JP2000112982A - Different kind of data base integration system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optimize retrieval by means of integrating individual data bases. SOLUTION: An application 2 inquiry expression is issued. An application interface 12 decomposes the tree structure of an inquiry expression into partial trees in accordance with a data base server managing a table being the operand of an inquiry expression and transfers them to corresponding wrappers 11. The each wrapper 11 evaluates the received part of the inquiry expression and calculates the timewise cost of retrieval and the expected number of retrievals at every node of the tree structure of the inquiry expression. An optimizer 13 deforms the inquiry expression based on timewise cost and the expected number of retrievals, which the wrapper 11 writes. The wrappers 11 are caused to execute the parts of the deformed inquiry expression corresponding the respective wrappers 11. At that time, an executor 14 is caused to execute it when the wrappers 11 cannot execute a processing.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a heterogeneous database integration system using a wrapper architecture.

[0002]

2. Description of the Related Art For a wrapper architecture,
“Cornel University: Predator home page, http: // www.
cs.cornell.edu/Info/Projwcts/PREDATOR/,Aug 1997 ”
"Mary T. Roth, Peter Schwarz: Don't Scrap It, Wra
p it! A Wrapper Architecture for Legacy Data Sourc
es, 1997 ".

[0003]

However, in the above-mentioned prior art, optimization of search of individual databases in different types of databases is disclosed, but optimization of search by integrating individual databases is not disclosed. Not disclosed.

An object of the present invention is to enable search optimization by integrating individual databases.

[0005]

According to a first aspect of the present invention, there are provided an application for issuing a search query expression which can be expressed in a tree structure of an operator, and a plurality of tables each managing a table which is an operant of the operator. Database server, and respectively correspond to the database server, execute each of the operators of the query formula with respect to the database server, and time when executing each of the operators with respect to the database server. Wrapper for estimating the cost and the expected number of searches, first means for transforming the query formula into another query formula that can be expressed by an operator tree structure based on the estimation result, and a query formula for the transformed query formula. A second means capable of executing each operator on behalf of the wrapper, based on the estimation result, Third means for each operator query expression after deformation respectively executed in either one or more of the plurality of wrapper and said second means is a heterogeneous database integration system comprising a.

Therefore, the time cost and the estimated number of searches performed by each wrapper on each database server when performing a search are estimated, and the query formula for the search is transformed into another query formula based on the estimation result. If it is inappropriate to cause the wrapper to execute the search from the estimation result, the second means is substituted.

According to a second aspect of the present invention, in the first aspect of the present invention, the application can add a time cost and a predicted number of searches to a query formula to be issued. Means uses the added time cost and the predicted number of searches as an estimation result.

[0008] Thus, the application can give an estimate of the time cost and the expected number of searches.

[0009]

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

[0010] As shown in FIG.
Issues a data inquiry formula using the heterogeneous database integration system 1. This query expression can be represented by an operator tree structure.

The heterogeneous database integration system 1 has a first
A database server 3, a second database server 4, a third database server 5,... Each of the plurality of database servers manages the data of a table that is an operant of the query expression. In addition, as long as a data set can be sequentially read, a format having a limited function such as a file may be used.

The wrapper 11 is prepared corresponding to each of the above-mentioned database servers, and mediates between the heterogeneous database integration system 1 and each of the above-mentioned database servers.
Specifically, the application interface 12
From the following (described later), a query formula for the corresponding database server is received, and the time cost required to execute the search and the estimated number of data items to be obtained are calculated for each operator constituting the query formula. Further, it converts the query-type operator given from the optimizer 13 (described later) into a corresponding command for the database server and executes the command.

Application interface 12
Divides the query formula issued by the application 2 into subtrees for each of the above database servers, and passes them to the wrapper 11 corresponding to the database server.

