via a network to perform data processing in parallel.

FIG. 3 is a schematic diagram of a database system in which each node can function. The database system stores a plurality of application programs 303 and 304 created by a user, a database management system 300 that manages the entire database system such as processing of data operations and inquiries from the application programs and resource management, and data. It comprises a database 301 and a dictionary 302 for storing database definition information.

The database management system 300 can be divided into a logical processing unit 310, a physical processing unit 320, and a system control unit. The logic processing unit 310 analyzes and executes an inquiry from the application program.
The inquiry passed from the application program is S
It is described in a database query language such as QL. The query analysis unit 311 performs syntax analysis and semantic analysis on the query sentence. The analysis result selects the optimal processing procedure for the inquiry by the static optimization processing 312.

The code generator 313 generates an executable intermediate code according to the selected processing procedure. The dynamic optimization process 314 selects an optimal process according to the state at the time of execution from the process candidates selected by the static optimization process. The code analysis execution 315 executes the generated intermediate code according to the selection of the dynamic optimization processing. Physical processing unit 320
Is composed of an exclusive control 321 for exclusive control of resources shared by the system, a data access process 322 for editing data and determining conditions, and a database buffer control 323 for reading and writing data. The system control unit 330 performs input / output management and the like.

Next, a detailed explanation will be given using an example. The contents to be explained are (1) Means of expressing the relation between data (2) Example data (3) Definition method of related data and management of definition information (4) Storage method of related data (5) Relation (6) Processing method of inquiry including related operation (7) Example of storage of example.

(1) Means for expressing the relationship between data The data of data type A (401) is data type B (402)
Has a reference (400) to the data of type A, the data of type A and the data of type B are said to be related. At this time, the data of type A is an association source, and the data of type B is an association destination. FIG. 4 is a model diagram showing that type A data and type B data are related. The arrow indicates the relation, the tip of the arrow is the relation destination, and the source is the relation source.

FIG. 5 shows a method for managing the related data of FIG. 4 in a database. Table 500 shows type A
Table 510 is a table for managing type B data. Type A has a reference to type B. This is stored in a column 502 for reference to type B. Table 512, which is the destination of type A data, has a column called its own reference value. Related data has the same value in this column. For example, since the reference value B_REF value (503) of the type A data 501 to the type B is equal to the own reference value B_REF value (503) of the type B data 511, the data 501 and the data 511 are related. In this case, the data 501 is an association source, and the data 511 is an association destination. The associated name is a reference to type B. FIG. 6 shows the structure of a reference value representing the association. A value that identifies the data or row 60
1 and a value 600 for identifying the node.

(2) Example Data The example data will be described. FIG. 8 shows an example of related data. Employee type data represents a company employee, and Project type data represents a company project. The relation 820 is a relation name myProject, the relation source is an Employee type, and the relation destination is a Project type. Refers to the project for which an employee is responsible. The relation 830 is a relation name “manager”, the relation source Project type is, and the relation destination is Em.
It is a ployee type. Refers to the manager of a project. The relation 840 is the relation name boss and the relation source is Employee
The type and related destination are also Employee type. Refers to the boss of an employee.

The data 801 is an employee whose name is Emp1 (Employee type data). My project is Data 820 and I have no boss. Data 802 is
The name (name) is an employee with Emp2 (Employee type data). The project in charge is data 812, and the supervisor is data 806. The data 803 has the name Em
Employee of p3 (Employee type data). My project is Data 813, and I have no boss. Data 80
Reference numeral 4 denotes an employee whose name (name) is Emp4 (Employee-type data). The project in charge is data 813, and the supervisor is data 803.

The data 805 is an employee whose name (name) is Emp5 (Employee-type data). The project in charge is data 811, and the supervisor is data 801. Data 806 is an employee whose name (name) is Emp6 (Employee-type data). My project is data 81
2. The boss is the data 807. Data 807 is an employee (Employee type data) whose name (name) is Emp7. The project in charge is data 812, and the supervisor is data 803. The data 808 has the name (Emp8)
(Employee type data). The project in charge is data 813, and the supervisor is data 803.

