JP2001331485A - Distributed data base arithmetic processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that a selected arithmetic processing can not be changed even in the case that it is considered that throughput is to be raised by a processing method other than the arithmetic processing selected by an evaluation function based on a fixed evaluation index in the case of carrying out a data connection arithmetic processing to a distributed data base system. SOLUTION: Plural composite data base arithmetic processings are generated by a decomposition generation part 11 from a composite data base arithmetic operation supplied from an operator, the evaluation function stored in an evaluation function set file 4 is applied for the plural composite data base arithmetic processings and evaluation information is acquired by an evaluation information acquisition part 12. The composite data base arithmetic processing of high evaluation is selected from the plural composite data base arithmetic processings by an evaluation part 13 based on the acquired evaluation information and performed by an arithmetic operation performance part 14.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed database operation processing device for performing data combining operation processing in a distributed database system.

[0002]

2. Description of the Related Art In a conventional arithmetic processing device having a decentralized database, a database composite operation given by a user is decomposed into basic operations by a fixed evaluation method in which evaluation is performed in advance under a specific evaluation index. , Rearrange the basic operations obtained by decomposition, generate a combination of a plurality of feasible basic operations, evaluate by the above fixed evaluation method, and combine a plurality of feasible basic operations from the evaluation result. Database composite operation processing by selecting and executing the optimal combination from the database.

[0003]

It is anticipated that the conventional decentralized database composite arithmetic processing unit can perform the optimal database composite arithmetic processing with relatively high probability. However, depending on the configuration of the database, the state of distribution of the database, and the processing capacity of the computer configuring the system, it is better to select a processing method other than the arithmetic processing selected by the evaluation function based on a predetermined fixed evaluation index. The best case is considered, but the selection could not be changed. Therefore, there have been problems that processing time is unnecessarily long, computer resources are unnecessarily wasted, and a network load is increased.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and has a plurality of evaluation functions and evaluation indices prepared in advance with respect to a conventional decentralized database combination arithmetic processing apparatus. Alternatively, by holding the evaluation function and the evaluation index added and changed by the user, performing evaluation using the evaluation function based on the evaluation index, and selecting an operation realization processing method that is estimated to be optimal, thereby efficiently performing the arithmetic processing. I do. In addition, as a result of evaluating the results of the arithmetic processing even in the middle of the arithmetic operation, when it is determined that the other arithmetic processing is more efficient, the arithmetic processing method is switched to an arithmetic processing method different from the initial evaluation and the arithmetic processing is continued. Hold. As a result, it is possible to select an optimal operation processing method and realize a flexible and efficient distributed database processing.

[0005]

According to the present invention, a distributed database operation processing for executing a composite database operation processing having two or more basic operators of a database by using information distributed and stored in a plurality of databases. The apparatus inputs and analyzes the execution request of the composite database operation process, and generates another plurality of composite database operation processes capable of obtaining the same operation result as the operation result obtained by executing the composite database operation. A decomposition generation unit;
An evaluation information acquisition unit that simulates each of the plurality of composite database operation processes generated by the decomposition generation unit based on the distribution status of the plurality of databases, and acquires evaluation information for evaluating the composite database operation process, An evaluation unit that compares the evaluation information of the plurality of composite database operation processes acquired by the evaluation information acquisition unit and determines an operation process with a high evaluation from among the plurality of composite database operation processes; And a calculation execution unit that executes the composite database calculation process as the input composite database calculation process.

[0006] The distributed database operation processing device according to the present invention further includes an evaluation function storage unit for storing an evaluation function for evaluating the processing capability of the plurality of database operation processes based on the distribution status of the plurality of databases. Comprising, the evaluation information acquisition unit simulates a plurality of composite database operation processes generated by the decomposition generation unit based on the evaluation function stored in the evaluation function storage unit, to obtain the evaluation information. Features.

