JP2017016208A - Virtual database system management apparatus, management method, and management program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve processing efficiency in an environment of a virtual database system management apparatus.SOLUTION: A method of managing a virtual database system includes: managing physical table information for managing entry numbers associated with data of database systems, and logical table information to be accessed from a user system, so as to associate the entry numbers with the physical table information and the logical table information; holding definition information for tables managed by the database systems; generating a query graph from a query text; referring to a logical table, the physical table information, and the table definition information, to generate corresponding sub-query sentences by decomposing the query graph; estimating a resulting size of the sub-query sentence; determining the order of executing the sub-query sentences and data arrangement of a result of the execution, on the basis of the resulting size; and executing the sub-query sentences, on the basis of the order.SELECTED DRAWING: Figure 2

Description

  The present invention generates a subquery to be executed in each database system in response to a query from a user in an environment where a plurality of database systems exist, and the remaining query is based on the result of each subquery. The present invention relates to a virtual database system management apparatus, a management method, and a management program that improve the efficiency of processing when the above processing is executed.

  A data virtualization system (or a multi-database system), for example, as shown in Patent Documents 1 to 4, each database is used to virtually show a plurality of database systems with different interfaces and data management methods as one database system. Data that the system has is logically integrated and managed, subquery statements are output to the database system corresponding to the given query statement, and the processing results of the subquery statements executed in each database system are collected centrally. , A system that outputs the result by integrating them into one according to a virtual schema (defining a process for converting an actual physical table into a logical table provided to a user).

  These data virtualization systems basically connect to multiple database systems and manage multiple databases as physical database models, while creating logical database models and associating them with physical database models. The physical database model is handled as one logical database model, and a query statement can be executed on the logical database model.

In the above-described virtual database system, a subquery sentence that can be executed in each database system connected to the virtual database system is generated and issued for a query sentence given by a user, and the processing result of the subquery sentence is virtualized. The database system receives and executes the remaining processing of the query statement and returns the result to the user.
However, when executing the remaining processing of the query statement, depending on the result of the sub-query statement executed in each database system, the amount of data to be processed in the virtual database system becomes enormous and the processing time becomes long, or the resource is insufficient There is a problem that the process may not be completed and may end abnormally.

  Therefore, as shown in Patent Document 5, the inventors reconstructed data when generating a subquery for each database system from a query in the environment of a virtual database system management apparatus having a plurality of database systems. We have come up with a virtual database system management device that can improve the processing efficiency by arranging the query and accompanying query transformation.

JP 09-091309 A Japanese Patent Laid-Open No. 10-307743 Special table 2013-508857 gazette Japanese Patent Application No. 2014-057080 Japanese Patent Application 2014-2226050

In the case of this virtual database system management apparatus, when the user system inputs a query statement to the virtual database system, the virtual database system generates a subquery statement corresponding to each database system.
For this reason, the subquery generated by the virtual database system has a one-to-one correspondence with each database system, and thus may be a subquery including a direct product. If it becomes a subquery including a direct product, the processing result becomes enormous and the processing time becomes long, causing the system to run out of resources and causing abnormal termination.

  The present invention has been proposed in view of the above circumstances, and when generating a subquery in a virtual database system, by using a query graph subjected to a decomposition process, the generation unit of the subquery is subdivided, It is an object of the present invention to provide a virtual database system management apparatus, a virtual database system management method, and a virtual database system management program that can improve processing efficiency by preventing the occurrence of a direct product.

  In order to achieve the above object, the present invention generates a query graph from a query statement input by a user in a virtual database system, resulting in an inefficient relationship (a relationship between a plurality of existing database systems, a many-to-many relationship). The query graph of the virtual database system management apparatus is made efficient by decomposing the query graph on the basis of the relationship (1) and generating a subquery corresponding to the decomposed query graph.