The data 811 is a project (Project type data) whose name (name) is PJ1 and number (projectNo) is 1945. The administrator is the data 801. Data 8
12, the name (name) is PJ2 and the number (projectNo) is 645
(Project type data). The administrator is data 807. Data 813 is a name
Is PJ3 and the project number (projectNo) is 1192 (Proj
ect type data). The administrator is the data 803.

(3) Definition method of related data and management of definition information FIG. 9 is a data type model diagram of the data of FIG. Employ
ee type 900 is related to char (30) type name and Project type myP
It has a roject 820 and an associated boss to the Employee type. Also,
Project type 910 is char (30) type name and integer type proj
Has related manager to ectNo and Employee type.

The Employee type and the Project type can be defined in the database as shown in FIG. This is SQL
3 is the definition of the named row type. In FIG. 10A,
Type Employee is defined. Indicates the meaning of each line. 1000: The row type name is Employee 1001: There is a field called name of CHAR (30) type 1002: There is a field called boss related to Employee type 1003: There is a field called myProject related to Project type

In FIG. 10B, a type Project is defined. Indicates the meaning of each line. 1010: The row type name is Project 1011: It has a field of name of CHAR (30) type 1012: It has a field of projectNo of INTEGER type 1013: It has a field of manager related to Employee type

The type definition is a description of the nature of the data, and does not manage the substance of the data. In a relational database system, data is managed by tables. In the present embodiment, a method of managing related data using a table will be described. Employee type data is table employees
Manage with. Project type data is managed in the table projects. In the present invention, the table definition method is extended so that the data entity storage method can be specified at the time of table definition.
The extended storage method specification consists of two methods: one is to specify the related destination to be stored in the same node as the related source, and the other is to specify the related source to be stored in the same node as the related destination. You can choose the specification. However, not all data is stored as specified.
If the data is related to multiple data, the specified storage method may not be implemented. Actually, focusing on one association, data to be stored later is stored in the node of the already stored data.

FIG. 11 shows a method of designating a relation destination to be stored in the same node as the relation source. 1101
In 1106, CLUSTERD indicating that the data is to be stored in the same node is specified in the SCOPE FOR clause of the association destination, and the node in which the data of the association destination table exists. FIG. 11A is a table.
This is the definition of employees. Indicates the meaning of each line. 1100: Define a table employees with the row type Employee as a row. 1101 to 1103: Related employees are related to table employees
And stored in the same node. The storage destination of the table employees to be stored in the same node is Node1 or Node
2 or Node3. Here, Node1, Node2, and Node3 are names for identifying node 1, node 2, and node 3, respectively. 1104-1106: Related myProject is related to table proje
Limited to cts data and stored in the same node. The storage destination of the table projects to be stored in the same node is Node1 or N
ode2 or Node3.

FIG. 11B is a definition of the table projects.
Here, the storage method relating to the association is not specified. Indicates the meaning of each line. 1110: A table projects having a row of the row type Project is defined. 1111: The relation destination of the relation manager is limited to the data of the table employees. 1112 to 1113: The storage destination of the table projects is distributed to Node1, Node2, or Node3 by a hash function called HASH1 using the value of the field projectNo as a key.

(4) Method of Storing Related Data FIG. 12 shows a method of designating a relation source to be stored in the same node as a relation destination. At 1212, the table
CLUSTERD BY is specified to indicate that the data is stored in the same node as the related source of employees related myProject. FIG.
(A) is a definition of the table employees. Indicates the meaning of each line. 1200: Define a table employees having a row of row type Employee. 1201 to 1203: Related boss is related to table employees
And stored in the same node. The storage destination of the table employees to be stored in the same node is Node1 or Node
2 or Node3. 1204 to 1206: Related project of related myProject is table proje
Limited to cts data. No storage method is specified for associations.

