JP2001117945A - Method and system for managing database - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for storing/retrieving a database, with which plural methods for storing databases are managed and a database matched to the retrieving method of data can be retrieved at high speed. SOLUTION: In a database system, plural databases for storing data are designated, the data storage method is changed for each database, plural databases are provided, and when inserting/updating/deleting data, data are double inserted/updated/deleted in the plural databases. When retrieving data, the database to be retrieved is automatically selected from retrieval conditions so that the data in the database can be retrieved at high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a database management method, and more particularly to a database storage method specifying a method of storing a database in a plurality of storage devices in accordance with a certain rule, and a high-speed data search method. .

[0002]

2. Description of the Related Art In a conventional database system, one data storage method is designated or omitted when a table is defined, and only a storage method and a database to be stored are determined.
It had multiple databases and could not store and retrieve data by changing the storage method for each database.

For example, when storing data, see Japanese Patent Application Laid-Open
Japanese Patent Application Laid-Open No.-60122 discloses a method of storing table data into subfiles by changing the dividing method in one database based on the contents of a database operation statement. However, a plurality of different data storing methods are designated. Can not.

[0004] In Japanese Patent Application Laid-Open No. 11-3354, a data cube which is considered to be used is created in advance, and a method of selecting an appropriate search route when a user requests aggregation is disclosed. Discloses the reduction of the number of aggregations by using the dependency between the aggregation keys because the number of combinations of aggregation results becomes very large compared to the method of creating the data cube related to the aggregation when the aggregation is requested are doing.

On the other hand, Japanese Patent Application Laid-Open No. Hei 5-334368 discloses a database query processing method in which a cost evaluation system calculates a cost for an input query and generates an optimal processing procedure for a search. I have.

[0006]

In the prior art database system, only one data storage method can be selected at the time of defining a table. Therefore, what kind of search content is and what the search conditions are are assumed in advance. In addition, it is necessary to select a storage method that speeds up assumed data retrieval. However, it is difficult to assume search contents and search conditions in advance, and even if they are assumed, they may change. In addition, when a plurality of search methods are assumed, it is necessary to select a search method that will be searched more frequently and to select a data storage method.

An object of the present invention is to provide a database management method and system which can select a database suitable for a data search method by managing data by a method of storing a plurality of databases.

[0008]

The above object is achieved in a database management method by specifying a plurality of data storage areas for storing data and specifying a data storage method for each of the plurality of data storage areas. Further, the database system includes a dictionary for storing a data storage method, a database name to be stored, a storage node (position information of each database system connected to a network), and a data update timing,
The dictionary management unit that manages the dictionary, receives instructions from the user for data operations (storage, addition, deletion, update, search, etc.), and processes each unit (dictionary management unit,
Equipped with a database control unit that controls the database input / output unit / optimization unit, specifies a plurality of databases that store data, and specifies a data storage method and update timing (database synchronized when inserting / updating / deleting data) for each database. Update, or update the database collectively later using the journal output when data is inserted / updated / deleted), specify the data storage method, the node to store, the name of the database to be stored and Remembers the timing of data update and inserts / updates data
A database input / output unit having a plurality of database update units that separates a database that is updated synchronously at the time of deletion and a database that is updated collectively from the journal, and receives a data assurance level from the input / output unit when searching for data.
This is achieved by providing an optimizing unit having a database selecting unit for selecting a database to be searched from a data storage method and a search condition, so that data can be searched at high speed. When retrieving data, the optimizing unit that selects the data storage area suitable for retrieving the data from the data storage method for each data storage area is described in Japanese Patent Laid-Open Publication No.
The method disclosed in Japanese Patent No. 34368 is also used. Note that a program for realizing the functions realized by the above-described apparatus and method and a storage medium storing the program may be used.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a database system according to an embodiment of the present invention. FIG. 2 is an example in which a table is defined in a plurality of data storage methods. FIG. Flowchart showing the processing operation of FIG. 4,
Is a diagram for explaining the contents of a dictionary managed by the dictionary management unit, FIG. 5 is an example of defining a table in a plurality of data storage methods, FIGS. 6 and 7 are diagrams for explaining the contents of a dictionary managed by the dictionary management unit, 8 is a block diagram showing a configuration of a plurality of database systems connected to a network according to the present invention. FIG. 9 is a configuration of a plurality of database systems connected to a network, in which a table is defined in a plurality of data storage methods. FIG. 11 is a diagram for explaining the contents of the dictionary managed by the dictionary management unit. FIG. 12 is a flowchart showing a processing operation when the input / output unit updates the table data. FIG. 14 is a flowchart showing a processing operation when searching, and FIG. 14 stores a table used in a relational database and table data. Is a diagram illustrating an example of the structure of the database. 1, 4, 6, 7, 8, 1
In FIGS. 11 and 14, reference numeral 1 denotes an input / output unit for instructing a database control unit 3 of a database system 2 to define a table, search, insert, update, and delete data in a database; 2, a database control unit 3; A database system including a management unit 4, a database input / output unit 5, a plurality of database update units 6, an optimization unit 7, and a database selection unit 3. Reference numeral 3 denotes an input / output unit 1, a dictionary management unit 4, a database input / output unit 5, The database control unit 4 that controls the data and processing of the optimizing unit 7, and when the table of the input / output unit 1 is defined, the table definition information is stored in the dictionary 9.
Table information management table 41, segment information table 4
2. Stored in the cluster key information table 43 and the division information table 44, output, process, and optimize the definition information of the target table when searching, inserting, updating, and deleting the data of the database of the input / output unit 1. Dictionary management department to pass to 5
Is a database input / output unit that performs database input / output in accordance with the input / output order passed from the optimization unit when searching, inserting, updating, and deleting data. The database update unit 7 updates a plurality of databases at the time of insertion, update, and deletion of a database. The database update unit 7 determines the order of database input / output based on table definition information passed from the dictionary management unit 4 at the time of database search. An optimization unit that communicates with the database input / output unit 5; a database selection unit 8 that selects an optimal database for searching when a database is searched under the optimization unit 7 when there are a plurality of databases; Dictionary for managing definition information,
Numeral 10 stores the update history of the database at the time of data insertion, update, and deletion, specifies that the database is inserted, updated, and deleted asynchronously at the time of database recovery or in the definition of the table, and inserts, updates, and deletes data. Journal 11 used when reflecting the data, 11 is the input / output unit 1 and the business result table 141
When defining a database, a database using a salesperson name as a cluster key when two are specified: a database using a salesperson name as a cluster key and a database using a product name as a cluster key. At the time of definition of 141, a cluster key of the salesperson name of the database 11 using the salesperson name as a cluster key when two are specified, a database using the salesperson name as a cluster key and a database using the product name as a cluster key, 13 Is the input / output unit 1 and the sales report 1
At the time of definition of 41, a database using a product name as a cluster key when two are specified, a database using a sales person's name as a cluster key and a database using a product name as a cluster key. When defining 141,
The cluster key of the product name of the database 13 using the product name as the cluster key when two are specified, the database using the sales person name as the cluster key and the database using the product name as the cluster key. Table information management table for managing the table name, the number of segments, the number of columns, the column name, and the data type specified in the definition of the I / O unit 42, the table name, the segment number, the update timing, the cluster specified in the definition of the table of the input / output unit 1 A segment information table holding the presence / absence of a key, a division type, a node name where a data set for storing data exists, and a data set name for storing data. 43 is a table name specified in the table definition of the input / output unit 1; Number, column name specified for cluster key, node name where data set storing cluster key exists, data set name storing cluster key The cluster key information table 44 to be held is a table name, a segment number, a column name specified by division, a division condition, and a node in which a data set for storing data specified by division exists in the table definition of the input / output unit 1. Name, a division information table that holds a data set name for storing data specified for division, a network 61 when the database system 2 is connected to the network,
Reference numerals 62, 63, and 64 denote database systems 2 when the database system 2 is connected via the network 61.
The dictionary 9, the journal 10, and the nodes indicating the positions under the network 61 including the database, 65, 66, and 67 are the dictionary 9, the journal 10, and the database managed by the database system 2.