The optimizer 13 is the first of the present invention,
Implements the third means, and transforms a query formula issued by the application 2 into another query formula that can be expressed by an operator tree structure based on an estimation of the time cost and the number of data received from the wrapper 11. I do. Also, based on the estimation of the time cost and the number of data received from the wrapper 11, the processing of each subtree of the transformed query formula is performed by any of the wrappers 11, 11,... And the executor 14 (described later). One or more of them are executed.

The executor 14 implements the second means of the present invention, and executes the processing of each subtree of the query formula after the transformation by taking over the wrapper 11. This is performed when the wrapper 11 cannot execute the processing for the operator in the inquiry expression, or when the optimizer 13 determines that the processing by the executor 14 is faster than the processing by the wrapper 11.

Next, a query formula processed by the heterogeneous database integration system 1 will be described. This query expression can be expressed by a combination of relational algebras. The combination of relational algebras is represented by a tree structure of various operators as described below.

Access: Data of the database server is sequentially retrieved. Select: Narrow data according to conditions. Project: Extract some of the elements that make up the data. Join: Creates a join between two tables under certain conditions. Union: Creates the sum of two tables. Intersect: Creates the product of two tables. Difference: Creates the difference between two tables.

Of the above, Access is not a relational algebraic operator, but is required to access the data of the database server. The operands for each operator are one or two tables, and the outputs of the operators are also tables. In addition to the above, any operator using two or less tables as operands can construct a query expression. However, the operator constructing the inquiry formula needs to be able to interpret and process it by the wrapper 11 or the executor 14.

Next, the processing executed by the heterogeneous database integration system 1 will be described. First, the application 2 creates a query formula according to the user's input.
If the condition of Select or Join includes logical sum, logical product, or negation, use the combination of Union, Intersect, and Difference to select only one condition term in Select or Join condition. Convert to

Application interface 12
Decomposes the query expression tree structure into sub-trees and passes them to the corresponding wrapper 11 in accordance with the database server that manages the table that is the operand of each operator of the query expression.

The wrapper 11 evaluates the received query formula part, and calculates the search time cost and the expected number of searches for each node of the query formula tree structure from its own catalog information. At this time, if the time cost of search and the expected number of searches have already been written in the query formula, they are used instead. In the calculation of the temporal cost, a temporal cost for another execution plan may be returned as long as the same processing as the query expressed by the subtree below the node is performed. For example, if the combination of Union, Intersect, and Diffrence can be processed by a single Select, the execution cost of that Select is written to the corresponding node.

The optimizer 13 transforms the query formula based on the time cost written by the wrapper 11 and the expected number. Then, the wrapper 11 executes a portion corresponding to each wrapper 11 of the transformed inquiry formula. At this time, if it is determined that the wrapper 11 cannot be processed, the executor 14 executes the processing. If the executor 14 determines that the processing is faster when executed by the executor 14 based on the expected number, the executor 14 also causes the executor 14 to execute.

The above processing will be described with reference to a specific example. FIG. 2 is a block diagram illustrating a tree structure of an example of a query expression issued by the application 2. An example of this query formula is "search for the affiliation and address of a person whose name includes a (specific surname) and who lives in Tokyo". The tree structure shown in FIG. 2 is a binary tree structure. Then, one of the lower nodes has a table A (“name + affiliation”) in which the name and the affiliation are associated with each other as an operant, and narrows down data including a in the name from this table A by selecting Select ( Select “name = a”) as the operator 21 and the other as an operant table B (“name + address”), which is a table in which names and addresses are associated with each other. Refine data where Tokyo is Tokyo
lect (Select “address = a”) as the operator 22. Then, both lower nodes join (Join) which joins Table A and Table B for data having the same last name.
"A. name = B. Name"). In this example, it is assumed that the first database server 3 manages a table A in which names and affiliations are associated, and the second database server 4 manages a table B in which names and addresses are associated.

When the application 2 creates the query formula shown in FIG. 2 according to the user's input, the application interface 12 breaks down the tree structure of the query formula into subtrees and matches the subtree of the operator 21 to the first database server 3. The subtree of the operator 22 is passed to the wrapper 11 corresponding to the second database server 4.

