JPH0269871A - Inquiry processing system for distributed data base system - Google Patents

Abstract

PURPOSE:To speed up inquiry processing by controlling the conversion of data according to information to show need or needlessness in the case where data base inquiring data is edited by a data converting means when data base inquiry is inputted. CONSTITUTION:The information to show the need or the needlessness of the data conversion in each level is set by using a mapping level adding means. Namely, a data conversion level in each level is set by using the mapping level adding means, and when the data base inquiry is inputted, at the time of the editing of the data base inquiring data, the data converting means converts the data according to a mapping level, and transfers it to an origin of a request. Thus, needless data conversion can be excluded, and the speedup of the inquiry processing can be attained.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a query processing method for a distributed database system, and in particular to a query processing method for a 9-distributed database system suitable for speeding up data conversion between multi-layered schemas in query processing. This is related to the processing method.

[Conventional technology]

Conventionally, in a distributed database system query,
To achieve distributed transparency to users, a data model is layered that hides the location and structure of the distributed database.

For example, in ``Configuration of a Distributed Database Management System (DEIMS-3)'' (Database System Study Group Materials 37-1, 1983) by Meiki et al., in order to achieve this, the conventional global schema and global external schema were It is shown to be set.

[Problem to be solved by the invention]

However, as described above, when schemas are multi-layered, the number of mapping stages between schemas increases in database queries, leading to an increase in query processing time overhead. There is no particular description of a method for speeding up mapping between schema that have been created.

The present invention has been made in view of the above circumstances, and its purpose is to solve the above-mentioned problems in the query processing method of conventional distributed database systems, and to reduce the overhead of database query conversion when schemas are multi-layered. An object of the present invention is to provide a query processing method for a distributed database system that can reduce the number of queries and speed up query processing.

[Means to solve the problem]

The above-mentioned object of the present invention is composed of a plurality of nodes having a processor and a database, and a communication network that interconnects the plurality of nodes, and each node processes a process of accessing the database of the plurality of nodes online. In a distributed database system configured as shown in FIG. A means for adding a mapping level indicating the need for conversion of query data corresponding to the level between the multi-layered schemas, and a data conversion means for converting data based on the mapping level are provided. Using the ping level addition means, set information indicating whether data conversion is necessary or unnecessary at each level, and when a database query is input, the data conversion means will add the above information when editing the database query data. A query processing method for a distributed database system characterized by controlling data conversion based on the above, or a directory creation means for creating a table that integrates the multi-layered schemas, and a distributed database system based on the table. A query processing means for making a database query is provided, the directory creation means is used to convert a multi-layered schema into a single unified table, and when a database query is input, the distributed database query processing means This is achieved by a query processing method of a distributed database system characterized by performing direct mapping between schemas based on the table.

[Effect]

In the query processing method of the first distributed database system according to the present invention, the data conversion level at each level is set using the above-mentioned mapping level addition means,
When a database query is input, when editing the database query data, the data conversion means converts the data based on the above mapping level and transfers it to the request source, eliminating unnecessary data conversion. This makes it possible to speed up query processing.

In addition, in the query processing method of the second distributed database system according to the present invention, a multi-layered schema is converted into a single unified table using the above-mentioned directory creation means, and a database query is input. Then, since the distributed database query processing means performs direct mapping between schemas based on the above table, the number of mapping processing stages can be reduced, thereby speeding up the query processing. It is possible.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 2 is an overall configuration diagram of a database system (hereinafter referred to as "rDB system") showing a first embodiment of the present invention. In the figure, 2a and 2b are no F, 3a and 3b
is a user terminal, 4a, 4b are data dictionaries, 5a, 5
b indicates a database, and 1 indicates a communication path connecting nodes 2a and r2b.

Note that 6a and 6b in nodes 2a and 2b are distributed database functions (hereinafter referred to as "distributed 08 functions"), 30
a, 30b are mapping management tables, 7 a, 7 b are UA
P (user program) is shown.

An overview of the database query processing procedure in the database system of this embodiment is as follows.

When a database inquiry is issued from the user terminal 3a or the UAP 7a, the distributed 08 function 6a receives the inquiry request and defines it in the data dictionary 4a. Interpreting and executing the query using the database structure related to the query request. As a result of the interpretation, if the database corresponding to the query exists in the node 2a, the data is retrieved from the database 5a, converted according to the database structure, and transferred to the UAP 7a.