The input / output unit 1 performs table definition, table data search, insertion, update, and deletion (hereinafter, insertion, update, and deletion are collectively referred to as update).

First, the definition of a table used in the present invention will be described with reference to FIG.

(1) A table name, a column name constituting a table, a column data type, a segment number for having a plurality of databases from the input / output unit 1 (the segment number is a designation of a storage method) And does not indicate a physical storage location), and specifies whether to update the database synchronously or to update it later collectively using the journal 10 when updating data for each segment. For each segment, database storage method, division type, division column name, division condition, storage node name, storage data set name, presence or absence of cluster key for each segment, cluster key column name, node for storing cluster key The database control unit 3 receives the name and the data set name for storing the cluster key as a table definition statement (step 302).

(2) The database control unit 3 includes the input / output unit 1
Is passed to the dictionary management unit 4, the dictionary management unit 4 analyzes the table definition statement, and stores the table name, the number of segments, the number of columns, and the column name in the table information management table 41 of the dictionary 9. , Column data type in the segment information table 42, table name, segment number, update timing, presence or absence of cluster key, division type,
If the table is not divided, the name of the node storing the database and the name of the data set storing the database are stored in the cluster key information table 43 in the table name, segment number,
The column name of the cluster key, the node name for storing the cluster key, and the data set name for storing the cluster key are stored in the partition information table 44 as a table name, a segment number, a column name specified in the partition, a partition condition, and a database for each partition. Is stored, and a data set name for storing a database for each division is stored (step 303).

In the above-described step 302, when data is updated for each segment, whether to update the database synchronously or to collectively update the data later using the journal 10 is to specify a plurality of data storage methods. In order to prevent the data update performance from becoming slower than when only one data storage method is specified by using the journal 10, the data update is not delayed by specifying the batch update using the journal 10. It becomes possible to. If it is not necessary to guarantee the latest data update in a data search by an information system or the like, the high-speed database data search of the present invention can be used.

Next, regarding the definition of the table used in the present invention,
Three definitions: a table definition when two data storage methods are specified, a table definition when five data storage methods are specified, and a table definition specifying five data storage methods when the network 61 is connected Each method will be described. However, these are only one embodiment of the present invention, and there is no limit on the number of data storage methods to be specified or the number of database systems 2 connected to the network 61.

First, referring to FIG. 1, FIG. 2 and FIG.
A definition of a table provided in two database systems and specifying different data storage methods for two data storage areas (segment number 1 and segment number 2) will be described. Here, the node of the database system of FIG. The definition of the table is the definition shown in FIG. 2, and it is possible to specify different data storage methods for a plurality of data storage areas with one definition. The definition of this table is received by the database control unit 3 as described above, and the dictionary management unit 4 analyzes the definition sentence. The dictionary management unit 4 analyzes the definition sentence of FIG. 2 and results in the four tables shown in FIG. Hereinafter, the contents of the definition (FIG. 2) of this table and the analysis result (FIG. 4) of the definition will be described.

First, the name of the table is defined as “sales report”, and the column names that make up the table are “salesperson name”, “customer name”,
Change the data type of “product name”, “quantity”, and column to “CHAR (3
0) ”,“ CHAR (30) ”,“ CHAR (2
0) ”and“ INT. ”The contents are the contents defined in the first and second lines of FIG. 2, and as a result of analyzing the contents, the contents (tables) shown in the table information management table of FIG. Name, number of columns, column name, data type).