That is, claim 1 is characterized in that a virtual database system management apparatus to which a plurality of database systems are connected and data is output according to an inquiry from a user system has the following configuration.
Each entry number is assigned to the physical table information and the logical table information from the physical table information for managing the entry number associated with the data of each database system and the logical table information to be accessed from the user system. In addition to managing the association, definition information for each table managed by the database system is held.
Query sentence analysis department. The query statement analysis unit generates a query graph from a query statement input from the user system, performs decomposition processing on the query graph with reference to the logical table, physical table information, and table definition information. Decide the corresponding subquery and data relocation plan.
Result size estimation unit. The result size estimation unit is called from the query statement analysis unit and estimates the result size of the subquery statement.
Query statement generator. The execution plan generation unit is called from the query statement analysis unit and generates the subquery statement.
Execution plan generator. The execution plan generation unit generates an execution plan based on the output result of the query statement analysis unit.
Execution control unit. The execution plan generation unit executes an inquiry to the database system based on the execution plan output by the execution plan generation unit.

Claim 2 is the virtual database system management apparatus according to claim 1,
The decomposition process for the query graph is related to the relationship between graph nodes.
The query graph is decomposed by deleting different edges in the physical database and deleting edges that are many-to-many with reference to the table definition information.

Claim 3 is the virtual database system management device of claim 2,
The many-to-many relationship is characterized in that the relation for associating the tables connecting the decomposed query graphs is an item with no uniqueness constraint.

Claim 4 is the virtual database system management apparatus of claim 1,
The query statement analysis unit
For the query statement from the user system, the query graph is decomposed to create a subquery statement corresponding to each graph, and a predetermined criterion for a result size estimated from each subquery statement Has the function of aligning in an order distinguished by “sufficiently large” and “approximately equal”,
When one or more “sufficiently large” exists in the result size alignment result, it is estimated that the result size can be reduced by combining the largest subquery statement with the result of another subquery statement. A function of selecting one query statement and adding a table linked to the execution result of the subquery statement of the selected decomposition graph to the logical table information;
The inquiry execution unit
It has a function of performing data rearrangement for adding a table as an execution result of the single database subquery statement to the physical database management information based on the execution order determined by the query input analysis unit.

Claim 5 is the virtual database system management device of claim 4,
The “sufficiently large” is a case where the result size difference is two or more digits,
The “substantially equal” is characterized in that the result size difference is one digit or less.

Claim 6 is a virtual database system management method in which a plurality of database systems are connected and data is output in response to a query statement from a user system.
Each entry number is assigned to the physical table information and the logical table information from the physical table information for managing the entry number associated with the data of each database system and the logical table information to be accessed from the user system. While linking and managing, while holding definition information for each table managed in the database system,
A procedure for generating a query graph from the query statement, referring to the logical table, physical table information, table definition information and performing a decomposition process on the query graph to generate a corresponding subquery statement;
A procedure for estimating a result size of the subquery statement;
A procedure for determining the execution order of the subquery statement and the data arrangement of the execution result based on the estimated result size for the subquery statement;
And a procedure for executing a subquery based on the determined execution order.

Claim 7 is the virtual database system management method,
Each procedure according to claim 6 is repeatedly performed until the number of graphs (elements) subjected to the decomposition processing on the query graph becomes one.

Claim 8 is a virtual database system management program that builds each part of claim 1 on a computer,
The decomposition process for the query graph is related to the relationship between graph nodes.
Processing for generating a subquery statement corresponding to the decomposed query graph by deleting the edge in which the physical database is different and decomposing the query graph by deleting the edge that becomes many-to-many with reference to the table definition information It is characterized by doing.

According to the present invention, in a virtual database system management apparatus in an environment where a plurality of database systems exist, when generating a subquery for each database system from a query statement, a query graph generated from the query statement is generated. As for the relationship between the graph nodes, an edge in which the physical database is different or an edge that is many-to-many is deleted and decomposed to create each subquery statement to generate an execution plan.
Therefore, when generating a subquery from a query, a subquery that avoids the enlargement of intermediate results due to a direct product is generated, so that the processing efficiency of the query in the virtual database system can be improved.