FIG. 11B is a definition of the table projects.
Indicates the meaning of each line. 1210: Define a table projects having rows of the row type Project. 1211: The relation destination of the relation manager is limited to the data of the table employees. 1212 to 1114: Stored in the same node as the related source of related myProject of table employees. No to store node
de1 or Node2 or Node3. However, if it cannot be stored, the storage destination of the table projects is the field projecte
Node1 with a hash function called HASH1 that uses the value of tNo as a key
Or assign to Node2 or Node3.

FIG. 13 is a flowchart 1300 of the definition registration of related data. The syntax analysis (1301) and the semantic analysis (1302) of the definition information are performed. The analyzed definition information is registered in the dictionary (1303). If there is no designation of the storage method of the related data in the definition information (1304), the process ends (130).
6) Yes. If the storage method of the related data is specified (1304), the consistency of the storage method specification of the related source and the related destination is checked and registered as related information in the dictionary. The consistency check for specifying the storage method is a check as to whether the storage node of the relation source and the storage node of the relation destination can be stored in the same node.

FIG. 14 is an example of a table for registering the definitions of the related data shown in FIG. This table is a table (1400) storing column (field in the row type) information for storing related reference values. Table identifier (1410),
It is composed of column identifier (1411), column data type identifier (association is managed as association type), association data type identifier, scope designation destination table identifier, storage method designation, and other information about columns (1412). I do.

In the table definition of FIG. 11, the related manager (14
20) has no storage method specification (1416), but the related boss
And the storage method specification of the related myProject is "same node". From this table, it is understood that the related boss of the table employees is stored in the same node as the data of the related table employees, and that the related myProject of the table employees is the related table project
It can be seen that the data is stored in the same node as the data of ts.

FIG. 15 is a table (1500) for registering related information when the storage method is designated in FIG.
Origin table identifier (1510), association identifier (151)
1), an associated destination table identifier (1512), an associated destination data type identifier (1513), a storage node (1514), and a storage state (1515). 1520-152
2 is related information of the related boss of the table employees, which is expanded to the nodes (Node1, Node2, and Node3) to be stored.
Also, 1523 to 1525 are related myProje of the table employees.
Related information of ct, also the node to be stored (Node1 and
Node2 and Node3).

At the time when the table is defined, since the data does not exist, the storage state is initialized to 0. In the present embodiment, the storage state is represented by the ratio (percentage) of the number of related destination data stored in the same node to the total number of related source data of a certain node.

FIG. 16 shows an example of a storage operation for storing related data in the same node. FIG. 16A shows an example of a storage operation of the relation destination of the relation manager of the table projects without specifying the storage method. Because there is no storage method specification, the table employee
In order to store the ees data in the same node as the related source data of the related manager, it is necessary to explicitly specify the storage node. Update the reference value of the related manager of the related source after storing the data of the table employees first. Indicates the meaning of each line. 1600 to 1601: Employee type data ("Emp3", NULL, NULL) is stored in the table employees of Node1. 1610-1613: Obtain data in which the column name is "Emp3" in the data of the table employees, and set it as the related destination in the related manager of the data p in the table projects. p is the data already acquired.

FIG. 16 (b) shows an operation of storing the relation source in the same node as the relation destination data in which the relation destination data has already been stored. Indicates the meaning of each line. 1620 to 1621: Emp whose name set in the related myProject is "Emp3" with the data p of the table projects as the related destination
Store loyee type data in the table employees. p is the data already acquired. & p represents the reference value of the data. The storage node depends on the storage node of p.

FIG. 16 (c) shows an operation of storing the relation destination in the same node as the relation source data in which the relation source data has already been stored. Indicates the meaning of each line. 1620 to 1621: Project type data ("PJ2", 645, NULL) is stored in the table projects. Storage node is table employee
Depends on the storage node of ees data e. e has already been obtained. The stored data is set as the related destination of the related myProject of the data e in the table employees.

FIG. 17 is a flowchart of the storage operation process of FIGS. 16 (b) and 16 (c). FIG. 17A is a processing flowchart (1700) of the data storage processing up to code generation. Syntax analysis of storage instruction (1701)
And a semantic analysis (1702). As a result of the analysis, it becomes clear which table and which relation is to be processed, so it is checked whether or not there is a storage method designation of related data based on the column information on the relation (step 1703). If the storage method is not specified, the storage node is determined from information other than the related information, and an intermediate code is generated (1705). If the storage method is specified, the storage node specification method is determined from the related information (1704), and an intermediate code is generated (170).
5).