In this example, segment numbers "1" and "2" are defined to indicate a case where there are two data storage areas. When updating data for each segment, whether to update the database synchronously or to collectively update later using the journal 10 (designation of update timing) is set to “ON TIME” (synchronous Segment number "2" is set to "ON TIM".
E ". This is the contents of the third and sixth lines in FIG. 2. As a result of analyzing this, the number of segments in the table information management table, the segment information management table, and the cluster key information table in FIG. The update timing is the content of the update timing of the segment information table.

In FIG. 2, after each segment is specified and the update timing is specified, a database storage method is specified for each segment. The storage method for the segment number "1" is specified in the fourth and fifth lines in FIG.
The designation of the storage method for the segment number “2” is shown on the seventh and eighth lines in FIG.

Here, the storage method includes a division type, a division column name, a division condition, a storage node name, a storage data set name, presence or absence of a cluster key, a column name of a cluster key, a node name for storing a cluster key, A method of storing data based on the designation of each item such as a data set name for storing a cluster key. In the present embodiment, if there is no condition specification in parentheses of IN as shown in the fourth line of FIG. 2, it is determined that there is no division. As shown in the tenth to thirteenth lines in FIG. 5, when a condition is specified in parentheses of IN, it is determined that the data storage area is divided. Of course, the method of specifying the division is merely an example, and the division type may be specified by another description. For example, a method may be used in which a specification statement for specifying division is inserted before IN.

For the segment number "1", the division type is not specified, the node name is NODE1, and the data set name of NODE1 is DATASET1 (see FIG. 2).
4th line). For the segment number "2", the division type is not specified, the node name is specified as NODE1, and the data set name is specified as DATASET3 (the seventh line in FIG. 2). These analysis results are indicated by the contents of the node name and the data set name in the segment information table of FIG.

For the segment number "1", there is a cluster key, the column name of the cluster key is the salesperson name, the node name is NODE1, and the data set name of NODE1 is DATA.
Specify SET2 (fifth line in FIG. 2). For the segment number "2", there is a cluster key, the column name of the cluster key is designated as a product name, the node name is designated as NODE1, and the data set name is designated as DATASET4 (8th line in FIG. 2). The above data set specification is shown as an example,
The data set that stores the cluster key may be the same as the data set that stores the database.

As described above, the database control unit 3 receives the contents defined in FIG. 2 as a table definition sentence. In step 303, the database control unit 3
Is passed to the dictionary management unit 4. The dictionary management unit 4 analyzes the definition statement of the received table, and stores items in each table of the dictionary 9.

In the table information management table 41, the table name “sales report”, the number of segments “2”, the number of columns “4”, the column names “salesperson name”, “customer name”, “product name”, “Number”,
Column data types “CHAR (30)”, “CHAR (3
0), "CHAR (20)", and "INT" are stored in the segment information table 42. The table name is "sales report", the segment numbers are "1" and "2", and the update timing is "ON TIME". , “ON TIME”, presence or absence of cluster key “Yes”, “Yes”, division type “No”,
"None", the node names "NODE1" and "NODE1" for storing the database because the table is not divided, and the data set name "DATASET" for storing the database
1 and "DATASET3" are stored in the cluster key information table 43. The table name is "sales report", the segment numbers are "1" and "2", and the cluster key column names are "salesman name" and "product name". , The node names “NODE1” and “NODE1” storing the cluster key, the data set names “DATASET2” storing the cluster key, and “DATA”
SET4 ″ is stored. In the division information table 44, since there is no designation of division in the definition of FIG. 2, no item is stored.

The definition of the table (FIG. 2) and the table (FIG. 4) storing the results of the analysis are shown above. As shown here, different storage methods for different data storage areas can be specified in one definition statement.

Second, the definition of a table specifying five data storage methods will be described with reference to FIGS. Here, a description will be given of a difference from the definition of the table specifying the two data storage methods described above. Step 301, 30
2, in addition to the segment numbers "1" and "2" for having a plurality of databases (or for specifying a plurality of data storage areas), "3", "4",
Specify “5”. At the time of updating data for each segment, the designation of whether to update the database synchronously or to update the data collectively later using the journal 10 is added to the segment numbers “1” and “2”, and the segment number “3” is “ DEL
AY ”(designation to collectively update later using journal 10), segment number“ 4 ”is“ ON TIM ”
E ”and segment number“ 5 ”specify“ DELAY. ”For each segment, the database storage method, division type, division column name, division condition, storage node name, and stored data set name are segment numbers“ 1 ”. , “2”, and for the segment number “3”, the division type is “key range”, the division column name is “number”, and the division conditions are “number <100”, “number <200”, “ Others ”
Specify “NODE1” for each node name to be stored.
The data set names are "DATASET5", "DATA
"SET6" and "DATASET7" The specification of the storage method of these segment numbers "3" is shown in the tenth to thirteenth lines of the description sentence of the definition in FIG.

For the segment number "4", the division type is "key range", the division column name is "number", and the division condition is "number <
“200”, “number <400”, “others”, and “NODE1” for each data set as the node name to be stored (in this embodiment, a single node has a plurality of data storage areas. Therefore, the node name is the same node name for any data set.), And the data set names are “DATASET8”, “DATASET9”, “D
ATASET 10 ”. The storage method of these segment numbers“ 4 ”is specified in the description statement 1 of the definition in FIG.
Lines 5 to 18 indicate this.

For the segment number "5", the division type is "hash", the division column name is "salesperson name", there is no division condition,
The node name to be stored is “NODE” for each data set.
1 ”and the data set name is“ DATASET1
4 "," DATASET15 "," DATASET1 "
6 and "DATASET 17." The designation of the storage method of these segment numbers "5" is shown in lines 24 to 28 of the description sentence of the definition in FIG.