It is a block diagram which shows the structure of the virtual database system management apparatus of this invention. It is a flowchart figure which shows the process sequence in a query sentence analysis part. It is an example of the query sentence Q. It is a virtual database system block diagram on which logical table information, physical table information, and table definition information are displayed. It is an example of the query graph produced | generated from the query sentence. It is an example of the query graph after decomposition | disassembly. It is an example of a subquery. It is an example of the query graph after a connection. It is a virtual database system block diagram in which the data retention status (when TBL_Q3 is added) is displayed. It is an example of a subquery sentence (loop 2nd time). It is an example of the query graph after connection (loop 2nd time). It is a virtual database system block diagram in which the data retention status (when TBL_Q2 is added) is displayed. It is an example of the subquery sentence obtained at the end. It is an example of the subquery sentence after rewriting. It is a virtual database system block diagram in which the data retention status (when TBL_Q1 is added) is displayed.

An example of an embodiment of a virtual database system management apparatus of the present invention will be described with reference to the drawings.
The virtual database system is a system that enables execution of query contents by selecting a database system according to the query contents of the user in an environment where a plurality of database systems having the same data exist, as shown in FIG. The system includes a user system 1 on which a user posts an inquiry, a virtual data system management device 2, and one or more database systems 3 used by the virtual data system management device 2. The user system 1, the virtual data system management device 2, and the database system 3 are connected via a network. The virtual data system management device 2 is constructed on a computer by installing it by downloading software via a recording medium storing a virtual database system management program or the Internet.

The computer in which the virtual database system management apparatus is constructed includes a ROM that stores a basic program including an operating system (OS) and various basic devices, a hard disk drive (HDD) that stores various programs and data, A media drive device that reads programs and data from a storage medium such as a CD-ROM or DVD, a CPU that executes the programs, a RAM that provides a work area for the CPU, and an input / output interface (I / F) A general configuration mainly includes a pointing device such as a display, a keyboard and a mouse, and a parallel / serial I / F communicating with an external device.
In the virtual database system management apparatus 2 of the present embodiment, a virtual database system management program is input from a serial / parallel I / F or read by a media drive device and stored in advance in an HDD. The virtual database system management program is stored in a storage medium, read by the media drive device, and installed in the HDD.

  A virtual database system management device 2 constructed on a computer defines logical database information (database, table) to be accessed by a user in an environment where a plurality of database systems 3 exist, and each database Assign one or more database management systems, databases, and table sets in the system.

  The user system 1 transparently acquires query statement processing results for a plurality of database systems 3 by posting a query statement using an integrated data model provided by the virtual database system management device 2.

  The database system 3 includes a database 31 that actually stores data and a database management system 32 that provides processing for the data. The database system 3 provides processing for data on the database 31 via an interface of the database management system 32.

  The virtual database system management device 2 includes a database interface unit 21 connected to a plurality of database systems 3, a database management unit 22 that acquires and manages information of each database system 3 via the database interface unit 21, and a user system 1 A user interface unit 23 for inputting / outputting data, a query input analyzing unit 24 for analyzing a query statement from the user system 1, and a query statement generating unit for generating a subquery statement from the query statement to each database system 3 25, a result size estimation unit 26 that estimates a result size of a subquery statement, an execution plan generation unit 27 that generates an execution plan of a series of subquery statements, and an execution control unit 28 that executes a subquery statement It is prepared for.

  The user interface unit 23 receives an inquiry sentence from the user system 1 and outputs it to the inquiry sentence analysis unit 24.

  The database management unit 22 defines physical table information for managing entry numbers associated with the database 31 and a logical database to be accessed from the user system 1, and sets each entry number corresponding to the physical table data. The table information database 221 that stores logical table information that is managed in association with each other and the table definition information database 222 that stores definition information for the tables managed in the database 31 are provided to manage the information of each database system 3. doing.