FIG. 17B is a flowchart of a code execution process in the data storage process. The intermediate code designated in (a) is analyzed (1721), and if there is no designation of a storage method of related data, data to be stored is generated (1724). If the storage method is specified, node information is acquired from the related data (related destination or related source data already stored), and the storage position is determined (17).
23), and generate data (1724). The generated data is stored in the designated (determined) storage location (172).
5). If the stored data is related data (17)
26), the related reference value of the related source is updated (172).
7).

(5) Storage state detection of related data FIG. 18 shows the data storage state of related data, for example,
FIG. 14 is a schematic diagram of a process of analyzing whether related data is stored in the same node. A database state analysis utility (18) disconnected from the network 100
00) scans the data of each node and checks the storage state. The result is set in the storage information in the dictionary.

FIG. 19 shows a method (190) in which the database state analysis utility detects the storage state of each node.
It is a flowchart of 0). The related information specifying the storage method of the corresponding node is acquired (1901), and all the acquired related information is acquired (1902) (1912).
The data of all the related sources are acquired, and the data of each related source are (1903), (1905), and (191).
0), the related reference value is acquired (1907), and it is checked whether the data of the related destination exists in the same node based on the reference value (1908). If they are in the same node, they are counted (1904) (1909). At the same time, the total number of related source data is also counted (1904) (1906), and for each piece of related information, the ratio of the number of related destination data stored in the same node to the number of related source data is calculated, and the storage state of related information ( 1515).

(6) Processing Method of Inquiry Involving Related Operation FIG. 20 shows an example of an inquiry including a related operation. Indicates the meaning of each line. 2000-2002: myProject such that the projectNo of Project pointed to by myProject of Employee is 645
Get the name of Project and the name of Employee. -> Is a related operation, in this case, a reference.

FIG. 21 is a table (2100) of search results for the inquiry of FIG. FIG. 22 shows the result of converting the query statement of FIG. 20 into a query of an outer join that does not use a related operation. Each row will be described. 2200: Reference related operation myProject-> name to projects.
Converted to name. Converted to employees.name for clarity. 2201: Add related table project to table list. 2202 to 2203: Insert a left outer join using the related reference value as a key. 2204: Reference related operation myProject-> projectNo proj
Converted to ects.projetsNo.

FIG. 23 is a flowchart of an inquiry process including a related operation. Particularly, it is a flowchart (2300) of a process up to code generation in the inquiry process. The inquiry sentence is analyzed (311), and if there is no related operation, processing equivalent to that of FIG. 3 is performed. If there is a related operation (230
1) Equivalently convert the query statement into a query statement using an outer join without related operations (2302). Static optimization of related operations is performed from the converted query statement and related information (including the storage state). After that, the process is the same as the process in FIG.

FIG. 24 is a flowchart of a process for converting the query sentence including the related operation shown in FIG. 23 into a query sentence using an outer join having no related operation (2302). A table storing related data is added to the table list (2400), and an outer join condition using a reference value as a key is set (2401). Converts a data reference to a column reference in the referenced table.

FIG. 25 is a flowchart of the static optimization of the related operation in FIG. 23 (2303). One measure of static optimization is the storage of related information. First,
The storage information is acquired (2500). A strategy is selected from, for example, the following three states depending on the storage state of the data related to the operation target (2501). (A) All sets of related source and related data are stored in the same node. (B) With some exceptions, a set of related source / destination data is stored in the same node. (C) Stored regardless of the storage method specification.

(A) is an ideal state. Select a strategy for performing outer join processing asynchronously for each node (25
02). In (b), a strategy to be processed is selected by distinguishing exceptional entities from related data stored in the same node (2503). Specifically, an inquiry as shown in FIG. 26 is executed (2600). The reference values of the relation source and the relation destination are compared (2601), and if there is a relation destination (2602), the combining process (2603) is performed. If it does not exist (exceptional existence), the related source data is transferred to the work node (2604). For the transferred data, the associated destination data is acquired from the node where the associated destination is stored, and the transfer destination node performs the joining process. (C)
In the case of, the optimization is not performed using the related information. Optimization is performed using other information.