The presence / absence of a cluster key for each segment, the column name of the cluster key, the node name for storing the cluster key, and the data set name for storing the cluster key are specified in addition to the segment numbers "1" and "2". Number "3"
Specifies "no cluster key". Segment number "4"
Indicates “with cluster key”, the column name is “customer name”, and the node name is “NODE1” for each data set.
The data set names are “DATASET11”, “DAT
ASET12 "and" DATASET13 "are specified, and the segment number" 5 "specifies" no cluster key. "The database control unit 3 receives the above definition as a table definition statement.

In step 303, the database control unit 3 passes the table definition sent from the input / output unit 1 to the dictionary management unit 4. The dictionary management unit 4 analyzes the definition statement of the received table and stores it in the items of each table of the dictionary 9. Next, the contents of each table in the dictionary 9 will be described. These four tables are shown in FIGS. 6 and 7.

The table information management table 41 sets the number of segments to "5", and the segment information table 42 sets the segment numbers "3" and "3" in addition to the segment numbers "1" and "2".
“4”, “5”, update timing “DELAY”, “O
N TIME ”,“ DELAY ”, presence or absence of cluster key“ none ”,“ exist ”,“ none ”, division type“ key range ”,“ key range ”,“ hash ”, database storage for dividing table NULL (no value) is stored as the node name to be executed and the data set name for storing the database.

In the cluster key information table 43, in addition to the segment numbers "1" and "2", the segment number "4", the column name "customer name" of the cluster key, the node names "NODE1", "" NODE1 ”,
“NODE1”, data set names to store the cluster key “DATASET11”, “DATASET12”,
“DATASET13” is stored.

The division information table 44 includes a table name “sales report”, segment numbers “3”, “4”, “5”, division column names “number”, “number”, “salesperson name”, division conditions "<1
00 ","<200"," None ","<200","<
400, "none", "none", node names "NODE1", "NODE1",
“NODE1”, “NODE1”, “NODE1”,
“NODE1”, “NODE1”, “NODE1”,
“NODE1”, “NODE1”, data set name “DATASET5” for storing the divided database,
“DATASET6”, “DATASET7”, “DA
TASET8 "," DATASET9 "," DATAS
ET10 "," DATASET14 "," DATASE
T15 "," DATASET16 "," DATASET
17 "is stored.

As described above, the definition of the table (the definition statement of the table shown in FIG. 5) and the result of analyzing the definition and storing it in the items of the dictionary table are shown (FIGS. 6 and 7). As shown here, by applying the present invention, it is possible to specify different division types for different data storage areas (segments). In particular, a plurality of key ranges can be specified, and a key range and a hash can be specified in one definition statement.

Third, referring to FIGS. 8, 10 and 11,
The difference between the table definition specifying the five data storage methods connected to the network 61 and the table definition specifying the five data storage methods not connected to the network 61 will be described. In the above description of FIGS. 1 to 7, the case where the definition statement in the case where a plurality of data storage areas are held in one node and the analysis result thereof are stored in the items of the table in the dictionary has been described. Hereinafter, a case will be described in which a plurality of nodes connected by a network respectively hold a plurality of data storage areas.

When three database systems 2 are connected to the network 61, the respective node names are "NODE1", "NODE2", and "NODE3". In steps 301 and 302, the database of the segment numbers "3", "4", and "5" and the node name for storing the cluster key are respectively set to the segment number "3".
Is "NODE2", segment number "4" is "NODE
The database control unit 3 receives the definition statement of the table in which “2”, segment number “5” is “NODE3”.

In step 303, the database control unit 3 passes the definition statement of the table received from the input / output unit 1 to the dictionary management unit 4. The dictionary management unit 4 analyzes the definition statement of the received table, and converts the node names of the segment numbers “3”, “4”, and “5” of the cluster key information table 43 of the dictionary 9 and the partition information table 44 into “NODE2”. "," NODE2 "," NODE3
Stored as “. The definition sentence of the table is shown in FIG. 9 for reference.

According to the present invention, as described above, even for different data storage areas connected by a network,
It is possible to specify different data storage methods.

Next, the processing operation of the input / output unit 1 for updating the table data to the database system 2 will be described with reference to the flow shown in FIG.

(1) When the input / output unit 1 issues a table data update request to the database control unit 3, the database control unit 3
Is the name of the table whose data is to be updated from the input / output unit 1,
Receiving the data update value (steps 1201, 120
2).

(2) The database control unit 3 passes the name of the table to the dictionary management unit 4, and the dictionary management unit 4 executes the table information management table 41, the segment information table 42, and the cluster key information table of the dictionary 9 of the target table. 43, read the division information table 44,
The data is passed to the database input / output unit 5 (step 1203).

(3) If the number of segments in the table information management table 41 is two or more, the database input / output unit 5 sends the table information management table 41 and the segment information table 42 passed from the dictionary management unit 4 to the multiple database update unit 6. , The values of the cluster key information table 43 and the division information table 44 are passed, and the multiple database update unit 6 verifies whether or not the number of segments has been completed in order to perform data update processing for the number of segments in the table information management table 41 (step 1204).

(4) The multiple database update unit 6 verifies the update timing of each segment in the segment information table 42, and if the update is collectively performed later from the journal (DELAY), performs the processing of the next segment, If the update is performed synchronously (ON TIME), the process proceeds to the next step (step 1205).

(5) The multiple database update unit 6 acquires the presence or absence of a cluster key and the type of division from the segment information table 42 (step 1206).

(6) If there is a cluster key, the multiple database updating unit 6 verifies the node name and data set name where the database is stored from the cluster key information table 43, and if there is no cluster key and the storage is not divided, the segment is updated. The node name and data set name where the database is stored are verified from the information table 42, and if there is no cluster key and the storage is divided, the node name and data set name where the database is stored is verified and updated from the split information table 44. Select the data node to be processed (step 1
207).

(7) The multiple database update unit 6 determines whether the node selected in step 1207 is the node from which the input / output unit 1 has requested data update (hereinafter referred to as its own node) or the input / output unit 1 It is verified whether the node is other than the node that made the request (hereinafter, referred to as another node) (step 120).
8).