The query statement analysis unit 24 analyzes the syntax of the query statement input via the user interface unit 23 and outputs the analysis result to the execution plan generation unit 27. In addition, the query statement analysis unit 24 refers to the logical table information, physical table information, and table definition information of the database system 3 stored in the database management unit 22 in the syntax analysis of the query statement. Determine data relocation. An appropriate subquery sentence means a subquery sentence in an element (graph) unit of a query graph obtained by performing a decomposition process on the query sentence. By making a subquery sentence in units of elements (graphs) of the query graph, it is possible to prevent a subquery sentence including a direct product.
Processing in the query statement analysis unit 24 regarding calculation of an appropriate subquery statement and determination of data rearrangement will be described later.

The query statement generation unit 25 is called from the query statement analysis unit 24, generates a subquery statement for each element (graph) of the query graph obtained by performing decomposition processing on the query statement, and rearranges data by the subquery statement Performs query statement conversion associated with.
The result size estimation unit 26 is called from the query statement analysis unit 24 and estimates the result size when executing a given subquery statement.
The execution plan generation unit 27 generates an execution plan for the determined series of query statements (subquery statement and converted query statement) based on the output result of the query statement analysis unit 24.

  The execution control unit 28 executes the execution plan generated by the execution plan generation unit 27, executes the subquery statement via the database interface unit 21, and sends the execution result to the user system 1 via the user interface unit 23. Output.

Next, in the virtual database system management apparatus 2, detailed processing (virtual database system management method) in the query statement analysis unit 24 for the query statement input from the user system 1 via the user interface unit 23 is shown in the flowchart of FIG. Will be described with reference to FIG.
The user system 1 inputs an inquiry sentence Q to the user interface unit 23 of the virtual database system management apparatus 2. When the query statement is input from the user interface unit 21, the query statement analysis unit 24 of the virtual database system management apparatus 2 executes the processing flow shown in FIG.
Hereinafter, the processing procedure in the query statement analysis unit 24 will be described by taking the query statement Q5 (one of 22 types of TPC-H benchmarks) of the TPC-H benchmark shown in FIG. 3 as an example.

FIG. 4 shows a configuration diagram of the virtual database system in which the correspondence relationship between the logical table information and the physical table information and the table definition information are clearly shown in FIG.
Two databases 31 (DB1 and DB2) are connected to the virtual database system management apparatus 2, a table lineitem is arranged in the database DB1, and tables customer, orders, suppliers, nations, and regions are arranged in the database DB2. It shall be.
The table information database 221 includes entry numbers (P1 to P6) for association, the type of the database 31 (DB1 or DB2), and the type of table in the database 31 (lineitem, customer, orders, supplier, nation, The physical table information associated with the region) and the logical table information to be accessed from the user system 1 are managed. In the logical table information, link items (P1 to P6) in which corresponding entry numbers are linked to the table type (lineitem, customer, orders, supplier, nation, region) are set, so that the storage destination is stored. The database 31 (database DB1 or database DB2) is associated.

  The table definition information database 222 stores definition information for each table type (lineitem, customer, orders, supplier, nation, region). For example, a plurality of pieces of definition information such as “C_CUSTKEY INTEGER NOT NULL unique” and “C_NATIONKEY INTEGER NOT NULL” are stored for the table customer. Here, if "unique" or "primary key" etc. exists in the definition information, it means that there is a uniqueness constraint for other tables, and if there is no "unique" and "primary key" etc., This means that there is no uniqueness constraint and the table can be connected to multiple tables.

First, when a query statement is input to the query statement analysis unit 24, a query graph as shown in FIG. 5 is generated for customer, orders, lineitem, supplier, nation, and region from which the query statement data is obtained. (Step 101). A graph node corresponds to a table of query statements, and an edge between graph nodes is composed of a relation that links the tables of query statements.
That is, customer, orders, lineitem, supplier, nation, region are graph nodes (corresponding to the query statement table) from the data string below from in FIG. 3, and customer and orders, lineitem and orders, , Lineitem and supplier, customer and supplier, supplier and nation, nation and region. Further, the data range is specified in the region and orders table from the data string below where in FIG.