On the other hand, if the result of the interpretation is that the database corresponding to one query does not exist in the node concerned but exists in another node 2b, the query request is transferred to the distributed 08 function 6b of the node 2b via the communication path 1.

The distributed 08 function 6b sends the inquiry request to the data dictionary i4b.
is defined in . The query is interpreted and executed using the database structure related to the query request, and data is retrieved from the database 5b and transferred to the node 2a via the communication path 1.

The distributed 08 function 6a edits the transferred data according to the previous database structure and transfers it to the UAP 7a, thereby completing one database inquiry process. A similar process will follow when the query is for database update.

FIG. 3 shows the schema structure of the data model in this embodiment. The local schema 13 shows the structure of the database existing in each node. The distribution schema 12 shows the arrangement status and arrangement conditions of the database of each node, and the global schema 711 shows the overall database structure for the user. Subschema 10 is
Shows the database structure for individual users. The database query shown above is executed by interpreting the schema structure defined for each database. This schema configuration improves distribution transparency and data independence for UAPs.

FIG. 4 shows an example of the database structure in this embodiment. In the figure, 20 is the node 2a, 2b
21 is a global database structure that transcends nodes, and 22 is a record type e+ indicating the set type of data set types with the same attribute.
'oρ1oyee, 23 is record type career
, 24 indicate the record type family. 2
5 is a set type emp-c indicating the set type of the access path between the employee 22 and the career 23 mentioned above.
ar, 26 is the above employee 22 and family
24 shows a Sera 1-type emp-fam.

The global database 21 is the node 2a.

They can freely access each other from 2b. On the other hand, the local database 20 can be accessed in a closed manner to each node 2a, 2b.

Figure 5 shows an example in which the database in Figure 4 is defined according to the schema configuration in Figure 3, where 10a is the subschema 5ubl, and lla is the global schema gd.
b, 12a indicates a distributed schema, and 13a indicates a local schema local-db. global schema l
In la, the record type emp that constitutes the schema
loyee, career, above set type em
Define component types that make up p-car and each record type. The component type is
Attributes of each component type - data type, precision,
Define length etc.-1.

This determines the attributes of each component and record - one type, one length and one position. subschema 10
In a, a structure that uses the global schema lla from a specific UAP is defined. In this example, employee
e-record component key.

It defines that the code components of the name and career records are used through the set type emp-car. Furthermore, the local schema 13a defines the structure of the local database shown in FIG. The distributed schema 12a is the global schema 11
The mapping between a and the local schema 13a is defined. Specifically, gdbla record emp1o
yce and career are 1ocal-db13
Corresponding to records employee and Career of a, components key, br of gdbll a
anch, nu+nber, code is 1oca
It corresponds to the component with the same name of l-db13a.

Furthermore, gdbLla is mapped to local-db13a of the two nodes according to the value taken by the branch component of the employee record. That is, if the value of branch is less than 100, it is Tokyo (node 2a).

If it is 100 or more, it indicates that it exists in osaka (node 2b). With the above schema definition, U
It is possible to expand database queries into local database queries via subschemas issued by the AP.

In this schema configuration, when performing a database query, data conversion is required because the record configuration, component configuration, and component attributes are different between the global schema 11 and the local schema 13. The present invention sets the necessity of data conversion in advance for each level of schema, record, and component, and
By referring to this information when executing a query, unnecessary data conversion is eliminated.

FIG. 6 is a diagram showing details of the mapping management table 30a or 30b. In the figure, 31 is a schema table, 32
indicates a record table, 33 a component table, and 34 a set table. Each table is associated with the arrows shown in the figure. In addition, each table has mapping information between corresponding schemas as well as a mapping level match flag indicating the necessity of data conversion at each level.

The operation of this embodiment configured as described above will be explained in more detail below.

Figures 1 (a) to (C) are flowcharts showing the procedure for setting the mapping level by inputting the global schema and local schema, and (cl) is the data conversion procedure using the set mapping level information. It is a flowchart which shows. Note that mapping levels between other schemas can also be set using a similar procedure. A detailed explanation will be given below.

FIG. 1(a) shows the mapping level setting process. In the figure, the schema table initialization step 41 secures and initializes the schema table 31.