(8) In the verification of step 1208, the multiple database updating unit 6 requests the database input / output unit 5 to update the data of the data set to be updated if the node is the own node. Updates the data of the target dataset. When the data is updated, the update history is output to the journal 10 (step 1209).

(9) In the verification at step 1208, if the verification is made at step 1208, the multiple database update unit 6 requests the corresponding node to update the data, and the node that has received the data update request performs steps 1201 to 1209. The data is updated (step 1210).

(10) Steps 1209 and 12
After the end of the process of 10, the process returns to the determination of step 1204.

If the data update timing of the segment information table 42 is determined to be updated collectively from the journal later (DELAY) in the verification in step 1205 described above, the journal 10 output at the time of data update in step 1209 is used. By performing the data update in a batch, the data update result is reflected in the database that has not been updated in synchronization with the data update. The data update result is reflected at certain time intervals or at the request of the input / output unit 1.

Next, the processing operation at the time of updating data will be described using the definitions of the three tables described above and the tables obtained by analyzing the definitions. Here, in the state where data is not included in the sales record table 141, "Saito Ichi" is displayed in "Sales employee name", "Yamano box office" is displayed in "Customer name", and "Product name"
An example in which data of "vacuum cleaner" is inserted into "." And "10" is inserted into "number" is taken as an example.

First, referring to FIGS. 4, 12 and 14,
The processing operation at the time of updating the data of the “sales report” defined by the definition of the table specifying the two data storage methods that are not connected to the network 61 will be described. Steps 1201, 120
2, when the input / output unit 1 requests the database control unit 3 to update the table data, the database control unit 3 sends the table name “sales report” to be updated from the input / output unit 1 , “Yamato Yoko”, “Vacuum Cleaner”, and “10” are received. Step 12
In step 03, the database control unit 3 passes the table name "sales report" to the dictionary management unit 4, and the dictionary management unit 4 executes the table information management table 41, the segment information table of the dictionary 9 of the target "sales report". 42, the cluster key information table 43, and the division information table 44 are read and passed to the database input / output unit 5.
At step 1204, the database input / output unit 5 determines that the number of segments in the table information management table 41 is "2".
The values of the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44 passed from the dictionary management unit 4 to the multiple database update unit 6 are passed to the multiple database update unit 6.
Performs data update processing twice for the number of segments in the table information management table 41. By step 1205, the multiple database update unit 6 determines that the update timing of the segment number “1” in the segment information table 42 is “ON TI
ME ”, the process proceeds to the next step.
According to 206, the multiple database update unit 6 acquires the cluster key “Yes” and the division type “No division” of the segment number “1” from the segment information table 42. Through steps 1207 and 1208, the multiple database updating unit 6 acquires the node name “NODE1” and the data set name “DATASET2” storing the database from the cluster key information table 43, selects its own node, and sets the segment information table 42. The data set name “DATASET1” in which the database is stored is obtained from At step 1209, the multiple database updating unit 6 causes the database input / output unit 5 to update the data set "DATASET1" and the data set "DATASET2" storing the cluster key in the database input / output unit 5.
Of the data set “DATASET1”, “DAT”
ASET2 "data is updated. As a result," Ito "is stored in the database 11 using the salesperson name as a cluster key.
H, "Yamano box office", "Vacuum cleaner", "10" are stored, returning to step 1204, and the multiple database updating unit 6 continues the data updating process.
05, the multiple database update unit 6 proceeds to the next step because the update timing of the segment number “2” in the segment information table 42 is “ON TIME”. In step 1206, the multiple database updating unit 6 acquires the cluster key “Yes” and the division type “No division” of the segment number “2” from the segment information table 42. According to steps 1207 and 1208,
The multiple database updating unit 6 determines from the cluster key information table 43 that the node name “N” where the database is stored is “N”.
ODE1 ”and the data set name“ DATASET4 ”, and selects its own node. From the segment information table 42, the data set name“ DA
TASET3 ″ is obtained.
The multiple database updating unit 6 includes a database input / output unit 5
Data set "DATASE" to be updated
T3 ", a data update request for the data set" DATASET4 "storing the cluster key is made, and the database input / output unit 5 sends the data set" DATA
The data of "SET3" and "DATASET4" are updated, thereby storing "ITOH", "Yamano Yokoyuki", "Vacuum cleaner", and "10" in the database 12 using the product name as a cluster key.Step 1204 Then, since the number of segments in the table information management table 41 has been completed, the data update process ends.

Second, referring to FIG. 6, FIG. 7, and FIG. 12, the processing operation at the time of updating the data of the “sales report” defined by the definition of the table specifying the five data storage methods without connecting to the network 61 The difference from the definition of the table specifying the two data storage methods not connected to the network shown in the first case will be described. At step 1204, the database input / output unit 5 determines that the table information management table 41 passed from the dictionary management unit 4 to the multiple database update unit 6 because the number of segments in the table information management table 41 is “5”
By passing the values of the segment information table 42, the cluster key information table 43, and the division information table 44, the multiple database update unit 6 performs data update processing five times for the number of segments in the table information management table 41. The data update processing of the segment numbers “1” and “2” is completed, and step 1
According to 204, the multiple database update unit 6 continues the data update process. At step 1205, the multiple database update unit 6 determines that the update timing of the segment number “3” in the segment information table 42 is “DELAY”.
Therefore, the process returns to step 1204. When updating the data of the segment number "3", the update result is collectively reflected later using the journal 10. Step 1204
Then, the multiple database update unit 6 continues the data update process. In step 1205, the multiple database update unit 6 proceeds to the next step because the update timing of the segment number “4” in the segment information table 42 is “ONTIME”. In step 1206, the multiple database update unit 6 acquires the cluster key “Yes” and the division type “Key range” of the segment number “4” from the segment information table 42. Step 1
207, 1208, the multiple database updating unit 6
Are the node names "NODE1" and "NODE1" in which the database is stored from the cluster key information table 43.
1 "," NODE1 ", data set name" DATASE
T11 ”,“ DATASET12 ”,“ DATASET
13 ", select its own node, and enter the split information table 4
4, the division column name “number”, the division condition “<100”, “<
200 "," none ", data set names storing the database" DATASET8 "," DATASET "
9 "and" DATASET10 ", and since the number of data to be updated is" 10 ", a data set" DATASET8 "matching the division condition and a data set" DATASET11 "storing the cluster key are selected. According to 1209, the multiple database update unit 6 issues a data update request to the database input / output unit 5 for the data set “DATASET8” to be updated and the data set “DATASET11” storing the cluster key. The unit 5 includes data sets “DATASET8”, “DATA
The data update of SET 11 is performed. Returning to step 1204, the multiple database update unit 6 continues the data update process. In step 1205, the multiple database update unit 6 updates the segment number “5” of the segment information table 42. Since the timing is "DELAY", the process returns to step 1204. The segment number "5"
In the data update, the update result is collectively reflected later using the journal 10. Returning to step 1204,
Since the number of segments in the table information management table 41 has been completed, the data update processing ends.