  Next, decomposition processing is performed on the generated query graph (FIG. 5) (step 102). In the decomposition process, first, the logical table information and physical table information in the table information database 221 are referred to, and the physical table placement destination is confirmed. For relationships between graph nodes, the physical database deletes different edges and decomposes the query graph. In the query graph of FIG. 5, since lineitem is stored in the database DB1 and other graph nodes (tables) are stored in the database DB2, the lineitem and orders and the lineitem and supplier are the decomposition points. That is, since the physical database is different between lineitem and orders, and lineitem and supplier, the edges are deleted and the elligraph is decomposed.

Further, the table definition information in the table definition information database 222 is referred to, edges that are many-to-many are deleted, and the query graph is similarly decomposed. The many-to-many relationship refers to the case where the relation for linking tables is an item that has no uniqueness constraint.
For example, according to the table definition information (FIG. 4), c_custkey = o_custkey, which is a customer-orders relationship, does not decompose between customer-orders because c_custkey includes unique, and there is a uniqueness constraint.
On the other hand, c_nationkey = s_nationkey, which is the customer-supplier relation, is judged as many-to-many because neither item can confirm a unique constraint such as unique. With this operation, the query graph is decomposed by deleting edges with the decomposition point between customer and supplier.

  FIG. 6 shows query graphs G1 to G3 after decomposition by the above processing. The sub-query sentence corresponding to the decomposed query graph eliminates a sub-query sentence having a large intermediate result size such as a direct product, so that efficient query processing can be performed in the virtual database system management apparatus.

Subsequently, using the decomposed query graph, an execution order that considers the result rearrangement of the subquery statement for the query statement Q is assembled.
The decomposition graph list of the query graph decomposed with respect to the query sentence Q is set as {Gi}, and the query sentence list is set as {φ} (step 103). The decomposition graph list {Gi} is {G1, G2, G3} obtained in step 102 of the preprocessing. The query statement list {φ} is initially an empty set. The query sentence Q is a query sentence input from the user system 1.

First, it is determined whether or not the number i of elements in the decomposition graph list {Gi} is greater than 1 (step 104).
In the case of the query statement (TPCH-Q5) in FIG. 3, the number of elements in the corresponding decomposition graph list {Gi} is “3”, which is “2” or more (the number i of elements is greater than 1). A subquery sentence Qi corresponding to each element Gi of the list {Gi} is generated (step 105).
The sub-query sentence corresponding to the element (graph) is to write the query sentence with only the table included in the node of each graph. A subquery sentence Q generated corresponding to each element Gi is as shown in FIG. The decomposition graph lists G1, G2, and G3 in FIG. 6 correspond to the subquery statements Q1, Q2, and Q3, respectively.

Next, for each generated subquery {Qi}, the result size estimation unit 26 estimates the output result size {Size (Qi)} and arranges them in order (step 106). The output result size need only be known and need not be exact. The order alignment by size is performed by “>> (sufficiently large)” and “≈ (approximately equal)” for the elements of the output result size {Qi}. Therefore, A >> B means “A is sufficiently larger than B”, and A≈B means “A is almost equal to B (A >> B or B >> A is not)”.
Note that “>> (sufficiently large)” and “≈ (approximately equal)” are classified according to a predetermined criterion. For example, “>> (sufficiently large)” means that the result size is 1 MB or more and a difference of about two digits or more. For example, when one result size is 1 MB and the other result size is 100 MB, it is determined to be “sufficiently large”. Further, when the result sizes are smaller than 1 MB or the difference in the result sizes is one digit or less, it is classified as “almost equal”.
Also, depending on the result size, “>> (sufficiently large)” may mean a difference of about 3 digits, and “substantially equal” may mean a difference of 2 digits or less.

  With respect to the size of each generated subquery sentence {Qi}, it is determined whether one or more >> exists (step 107). Select one subquery statement that is estimated to be able to reduce the result size by combining with the result of the subquery statement, combine the graphs corresponding to the selected subquery statement, and delete the target graph from the decomposition graph list Gi Then, the selected subquery statement is added to the query statement list, and the logical table information / physical table information update processing is sequentially performed (step 108).