Further, in the distributed database operation processing device according to the present invention, each of the plurality of composite database operation processes decomposed by the decomposition generation unit has at least one database operation expression, and is stored in the evaluation function storage unit. The stored evaluation function is an evaluation function that instructs to change the execution order of executing the database operation expressions of the plurality of composite database operation processes.

Further, the distributed database operation processing device according to the present invention is further characterized by further comprising an evaluation function updating unit for changing the evaluation function stored in the evaluation function storage unit.

Further, the distributed database operation processing device according to the present invention further includes a history storage unit for storing the composite database operation processing determined by the evaluation unit as history information, and the evaluation unit includes the decomposition generation unit. If there is a matching composite database operation with reference to the history information stored in the history storage unit for the plurality of composite database operation processes generated by Is determined as a complex database operation process having a high degree of complexity.

Further, in the distributed database operation processing device according to the present invention, the evaluation function has at least index information serving as an index for evaluating the processing capability, and the evaluation function storage unit relates to the evaluation function. Index information indicating an arithmetic expression is stored in association with the evaluation function, and the evaluation information obtaining unit refers to the index information stored in the evaluation function storage unit and stores the index information relating to each of the plurality of composite database arithmetic processes. An evaluation function is obtained, and evaluation information based on the index information of the evaluation function is obtained from the obtained evaluation function.

Further, the distributed database operation processing device according to the present invention further includes an evaluation function selecting section for selecting a predetermined evaluation function from the evaluation functions stored in the evaluation information storage section, and the evaluation information obtaining section. Is characterized by acquiring the evaluation information of a plurality of composite database operation processes generated by the decomposition generation unit based on the evaluation function selected by the evaluation function selection unit.

Further, the distributed database operation processing device according to the present invention is characterized in that the evaluation function selection section selects an evaluation function having predetermined index information stored in the evaluation function storage section.

Further, in the distributed database operation processing device according to the present invention, the plurality of composite database operation processes generated by the decomposition generation unit have one or more database operation expressions, and the evaluation function is the operation function. In the course of executing the database operation expression by the execution unit, the execution unit has at least evaluation policy information indicating that the database operation expression to be executed is evaluated and the execution order is reviewed. According to the evaluation policy information of the function, obtain evaluation information of a database operation expression to be executed from now on, and the evaluation unit performs the database operation expression to be executed based on the evaluation information obtained by the evaluation information obtaining unit. An arithmetic expression having a higher evaluation is determined, and the arithmetic execution unit determines the database determined by the evaluation unit. And executes the arithmetic expression.