Third, referring to FIG. 10, FIG. 11, and FIG. 12, the processing operation at the time of updating the data of the “sales report” defined by the definition of the table specifying the five data storage methods connected to the network 61 The difference from the definition of the table specifying the five data storage methods not connecting the network 61 described above will be described. By step 1204,
Since the number of segments in the table information management table 41 is “5”, the database input / output unit 5
To the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44 passed from the dictionary management unit 4,
The multiple database update unit 6 includes a table information management table 41
The data update process is performed five times for the number of segments. The data update processing of the segment numbers “1”, “2”, and “3” ends, and the multiple database update unit 6 continues the data update processing in step 1204. Step 1205
Accordingly, the multiple database update unit 6 proceeds to the next step because the update timing of the segment number “4” in the segment information table 42 is “ON TIME”. In step 1206, the multiple database update unit 6 acquires the cluster key “Yes” and the division type “Key range” of the segment number “4” from the segment information table 42. According to steps 1207 and 1208,
The multiple database updating unit 6 determines from the cluster key information table 43 that the node name “N” where the database is stored is “N”.
By obtaining “ODE2”, “NODE2”, and “NODE2” and selecting another node, the multiple database update unit 6 requests the database input / output unit 5 to update the data to the node “NODE2”, which is the target of data update, in step 1210. The database input / output unit 5 issues a data update request to the node "NODE2" to be updated, and the node "NODE2" that has received the data update request performs a data update process. The update unit 6 continues the data update process.By step 1205, the multiple database update unit 6
Since the update timing of the segment number “5” in the segment information table 42 is “DELAY”, the process returns to step 1204. When updating the data of the segment number “5”, the update result is collectively reflected later using the journal 10. Returning to step 1204, since the number of segments in the table information management table 41 has been completed, the data update processing ends.

Next, with reference to the flow shown in FIG. 13, the processing operation in which the input / output unit 1 searches the database system 2 for table data will be described.

(1) When the input / output unit 1 issues a table data search request to the database control unit 3, the database control unit 3
Is the name of the table to be searched for data from the input / output unit 1,
A data search condition and a data assurance level of whether the data to be searched is guaranteed to be latest or not to be updated are received (steps 1301 and 1302).

(2) The database control unit 3 passes the table name, the data search condition, and the data assurance level to the dictionary management unit 4, and the dictionary management unit 4 sends the table information management table 41 of the dictionary 9 of the target table. Segment information table 42, cluster key information table 4
3. Read the division information table 44 (step 130)
3).

(3) The dictionary management section 4 verifies whether the number of segments in the table information management table 41 is plural (step 1304).

(4) The dictionary management unit 4 verifies whether the data assurance level received from the database control unit 3 is at the level “ON TIME” that guarantees that it is the latest. The table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44 are passed to the optimizing unit 7 (Step 1305).

(5) The dictionary management unit 4 determines that the target table is a plurality of segments in the verification in step 1304, and the data assurance level is “ON TI” in the verification in step 1305.
If the update timing of the segment information table 42 is “MELA”, the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44 of the dictionary 9 excluding the segment whose update timing is “DELAY” are optimized. 7 (step 130)
6).

(6) After the processing in step 1306, and if the data assurance level is not “ON TIME” in the verification in step 1305, the optimizing unit 7 searches the data search conditions and the table information management table 41 and the segment information table 42. , The cluster key information table 43 and the division information table 44 are passed to the database selection unit 8, and the database selection unit 8 analyzes the data search condition and analyzes the table information management table 41, the segment information table 42, the cluster key information table 43, and the An optimal segment for retrieval is selected from the information table 44. Database selector 8
Uses syntax-based optimization or cost-based optimization when selecting the best segment for data retrieval (step 1307).

(7) The optimizing unit 7 determines the data input / output order of the data retrieval of the target table after the processing in step 1307 and when the segment is not a plurality of segments in the verification in step 1304, (Step 1308).

(8) The database input / output unit 5 searches the database according to the database input / output order of the optimizing unit 7 and returns the search data to the input / output unit 1 (step 130).
9).

Next, the processing operation at the time of data search will be described using the definitions of the above three tables. Here, a case where the data shown in the business performance table 141 of FIG. 14 is stored is shown as an example.