The record retrieval step 42 retrieves record types in the global schema one by one and searches for a record type in the local schema that has the same name as the record type. The result is determined in normality determination processing step 43.6 If the corresponding record type exists, it is determined to be normal, and if it does not exist, it is determined to be an error, and the mapping level setting process is terminated.

If it is normal, call the record level setting process 50,
This process is executed for all records in the global schema. At completion determination step 44, all records match determination (
Perform step 45). Only when all record types defined in the global schema exist in the local schema and all record level matches are set, all records are considered to match, and schema table 31
Then, schema level-number setting processing (step 46) is performed.

FIG. 1(b) shows details of the record level setting process 50 described above. Record table initialization step 51
secures and initializes the record table 32. The component retrieval step 52 sequentially retrieves components constituting the record type of the global schema selected in the record retrieval step 42 described above, and searches for components with the same name among the components constituting the local record type. Result is,
The determination is made by the normality determination process (step 53). If the corresponding component type exists, it is considered normal; if it does not exist, it is considered an error, and the process ends.

If it is normal, the component level setting process 60 is called and the process is executed for all components of the record in the global schema.

- At the completion determination step 54, it is determined whether all component processing is completed, and if the processing is completed, all component matching determination is performed (step 55). Only when all components 1 of the record type in the global schema exist in the record type in the local schema, and all components 1 to level match are set, all components 1 to 1 are considered to match. , record table 32
Then, record level-number setting processing (step 56) is performed.

FIG. 1(c) shows details of the component level setting process 60 described above. The component table initialization step 61 secures and initializes the component table 33. Attribute check step 62 refers to component attributes, sets data conversion rules, and checks the degree of matching of attributes between components. The attributes in this case include group/array/basic attributes, decimal/binary/character type attributes, integer/real number attributes, precision/length attributes, etc. In the attribute match determination step 63, the result of the attribute check is determined, and if the attributes match, component level-number setting processing (step 64) is performed in the component table 33.

In the case of the schema definition example shown earlier (Figure 5), the global schema lla matches the schema, and the record emplo
yee is record matching, component key.

branch, number, name, cod
e is set to component-match.

FIG. 1(d) shows a data conversion procedure using the mapping level information set in the above process. When data conversion is instructed by a query request or the like, the schema table retrieval step 71 retrieves the schema table 31 corresponding to the query and determines whether schema level matching is set (step 72). If it is set, the record unit assembly process (step 7) of the record is performed.
Do 5). If there is no match at the schema level, the record table 32 corresponding to the query is taken out in the record table retrieval step 73, and it is determined whether record level matching is set (step 74). If it has been set, the record unit assembling process (step 75) for the record is performed.

The record unit assembling step 75 performs processing such as moving or copying records all at once using only the overall length of the records without referring to individual component attributes. If there is no match at the record level,
The component table retrieval 76 sequentially retrieves the component table 33 of the record and determines whether component level matching is set (step 77
). If it has been set, a component-by-component assembly step 78 moves or copies components at once without undergoing attribute conversion processing. If not set, the attributes between the components are different, and the attribute conversion assembly (step 79) is called to process the data conversion procedure according to the attributes of both components.

The above component-based assembly process 78 is executed for all components (step 80). By using the data conversion procedure shown above, it is possible to realize data conversion without unnecessary attribute conversion.

In this embodiment, although the timing of creating the mapping management table is not specified, it is clear that the mapping management table should be created at least before the first database query using the schema occurs. Furthermore, from the viewpoint of performance, it is also easily possible to store the mapping management table in the data dictionary.

According to the query processing method of the distributed database system shown in this embodiment, the data conversion level at each level is set using the mapping level addition means, and when a database query is input, when editing the database query data, Since the data conversion means converts the data based on the mapping level and transfers it to the request source, unnecessary data conversion can be eliminated, and as a result, query processing can be speeded up. There is an effect.

Next, other embodiments of the present invention will be shown. In the embodiment shown above, in the schema structure shown in FIG. 5, when performing a database query, the record structure, component structure, and component attributes are different between the global schema lla and the local schema 13a. Set information indicating the need for necessary data transformation at each schema, record, and component level.
By referring to this information when executing a query, unnecessary data conversion is eliminated. On the other hand, in the embodiment shown below, when executing a database query in the schema configuration shown in FIG. By creating a table that integrates the global and local schemas as a directory in advance, the schema conversion required due to differences in record structure, component phone structure, and set structure can be performed at the time of executing one query. By referring to this table, unnecessary schema conversion is removed.