  That is, in this example, assuming that each generated subquery {Qi} is Size (Q1) >> Size (Q2) >> Size (Q3), there is one or more >> as a result of ordering. It can be seen that the result size of Q1 is the largest. This indicates that the execution result of Q1 is enormous and Q1 processing and intermediate result transfer can be a bottleneck in performing query processing.

  Similarly, from the result of ordering, the subquery sentence whose size is estimated to be smaller than Size (Q1) is {Qj}, and the result size of Size (Join (Q1, {Qj}) is estimated to be the smallest. In this example, Q3 is selected because the order of Size (Q1) >> Size (Join (Q1, Q2)) >> Size (Join (Q1, Q3)) is selected in this example. Q3 results executed in database DB2 are placed in database DB1 where Q1 is executed, and a join operation is performed to reduce the result size of Q1, which was a processing bottleneck, and to improve processing efficiency. Means you can.

  “TBL Q3” is connected to the decomposition graph list Gi, and G3 is excluded. “TBL Q3” is a table that temporarily stores the result of Q3. As a result, the graph after processing is as shown in FIG.

  Subsequently, {Q3 → 1} is added to the query statement list. This means that the Q3 result is rearranged in the database DB1.

  In order to temporarily relocate the Q3 result to the database DB1, the logical table information and the physical table information in the database management unit 22 are updated, and “TBL Q3” is added to the logical table information as shown in FIG. Is linked to P7, P7 and database DB1 are associated with the physical table information, and "TBL Q3" is rearranged in database DB1.

Subsequently, the process returns to the loop start of the processing flow (step 104), and it is determined whether or not the number of decomposition graph list Gi elements is two or more.
Since the number of decomposition graph list Gi elements is two, a subquery statement corresponding to each graph is generated. The generated subquery is shown in FIG. It is assumed that Size (Q1) >> Size (Q2) by size estimation and order alignment of the subquery. As a result of ordering, one or more >> exists. A combination that can reduce the result size is searched, the result is Size (Join (Q1, Q2)), and Q2 is selected. “TBL Q2” is connected to the decomposition graph list G1, and G2 is excluded.
The graph after processing is shown in FIG. Add {Q2 → 1} to the query statement list. In order to temporarily relocate the Q2 result to the database DB1, the logical table information and the physical table information are updated, and the configuration diagram of the virtual database system in which the logical table information, the physical table information, and the table definition information are clearly shown. As shown in FIG.

Again, returning to the loop start of the processing flow (step 104), it is determined whether or not the number of decomposition graph list Gi elements is two or more.
Since there is only one decomposition graph list {G1}, the loop is terminated. A corresponding subquery is generated from the remaining graph, and Q1 → DVS is added to the query list (step 109). This means that the result of Q1 is acquired on the virtual database system. The generated subquery is shown in FIG.

  The query statement is rewritten corresponding to “TBL Q1” placed in the virtual database system (step 109). FIG. 14 shows the query sentence after rewriting. The logical table information and physical table information are updated (step 109), and as a final state, the configuration diagram of the virtual database system in which the information of Q1 is reflected in the table definition information is as shown in FIG.

  By executing a series of processing flows shown in FIG. 2, a subquery sentence, a result rearrangement and order of the subquery sentence, and a rewritten query sentence are obtained.

According to the above-described virtual database system management device, management method, and management program, when there are a plurality of database systems 3 and they are handled in an integrated manner, the virtual database system has become a bottleneck in the conventional virtual database system Part of the exchange of large amounts of data between the management apparatus 2 and each database system 3 can be eliminated.
In the virtual database system, when generating a subquery from a query statement, the physical database decomposes the query graph by deleting different edges and deleting the many-to-many edges by referring to the table definition information. Therefore, it is possible to generate a subquery that avoids the enlargement of the intermediate result due to the direct product, and the query processing efficiency in the virtual database system can be improved.
As a result, the query processing in the virtual database system management apparatus 2 is made more efficient, and it is possible to avoid the situation where the process takes a long time or the situation where the process ends abnormally due to a lack of resources, and the query result can be obtained. It becomes possible.