First, referring to FIG. 4, FIG. 13 and FIG.
The processing operation at the time of data search of the “sales report” defined in the definition of the table that specifies two data storage methods that are not connected to the network 61 will be described. Steps 1301, 130
2, when the input / output unit 1 issues a table data search request to the database control unit 3, the database control unit 3 sends the table name “sales record table”, And “ON TIME” which guarantees that the data to be searched is the latest. In step 1303, the database control unit 3 stores in the dictionary management unit 4 the table name “sales report”, the data search condition “data of salesperson name”
The data assurance level “ON TIME” is passed, and the dictionary management unit 4 stores the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44 of the dictionary 9 of the target table “sales report”. Read. Steps 1304, 130
5, the dictionary management unit 4 determines that the number of segments in the table information management table 41 is “2” and the data assurance level is “ON TIME”.
Proceed to 06. According to step 1306, the dictionary management unit 4 determines that the update timing of the segment information table 42 is “DELAY” and the table information management table 41 of the dictionary 9 except for the segment whose update timing is “DELAY” because the data assurance level is “ON TIME”. The table 42, the cluster key information table 43, and the division information table 44 are passed to the optimization unit 7. According to step 1307, the optimization unit 7 determines the data search condition “data of“ salesperson name ”is“ Itoichi ”” and table information management table 41, segment information table 42, cluster key information table 43, and division information table 44. Is passed to the database selecting unit 8, which analyzes the data search condition “data of“ salesperson name ”is“ I ”and“ IT ”and“ data ”, and stores the table information management table 41, the segment information table 42, the cluster key information table An optimum segment number "1" for search is selected from the division information table 43 and the division information table 44. When selecting the optimal segment for data search, the database selection unit 8 uses the “salesperson name” as a cluster key and defines the segment number “1” from the table search conditions in the syntax-based optimization. "And select the segment number" 1 "in which the cluster key is defined in the cost-based optimization.
Select “2” and select “salesperson name” from the search conditions in the table.
Is used as a cluster key to finally select the segment "1" in which the table is defined. By step 1308,
The optimization unit 7 searches for “DATASET2” storing the cluster key of “salesperson name” in the data input / output order of the data search of the target table “sales report”,
When a cluster key whose “salesperson name” is “Itoichi” is found, the database input / output unit 5 is notified of a search for a data set “DATASET1” in the database using the “salesperson name” as the cluster key. Step 1
309, the database input / output unit 5
The database is searched according to the database input / output order, and the search data "Ito Yichi", "Yamano box office", "Vacuum cleaner", and "10" are returned to the input / output unit 1. In addition, in Steps 1301 and 1302, if the data search condition ““ product name ”is“ data of “vacuum cleaner” ”from the input / output unit 1, in Step 1307, the database selection unit 8
Analyzes the data search condition “data of“ product name ”is“ vacuum cleaner ”” and finds the optimal segment number for search from the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44. Select “2”. When selecting the best segment for data search, the database selection unit 8 uses the “product name” as the cluster key and defines the segment number “2” from the search condition of the table in the syntax-based optimization. Is selected, and in the cost-based optimization, the segments "1" and "2" in which the cluster key is defined are selected, and the table is defined using the "product name" as the cluster key from the search conditions of the table. Segment number "2"
Is finally selected.

Second, referring to FIG. 6, FIG. 7, and FIG. 13, the processing operation at the time of retrieving data of the “sales report” defined by the definition of the table specifying five data storage methods without connecting to the network 61 Network 6 shown above
The difference from the definition of a table that specifies two data storage methods that do not connect 1 will be described. As a result of steps 1304 and 1305, the dictionary management unit 4 determines that the number of segments in the table information management table 41 is "5" and the data guarantee level is "ON TIME".
Proceed to. At step 1306, the dictionary management unit 4 determines that the update timing of the segment information table 42 is “DELAY” and the segments 9 of the dictionary 9 except for the segments “3” and “5” because the data assurance level is “ON TIME”. Information management table 41, segment information table 42, cluster key information table 4
3. Pass the division information table 44 to the optimization unit 7. According to step 1307, the optimization unit 7 sets the data search condition “data of“ salesperson name ”to“ Itoichi ””, table information management table 41, and segment information table 4
2. The cluster key information table 43 and the division information table 44 are passed to the database selection unit 8, and the database selection unit 8 analyzes the data search condition “data of“ salesperson name ”“ ga ”“ IT ”and manages table information. Table 4
1. From the segment information table 42, the cluster key information table 43, and the division information table 44, the optimum segment number "1" for the search is selected. Step 13
01, 1302, when the data search condition ““ number ”is“ <100 ”data” from the input / output unit 1, in step 1307, the database selection unit 8 sets the data search condition “” number ”to“ < The data "100" is analyzed, and a segment number "4" optimal for retrieval is selected from the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44. When selecting the optimal segment for data retrieval, the database selection unit 8 defines, in the syntax-based optimization, the segment number “4” for which the table is defined using “number” as the divided column name based on the table retrieval condition. And select
In the cost-based optimization, a segment “4” in which a table is defined using “number” as a division column name is selected. Further, in steps 1301 and 1302, the data search condition "" number ""<100
If “DATA” is “DELAY” for which it is not necessary to guarantee that the data to be searched is the latest, step 13
05, the dictionary management unit 4 determines that the data assurance level received from the database control unit 3 is "DELAY" which does not need to guarantee that it is the latest.
The table information management table 41 of the dictionary 9, the segment information table 42, the cluster key information table 43,
The division information table 44 is passed to the optimization unit 7. Step 1
In 307, the database selecting unit 8 analyzes the data search condition “data of“ number ”is“ <100 ”” and analyzes the table information management table 41, the segment information table 42,
Cluster key information table 43 and division information table 4
From 4, the segment number "3" that is most suitable for the search is selected.
When selecting the most suitable segment for data retrieval, the database selector 8 defines, in the syntax-based optimization, the segment number “3” for which the table is defined using “number” as the divided column name from the retrieval condition of the table. , “4”,
Segment number “4” that matches division condition “<100”
Is finally selected. In cost-based optimization,
The segment numbers “3” and “4” for which the table is defined with “number” as the division column name are selected, and the division condition “<10
The segment number “4” corresponding to “0” is finally selected.