Here, it should be noted that:
This means that the storage state referred to when the code was generated may be different from the storage state at the time of executing the code.
Even if the storage state strategy of (a) is selected at the time of code generation, the storage state of (b) or (c) may be present when the code is actually executed. Therefore, at the time of code generation, a processing method is selected according to the storage state referred to at that time, but it is necessary to leave other processing methods as candidates as necessary. When generating code (a)
, It is necessary to leave at least the processing method corresponding to (b) as a candidate. Then, when executing the code, an appropriate one is selected from the candidate processing methods according to the data.

(7) Example of Storage of Example FIG. 27 shows an example in which the data of FIG. 8 is stored in the definition of FIG. In the table “projects”, reference numeral 2711 denotes its own reference value column, reference numeral 2712 denotes a name column,
13 is a projectNo column, and 2714 is a column of a related manager. In the table employees, 2741 is its own reference value column, 2742 is the name column, 274
3 is a column of the related boss, and 2744 is a column of the related myProject. The reference value column associated with the own reference value column stores the reference value assigned by the system. In this example, the reference value is represented by three characters, the first character is the identifier of the table to which the data belongs (e is the table employees, p is the table projects), and the middle character is the identifier of the node (1 Is the node 1, node 2 is the node 2, node 3 is the node 3), and the last character is the data (row) identifier in the table.

The storage example of FIG. 27 is in the storage state of FIG. A description will be given using the employees table of the node 1 as an example. Focus on related boss columns. Reference values are e17, e23,
e15, the data of the nodes 802 and 80
The boss of No. 6 is stored in node 1, but data 807
Indicates that it is stored in node 2. Therefore, the storage state is 66 (percent) because two of the three data are stored in the same node. On the other hand, focusing on the column of the related myProject, all the reference values are p11, indicating that all the related destinations are stored in the same node. Therefore, in the inquiry of FIG. 20, since all the related destinations of the related myProject are stored in the same node, the possible strategy for node 1 is 2502 in FIG. When the related destination is NULL, it indicates that there is no related destination. When the storage state is detected, it is treated as being stored in the same node.

[0050]

According to the present invention, for data to be stored in a distributed manner, there is provided a means for storing a set of related data (related) in the same node. Therefore, it is possible to perform parallel processing. Further, even if the data is not stored in the same node, a means for detecting the storage state of related data is provided, so that it is possible to select an optimal parallel processing. Therefore,
Inquiries including related operations can be speeded up.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of parallel processing of an inquiry including a related operation.

FIG. 2 is a hardware configuration diagram.

FIG. 3 is a configuration diagram of a database system.

FIG. 4 is a diagram illustrating an example of an associated data type.

FIG. 5 is a diagram showing an outline of a method of expressing related data.

FIG. 6 is a diagram showing a structure of a reference value indicating an association.

FIG. 7 is a flowchart of parallel processing of related operations.

FIG. 8 is a diagram showing an example of related data.

FIG. 9 is a data type model diagram of the data of FIG. 8;

FIG. 10 is a diagram showing an example of a data type definition in FIG. 9;

11 is a diagram showing a definition example 1 of a table using the data type of FIG. 10;

12 is a diagram illustrating a definition example 2 of a table using the data type of FIG. 10;

FIG. 13 is a flowchart of definition registration of related data.

FIG. 14 is a storage structure diagram of column information for storing related reference values.

FIG. 15 is a storage structure diagram of related information.

FIG. 16 is a diagram illustrating an example of a storage operation of data related to the tables defined in FIGS. 11 and 12;

FIG. 17 is a flowchart of a process of storing related data.

FIG. 18 is a schematic diagram of a database state analysis process.

FIG. 19 is a flowchart of a method for detecting a storage state of related data.