Each wrapper 11 that has received the subtree evaluates the received query expression part and, based on its own catalog information, determines the time cost of retrieval and the expected number of retrievals for each node of the query expression tree structure. calculate.

Here, the optimizer 13 determines that the wrapper 1
Based on the time cost written by 1 and the expected number of cases, the second
The function of the database server 4 is poor and the operator 22
If it is determined that Access cannot be performed but Access can be performed, for example, the query formula is modified as shown in FIG. That is, the operator 22
Change from Select to Access.

In this example, based on the time cost written by the wrapper 11 and the expected number of records, each subtree of the transformed query formula is converted into the first and second database servers without using the executor 14. Each wrapper 11 corresponding to 3 and 4 is executed. Then, by performing the process of the operator 23, it is possible to "search for the affiliation and address of a person whose name includes a (specific surname) and who lives in Tokyo".

When the time cost of retrieval and the expected number of retrievals have already been written in the query formula, the optimizer 13 uses them instead.

As is clear from the above description, according to the heterogeneous database integration system 1 of this embodiment, the time cost and the search time when each wrapper 11 executes a search for each of the database servers described above. Optimizer 13 estimates the number of predictions of, and based on the estimation result,
Transform a search query into another query. When the optimizer 13 determines from the estimation result that it is inappropriate to cause the wrapper 11 to execute the search, the optimizer 13 substitutes the executor 14 for the cost-based optimization across different databases. Can be done.

Further, if the time cost of search and the expected number of searches have already been written in the query expression, the optimizer 13 uses them instead. An estimate of the number of cases can be given to control cost-based optimization.

[0031]

According to the first aspect of the present invention, the time cost and the estimated number of searches when each wrapper executes a search for each database server are estimated, and a search inquiry is made based on the estimation result. The formula is transformed into another query formula, and if it is not appropriate for the wrapper to perform the search from the estimation result, the second means is substituted, so that the cost-based optimization can be performed on a different database. It can be done across the board.

According to a second aspect of the present invention, in the first aspect of the present invention, an application can control cost-based optimization by giving an estimate of a time cost and a predicted number of searches.

[Brief description of the drawings]

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

FIG. 2 is a block diagram showing an example of a tree structure of a query expression issued by an application of the heterogeneous database integration system.

FIG. 3 is a block diagram illustrating an example of a tree structure of a query expression after transformation by an optimizer of the heterogeneous database integration system.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Heterogeneous database integration system 2 Application 3 Database server 4 Database server 5 Database server 11 Wrapper

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Hideaki Nakayama 1-3-6 Nakamagome, Ota-ku, Tokyo Inside Ricoh Company (72) Inventor Hiroko Mano 1-3-6 Nakamagome, Ota-ku, Tokyo Stock F term in Ricoh Company (reference) 5B075 PP02 PP12 PP23 QS20 QT06 QT10

Claims (2)

[Claims] 1. An application for issuing a search query expression that can be expressed in a tree structure of an operator, a plurality of database servers each managing a table that is an operant of the operator, and a plurality of database servers respectively corresponding to the database server. A wrapper for executing each operator of the inquiry formula on the database server, and estimating a time cost and a predicted number of searches when executing each operator on the database server; First means for transforming the query formula into another query formula which can be expressed by an operator tree structure based on the estimation result; and a second means capable of executing each operator of the transformed query formula by taking over the wrapper. 2 means, and each operator of the inquiry formula after the deformation based on the estimation result Third to each perform any one or more of the plurality of wrapper and said second means
Means for integrating a heterogeneous database. 2. An application can add a time cost and a predicted number of searches to a query formula to be issued, and the first and third means include an estimation of the added time cost and a search prediction. 2. The heterogeneous database integration system according to claim 1, wherein the number of cases is used as an estimation result.