Further, the distributed database operation processing device according to the present invention is characterized in that the evaluation function is described by an expression having a predetermined description grammar.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the distributed database processing apparatus of the present invention will be described below. Embodiment 1 FIG. FIG. 1 is a diagram showing a distributed database according to the first embodiment and an expression of a composite database operation input by a user. In FIG.
Reference numerals 2 and 3 denote computer sites (computer sites are also referred to as sites) which are dispersed and dispersed in each region.
Reference numerals 1, 2-1, and 3-1 denote databases existing in each computer site (the databases are also referred to as relations). FIG. 1 shows the stored contents. Each database is named, and database 1-1 is SUPPLIE
R, the database 2-1 is also called SUPLY, and the database 3-1 is also called SHIPMENT. The number of records (cardinal) of each database is the same as that of database 1-1.
Is assumed to be 200 records, the database 2-1 is assumed to be 5000 records, and the database 3-1 is assumed to be 20 records.
Each database is composed of a plurality of attributes. Supplier of database 1-1 is S
It is composed of NUM and NAME attributes, and their fields are 1-1-1 and 1-1-2. SUPLY of database 2-1 is SNUM, SHIPNUM
Are composed of two fields of 2-1-1 and 2-1-1, respectively.
-1-2. SHIPMEN of database 3-1
T is composed of two fields, SHIPNUM and SHIPNAME, which are respectively 3-1-1 and 3-1-2. Each attribute length has the following size. 1-1
-1, 2-1-1 is 4 bytes long, 1-1-2
NAME is 20 bytes long, 2-1-2, 3-1-1 SHIPNUM is 3 bytes long, 3-1-2 SHIPN
AME is 3 bytes long. 5 is computer site 1
Indicates the database operator at the current location.
An example of the expression of the complex database operation processing which is the database processing operation input by the operator 5 is 5-1.
And The flow of the database transfer between the databases required for executing the operation is defined as transfer paths 6-1, 6-2, 6-3, 6-4, 6-5, 6-6. FIG. 2 is a diagram illustrating the contents stored in the evaluation function set file according to the first embodiment. In FIG.
The evaluation function set file No. 4 registers and stores the evaluation functions of this database system. The evaluation function set file 4 is shared by the computer sites 1, 2, and 3. The set function file 4 includes, as an example, the connection operation UNI of 4-1.
It is assumed that a mixed operation evaluation function of ON and JOIN and an evaluation function of two combined operation JOIN operations 4-2 and 4-3 are registered in advance. The evaluation function set file 4 is
The evaluation function is indicated by the title of “content”, and the arithmetic expression related to the evaluation function is indicated by the title of “index”. FIG. 3 is a configuration diagram of the distributed database operation processing device according to the first embodiment. In the figure, 5-1 is the equation 5-1 shown in FIG. Reference numeral 10 denotes the distributed database operation processing device according to the first embodiment. The distributed database operation processing device 10 is a decomposition generation unit that inputs and analyzes Expression 5-1 and decomposes it into another plurality of composite database operation processes that have the same result as Expression 5-1. Reference numeral 4 denotes an evaluation function set file 4 shown in FIG. Reference numeral 12 denotes an evaluation information acquisition unit, which acquires evaluation information for evaluating the processing performance using the evaluation function stored in the evaluation function set file 4 for each of the plurality of composite database operation processes generated by the decomposition generation unit 11. I do. An evaluation unit 13 compares the evaluation information of the plurality of composite database operation processes acquired by the evaluation information acquisition unit 12 and determines a composite database operation process having a high processing capability. Reference numeral 14 denotes an operation execution unit that executes an operation on the databases 1-1, 2-1 and 3-1 based on the composite database operation process determined by the evaluation unit 13. Database 1-1, 2-1 and 3
-1 is the database shown in FIG. FIG.
FIG. 9 is a diagram showing a transition from the process to the middle stage of the composite operation processing.