FIG. 20 is a diagram illustrating an example of an inquiry including a related operation.

FIG. 21 is a diagram showing a result of the inquiry in FIG. 20;

FIG. 22 is a diagram showing the query of FIG. 20 converted into an outer join.

FIG. 23 is a flowchart of an inquiry process including a related operation.

FIG. 24 is a flowchart of a process of converting an inquiry sentence including a related operation into an outer join.

FIG. 25 is a flowchart of static optimization of a related operation.

FIG. 26 is a flowchart of a related operation execution process.

FIG. 27 is a diagram illustrating an example of distributed storage of the data in FIG. 8;

28 is a diagram showing the storage state of FIG. 27.

[Explanation of symbols]

100: network, 102: node for operating data, 105: stored information analysis / management node, 110: external storage device for storing data or data, 311: query analysis unit, 312: static optimization processing, 313 ... Code generation, 314: Dynamic optimization processing, 315: Code analysis execution

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification code FIG06F 15/40 380D (72) Inventor Morihiro Iwata 890 Kashimada, Saiwai-ku, Kawasaki-shi, Kanagawa Pref. 72) Inventor Shunichi Torii 890 Kashimada, Saiwai-ku, Kawasaki City, Kanagawa Prefecture Inside the Information and Communications Development Division, Hitachi, Ltd. (72) Inventor Yoshito Kamejo 5030 Totsukacho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Software Development, Hitachi, Ltd. In headquarters

Claims (9)

[Claims] A plurality of nodes for performing database processing, each of which is connected to another node in a database management system. Means for accepting the definition of the association and the storage method of the associated data and managing the associated data together with the storage state; holding a value for identifying the associated data as a value for managing the association with the associated data; Means for storing related data, which adds the same value to the associated data as a value for identifying itself, and a value for identifying the associated data added to the original data and the associated data A means for combining the data of the related source and the data of the related destination by comparing the The most suitable processing procedure is selected according to the analysis result of the above, the definition of the relation between the data and the method of specifying the storage of the related data, and the storage state of the node storing the data of the relation destination and the data of the relation source to be operated. A database management system comprising: a determination unit; and an execution unit that drives a unit that combines the data of the relation source and the data of the relation destination with an optimum node in accordance with a selected processing procedure. 2. The database management system according to claim 1, wherein the means for receiving and managing the definition of the relationship between the data and the storage method of the related data includes: Means for specifying the storage method to be stored in the database together with the relation definition, means for registering the specified relation definition and the storage method in the dictionary of the database system, and means for managing the storage state of data related to the storage method And a database management system. 3. The database management system according to claim 1, wherein the value for identifying the data of the association destination holds a value for identifying a node at which the association destination is stored. 4. The database management system according to claim 1, wherein the means for combining the data of the relation source and the data of the relation destination includes a query sentence including a relation operation, and data of the relation source and relation destination being managed. A database management system for converting a query into a query that does not include a related operation such as joining tables. 5. The database management system according to claim 1, wherein the deciding means for selecting the optimum processing procedure is such that a set of related destination data and related source data to be queried is stored in the same node. A database management system characterized in that when each group is stored in a distributed manner at each node, a joining process is selected asynchronously at each node to combine the data of the related destination and the data of the related source. 6. The execution means according to claim 1, wherein the node that stores the related destination is identified by a value that identifies the related destination data according to claim 3 that indicates the related destination to be processed. A database management system that performs a combining process at a node that stores related data or a working node that is not the same node. 7. A database management system comprising: means for storing related destination data and related source data in an appropriate node based on the stored related information and the storage method according to claim 1. 8. A database system including the database management system according to claim 1, wherein, for data having a relation stored in the database, means for detecting a storage state of data of a relation destination and data of a relation source, A database system comprising: means for evaluating the storage state for a method of storing related data; and means for storing the evaluated result together with the corresponding storage method. 9. The program of the database management system according to claim 1, information on a definition of a relationship between data and a designation method of a storage method of related data, and information on a relation destination and a relation source of related data. A storage medium storing a database storing a value for identifying an associated destination according to claim 3.