Third, referring to FIGS. 10, 11 and 13,
The processing operation at the time of data search of the “sales report” defined by the definition of the table specifying the five data storage methods connected to the network 61
The difference from the definition of a table that specifies five data storage methods that do not connect 1 will be described. In Steps 1301 and 1302, when the data search condition ““ number ”is“ <100 ”data” from the input / output unit 1, Step 1307
, The database selecting unit 8 analyzes the data search condition “data of“ number ”is“ <100 ”” and searches from the table information management table 41, the segment information table 42, the cluster key information table 43, and the division information table 44. Segment number "4" or "1", "2"
Select When selecting the optimal segment for data retrieval, the database selection unit 8 defines, in the syntax-based optimization, the segment number “4” for which the table is defined using “number” as the divided column name based on the table retrieval condition. , And in the cost-based optimization, select the segment number “4” in which the table is defined using “number” as the divided column name, or select the node where the database with the segment number “4” exists as another node. Therefore, the segment number “1” or “2” is selected without selecting the segment number “4”. Whether to select segment number "4" or segment number "1" or "2"
Depends on communication cost. Whether to select the segment number “1” or the segment number “2” of the own node depends on the access cost of the data set.

As described above, according to the present invention, it is possible to specify different data storage methods for a plurality of data storage areas. In the present embodiment, as an example in which a plurality of division methods can be specified in one definition statement, an example in which a plurality of key ranges is specified, a key range and a hash are specified, but another method can be specified. It is. For example,
It is also possible to specify multiple hashes. Also, at the time of retrieval, it is possible to select and search for the most suitable method from among the methods specified when storing.

As described above, according to the present invention, a plurality of databases for storing data are designated, the data storage method is changed for each database, and a plurality of databases are provided to meet various data search conditions. Data retrieval can be performed at high speed.

[0071]

As described above, according to the present invention, by having a plurality of data storage areas with different data storage methods, a plurality of data storage methods can be designated for each data storage area. At the time of a search, a search method suitable for the search can be selected from a plurality of designated storage methods.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a database system according to an embodiment of the present invention.

FIG. 2 is an example in which a table is defined in a plurality of data storage methods.

FIG. 3 is a flowchart illustrating a processing operation when an input / output unit defines a table.

FIG. 4 is a diagram illustrating the contents of a dictionary managed by a dictionary management unit.

FIG. 5 is an example of defining a table in a plurality of data storage methods.

FIG. 6 is a diagram illustrating the contents of a dictionary managed by a dictionary management unit.

FIG. 7 is a diagram illustrating the contents of a dictionary managed by a dictionary management unit.

FIG. 8 is a block diagram showing a configuration of a plurality of database systems connected to a network according to the present invention.

FIG. 9 is an example in which a table is defined in a plurality of data storage methods in a configuration of a plurality of database systems connected to a network.

FIG. 10 is a diagram illustrating the contents of a dictionary managed by a dictionary management unit.

FIG. 11 is a diagram illustrating the contents of a dictionary managed by a dictionary management unit.

FIG. 12 is a flowchart illustrating a processing operation when the input / output unit updates table data.

FIG. 13 is a flowchart showing a processing operation when the input / output unit searches table data.

FIG. 14 is a diagram illustrating an example of a table used in a relational database and a structure of a database storing data of the table.

[Explanation of symbols]

Reference Signs List 1 input / output unit 2 database system 3 database control unit 4 dictionary management unit 5 database input / output unit 6 multiple database update unit 7 optimization unit 8 database selection unit 9 dictionary 10 journal 11, 12, 13, 14 database 41 table information management table 42 Segment information table 43 Cluster key information table 44 Division information table 61 Network 62, 63, 64 Node 65, 66, 67 Database, dictionary and journal 141 managed by database system User table

Claims (9)

[Claims] In a database management method, a plurality of data areas for storing data are managed, and in accordance with a data storage method specified by a user for each of the plurality of data storage areas, the data storage method includes: A database management method characterized by storing data in a data storage area. 2. The database management method according to claim 1, wherein when searching for data, a data storage area suitable for searching for the data is selected from a data storage method for each data storage area. Database management method. 3. A database management system, comprising: a data storage area for storing data in different data storage methods, a database input / output unit for inputting / outputting the data storage area, and means for searching for the stored data. An optimization unit for selecting and notifying the database input / output unit, a dictionary for holding a table information table, a segment information table, and a cluster key information table, and managing the dictionary, analyzing a data storage method, and analyzing the analysis result. A database management system, comprising: a dictionary management unit that stores data in each dictionary table; and a control unit that controls the database input / output unit, the optimization unit, and the dictionary management unit. 4. The database management system according to claim 3, wherein said database management system comprises a dictionary for storing positional information of a plurality of database systems connected to a network, and said plurality of database systems or A database management system characterized by being applied to a form of a parallel database system or a plurality of data storage areas. 5. The database management method according to claim 1, wherein when storing the data in the data storage area, the database is updated in synchronization with the operation of the data, or a journal output when the data is operated. A method for managing a database, comprising specifying, for each data storage method, whether or not to update the database collectively later by using. 6. The database management method according to claim 1, wherein the data is searched by dividing a data storage area to be updated in synchronization with a data operation and a data storage area to be updated collectively from a journal. A database management method characterized in that data is searched separately in a data storage area. 7. The first, second, fifth and sixth aspects of the present invention.
A computer-readable storage medium storing a program for implementing the database management method according to any one of the above. 8. A database management method according to claim 1, wherein a plurality of cluster keys are specified for each data storage area when a data storage method is specified. 9. In the database management method, when a data storage area is specified to be divided, a data storage area is divided by a different key range, a data storage area is divided by a key range, and a hash division is specified. A database management method, characterized in that one of the following combinations is specified in a single definition statement: designation of division by hash for each storage area.