FIG. 7 shows an example of a mapping management table corresponding to a directory. In the figure, 30a is the entire mapping management table, 31a is a management table that summarizes the following individual tables, and 32a is a management table for the global schema 11. A global record table that holds information about global records, 3
3a is a global component table that holds information about global components; 34a is a global set table that holds information about global sets; 35
a is a distributed record table that holds information about distributed records of the distributed schema 12, 36a is a distributed component table that holds information about distributed components, 37a is a distributed set table that holds information about distributed sets, and 38a is a related node. Node management table to manage, 39a
is a local schema table that holds information about the local schema 13; 40a is a local record table that holds information about local records; 41a is a local component table that holds information about the raw component 1-; and 42a is a table that holds information about local sets. shows the local set table.

In the mapping management table 30a, the global schema 11 and the distributed schema 12 have a 1:1 correspondence, and the distributed schema 12 and the local schema 13 have a 1:n correspondence. Each table is directly linked by pointers. The global record table 32a and the distributed record table 35a are
Directly coupled 1:1. Therefore, once an arbitrary global record is determined, distributed records for it can be directly determined. The same applies to other global-dispersion relationships. The distributed record table 35a and the local record table 40a are directly connected to the circle.

Local records corresponding to a query are determined by evaluating the distribution conditions during execution. Even in this case, once the evaluation is determined, the local node and local record can be determined directly. The same holds true for other distributed-local relationships.

FIG. 8 is a flowchart showing an example of the procedure for creating the mapping management table 30a shown in FIG. In the figure, the global schema relationship table creation step 81 is performed from the global schema 11 to the global record table 328.
．． Global component table 338. A global set table 34a is created. In the distributed schema relationship table creation step 82, the distributed record table 358 is created from the distributed schema 12.
1 Dispersion component I - Table 3689 Dispersion set 1 Table 37
a, node management table 38a and local schema table 39
Create a.

The record association setting step 83 sets a pointer between the global record table 32a and the distributed record table 35a. The component related setting step 84 includes the global component table 33a and the distributed component table 3.
A pointer is set between 6a.

The set related setting step 85 includes the global set table 3
A pointer is set between 4a and the distributed set table 37a.

Through the above procedure, direct mapping between the global schema and distributed schema is set.

Next, in the local schema relationship table creation step 86, based on the node clause of the distributed schema 12, the local schema 1 corresponding to the local schema specified in the node clause is
3 are sequentially extracted and associated with the node management table 388 and the local schema table 39a by pointers, and the local record table 40a and the 0-cal component table 41.
a. Local Sera 1 - Create table 42a. Furthermore, the record association setting step 87 sets a pointer between the single distributed record table 35a and the local record table 40a.

The component related setting step 88 includes a distributed component association setting step 88.
A pointer is set between the phone 1 table 36a and the local component table 41a. Set related setting step 89
sets a pointer between the distributed set table 37a and the local set table 42a.

Completion determination step 90 determines whether processing of all local schemas specified in the node clause of one distributed schema 12 has been completed, and if not completed, local schema relationship table creation (step 86) and related setting processing are performed. (Steps 87 to 89) are repeated. Furthermore, if the processing of all local schemas has been completed, the mapping management table creation processing ends. By following the above procedure, you can set up direct mapping between global schema and local schema.

FIG. 9 shows an example of a distributed database query processing procedure using the mapping management table created by the above procedure. Hereinafter, using the example shown in FIG. 5, the one-query conversion processing procedure will be described assuming that the following database inquiry command is input.

FIND employee WHERE branch
h=50 The above command executes the component branch
This instructs to search for the record employee whose value is 50.

In the global record table search step 91, the global record table 32a is searched for the record name employee, and from this, the distributed record table 3 is searched via the pointer.
5a is determined directly. In the global component table search step 92, the component name bran
Search the global component table 338 for ch,
From now on, we will use the distributed component table 36 via the pointer.
Determine a directly.