  DESCRIPTION OF SYMBOLS 1 ... User system, 2 ... Virtual database system management apparatus, 3 ... Database system, 21 ... Database interface part, 22 ... Database management part, 23 ... User interface part, 24 ... Query sentence analysis part, 25 ... Query sentence generation part, 26 ... Result size estimation unit, 27 ... Execution plan generation unit, 28 ... Execution control unit, 31 ... Database management system, 32 ... Database, 221 ... Table information database, 222 ... Table definition information database

Claims (8)

In a virtual database system management apparatus in which a plurality of database systems are connected and data is output according to an inquiry from a user system,
Each entry number is assigned to the physical table information and the logical table information from the physical table information for managing the entry number associated with the data of each database system and the logical table information to be accessed from the user system. While linking and managing, while holding definition information for each table managed in the database system,
A query graph is generated from a query statement input from the user system, and a corresponding subquery statement and data are generated by performing decomposition processing on the query graph with reference to the logical table, physical table information, and table definition information. A query statement analysis unit that determines a relocation plan;
A result size estimation unit that is called from the query statement analysis unit and estimates a result size of the subquery statement;
A query statement generation unit that is called from the query statement analysis unit and generates the subquery statement;
An execution plan generation unit that generates an execution plan based on the output result of the query statement analysis unit;
An execution control unit that executes an inquiry to the database system based on the execution plan output by the execution plan generation unit;
A virtual database system management apparatus comprising: The decomposition process for the query graph is related to the relationship between graph nodes.
The virtual database system management apparatus according to claim 1, wherein the query graph is decomposed by deleting an edge having a different physical database and deleting an edge that is many-to-many with reference to the table definition information.   The virtual database system management apparatus according to claim 2, wherein the many-to-many relationship indicates a case in which a relation for linking tables that connect decomposed query graphs is an item that has no uniqueness constraint. The query statement analysis unit
For the query statement from the user system, the query graph is decomposed to create a subquery statement corresponding to each graph, and a predetermined criterion for a result size estimated from each subquery statement For the largest subquery when there is one or more “sufficiently large” in the alignment result of the result size. The logical table is a table that selects one subquery statement estimated to be able to reduce the result size by combining with the results of other subquery statements, and links to the execution result of the subquery statement of the selected decomposition graph. With the ability to add to information,
The inquiry execution unit
2. The virtual device according to claim 1, further comprising: a data relocation function that adds a table that is an execution result of the single database subquery statement to the physical database management information based on an execution order determined by the query input analysis unit. Database system management device. The “sufficiently large” is a case where the result size difference is two or more digits,
The virtual database system management apparatus according to claim 4, wherein the “substantially equal” is a case where the difference in the result size is one digit or less. In a virtual database system management method in which a plurality of database systems are connected and data is output according to a query statement from a user system,
Each entry number is assigned to the physical table information and the logical table information from the physical table information for managing the entry number associated with the data of each database system and the logical table information to be accessed from the user system. While linking and managing, while holding definition information for each table managed in the database system,
A procedure for generating a query graph from the query statement, referring to the logical table, physical table information, table definition information and performing a decomposition process on the query graph to generate a corresponding subquery statement;
A procedure for estimating a result size of the subquery statement;
A procedure for determining the execution order of the subquery statement and the data arrangement of the execution result based on the estimated result size for the subquery statement;
A virtual database system management method comprising: executing a subquery based on the determined execution order.   The virtual database system management method according to claim 6, wherein each procedure is repeatedly performed until there is one graph (element) obtained by performing a decomposition process on the query graph. Each part according to claim 1 is constructed on a computer,
The decomposition process for the query graph is related to the relationship between graph nodes.
Processing for generating a subquery statement corresponding to the decomposed query graph by deleting the edge in which the physical database is different and decomposing the query graph by deleting the edge that becomes many-to-many with reference to the table definition information Virtual database system management program to be performed.