The distribution condition evaluation step 93 evaluates the condition expression bran of the query command based on the distribution condition expression in the distribution record table 35a.
Evaluate ch=50. As a result, the local node management table 38a
(In this case, since the branch is less than 100, to
ky.

), local schema table 39a, local record table 40a, local component table 41
a. Directly determine the local set table 42a.

A local database query generation step 94 generates a database query command for the local database of the node from the above local schema information. A single distributed database query can be realized by transferring this command to the corresponding node.

Note that the timing of creating the mapping management table in this embodiment may be considered in the same manner as in the previous embodiment.

〔Effect of the invention〕

As described in detail above, according to the present invention, a plurality of nodes each having a processor and a database.

comprising a communication network that interconnects the plurality of nodes, configured such that each node processes online access to the database of the plurality of nodes, and a schema representing a database structure in the plurality of nodes; In a distributed database system configured such that the database structure is managed by a schema representing the overall database structure and data processing is performed according to the data structure, query data is converted according to the level between the multi-layered schemas. A means for adding a mapping level indicating necessity.

A data conversion means for converting data is provided based on the mapping level, information indicating whether data conversion is necessary or unnecessary at each level is set using the mapping level addition means, and a database inquiry is input. and the data conversion means controls data conversion based on the information when editing database query data, or directory creation means creates a table that integrates the multi-layered schemas. , providing a query processing means for performing a distributed database query based on the table, and using the directory creation means,
The multi-layered schema is converted into a single unified table, and when a database query is input, the distributed database query processing means is configured to perform direct mapping between schemas based on the table. This has the remarkable effect of realizing a query processing method for a distributed database system that can reduce database query conversion overhead when schemas are multi-layered.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the operating procedure of the first embodiment of the present invention, FIG. 2 is a diagram showing the overall configuration of the DB system of the embodiment, FIG. 3 is a diagram showing the schema hierarchy of the data model, and FIG. 4 is a diagram showing the schema hierarchy of the data model. The figure shows an example of a database, Figure 5 shows an example of a database schema definition, Figures 6 and 7 show an example of the configuration of a mapping management table, and Figures 8 and 9 show an example of a mapping management table. 2 is a flowchart showing the operating procedure of the embodiment. 2a, 2b: node, 3a, 3b: 1-the terminal, 4a,
4b: data dictionary, 5a, 5b: database, 6a,
6b: Distributed DB function, 7a, 7b: UAP, 30a,
30b: Mapping management table, 10.10a: Subschema, 11. lla Nigerlobal schema,
12゜■2: Distributed schema, 13.13a: Local schema, 31a: Management table, 32a Nigero global record table, 33a Nigero global component table, 34a
Nigero-balset table, 35a: Distributed record table, 36
a: Dispersion component table, 37a: Dispersion cellar 1-table,
38a: Node management table, 39a: Local schema table,
40a: Local record table, 41a: Local component table, 42a: Local set. Figure (Part 2) (c) Patent applicant Hitachi, Ltd.

Claims (1)

[Claims] 1. Consisting of a plurality of nodes having processors and databases, and a communication network that interconnects the plurality of nodes, each node processes online access to the databases of the plurality of nodes. In a distributed database system, the database structure is managed by a schema representing the database structure in the plurality of nodes and a schema representing the overall database structure, and data processing is performed according to the data structure. , a means for adding a mapping level indicating the need for conversion of query data corresponding to the level between the multi-layered schemas, and a data conversion means for converting data based on the mapping level, and the mapping level An additional means is used to set information indicating whether data conversion is necessary or unnecessary at each level, and when a database query is input, the data conversion means uses the information based on the information when editing database query data. A query processing method for a distributed database system characterized by controlling data conversion. 2. It is composed of a plurality of nodes having a processor and a database, and a communication network that interconnects the plurality of nodes, and is configured such that each node processes the process of accessing the database of the plurality of nodes online, and , in the distributed database system configured to manage the database structure using a schema representing the database structure in the plurality of nodes and a schema representing the database structure as a whole, and perform data processing according to the data structure, A directory creation means for creating a table that integrates schemas, and a query processing means for performing a distributed database query based on the table, and the directory creation means is used to combine multi-layered schemas into a single unified structure. When a database query is input, the distributed database query processing means
A query processing method for a distributed database system, characterized in that direct mapping between schemas is performed based on the table.