FIGS. 5 and 6 are flowcharts showing the operation of the distributed database operation processing device of the present invention. Next, the operation of the distributed database processing device will be described with reference to the flowcharts of FIGS. First, the operator 5 determines the expression 5-1 “JOIN (SUPPLIER, (JOIN
(SUPPLY, SHIPMENT) BY SHIPN
UN = SHIPNUM) BY SNUM = SNUM) "
Is entered. It is assumed that the expression of the complex database operation includes two or more basic operators. ADD (record addition), DE as basic database operators
In addition to LETE (delete record), MODIFY (change record), PROJECTION (selection), UNION
There are a plurality of such as (combination of equivalent relational DBs), JOIN (combination of relational DBs specifying fields). Equation 5-1
Is an expression of a composite database operation including two “JOINs”. Equation 5
1 is read (S1), the content of the operation is analyzed by the decomposition generation unit 11, and the result is decomposed into a plurality of other complex database operation processes having the same execution result as the result of executing Expression 5-1. Each of the plurality of composite database operation processes generated by decomposition has one or more basic operation expressions. As a result of the analysis, Equation 5-1 shows that the three database relations (SUPPLIER, SUPLY, and SHIPMEN)
T) is recognized to include the double JOIN processing. As shown in FIG. 1, databases 1-1, 2-1 and 3-
7 shows an example of operation decomposition when Expression 5-1 is specified by a user in an environment where 1 is distributed. Data transmission between databases (hereinafter, the database is also referred to as DB) between the computer sites is expressed as “TRAN (source site name, destination site name, relational DB name)”. For example, from site 1 ("site 1" is "A") to site 2
(Supplier is sent to “Site 2” as “B”), TRAN (A, B, SUPPLIEER)
Notation. Also, the name of the site where the composite operation is performed is described as “basic operation (site name)”. As a result, Equation 5-1 is decomposed into a plurality of sets as in the following example. Similar results are obtained in any of the sets. Note that “site 3” is “C”. (1) (SEMIJOIN (SUPPLIER, (SE
MIJOIN (SUPLY, SHIPMENT) BY
(SHIPNUM) BY SNUM)) (2) (SEMIJOIN (SUPPLIER, (SE
MIJOIN (SHIPMENT, SUPLY) BY
(SHIPNUM) BY SNUM)) (3) (SEMIJOIN ((SEMIJOIN (SU
PPLY, SHIPMENT) BY SHIPNU
M), SUPPLIER) BY SNUM)) (4) (SEMIJOIN ((SEMIJOIN (SH
IPMENT, SUPLY) BY SHIPNU
M), SUPLIER) BY SNUM)) (5) TRAN (SUPPLY) TO C JOIN (SUPPLIER, (TRAN (JOIN)
(SUPPLY, SHIPMENT) BY SHIPN
(UM) TO A) BY SNUM)) (6) TRAN (SHIPMENT) TO B JOIN (SUPPLIER, (TRAN (JOIN)
(SUPPLY, SHIPMENT) BY SHIPN
(UM) TO A) BY SNUM)) (7) TRAN (SUPPLY) TO C JOIN ((TRAN (SUPPLIER) TO
C), (JOIN (SUPLY, SHIPMENT)
(BY SHIPNUM) BY SNUM)) (8) TRAN (SHIPMENT) TO B JOIN ((TRAN (SUPPLIER) TO
B), (JOIN (SUPLY, SHIPMENT)
BY SHIPNUM) BY SNUM)) The above (1) to (8) are examples of a plurality of sets of composite database arithmetic processing obtained separately.

In principle, the processing in each location is processed with priority first. 5-
In equation (1), since all of the operations involve discrete relations, there is no priority processing in equation (5-1). Furthermore, since the basic operator of Expression 5-1 is an operation of only the JOIN function,
Analyze and evaluate the JOIN process. When performing the evaluation of the JOIN process, a registered evaluation function is selected. In S4, the index information is “HIS” in the evaluation function set file.
It is confirmed whether or not there is an evaluation function of “TORY”. This processing and S20 branching to YES in S4 to S20
The description of 22 will be given in a later embodiment, and the description is omitted here.

The evaluation information acquisition unit 12 determines in step S3 that the JO
The evaluation function set file 4 is read as an evaluation function of the IN processing, and the index having a JOIN index from an index in which a relational operation name at the head of the evaluation function is described as index information from the read evaluation function record 4-1. 4-2,
4-3 is selected. 4-1 is UNION, JOI
N, but the evaluation function 4-1 is excluded because the composite database operation processing obtained by decomposing in S2 does not include UNION. The remaining evaluation function 4-
If an operator is specified among 2, 4-3, the specified evaluation function is used. The operator explicitly specifies the evaluation function by the evaluation function selector 15. If the evaluation function is not explicitly specified, the evaluation function 4-1
And 4-2 are applied to a plurality of composite database operations obtained by decomposing and generating in S2, and each piece of evaluation information is obtained and compared. Is determined (S5 to S7, S30, S8). The acquisition of the evaluation information is performed by the evaluation information acquisition unit 12 and the evaluation unit 13
The evaluation information acquired by the evaluation information acquisition unit 12 is compared with each other to determine a composite database operation process having a high processing capability. Various methods of operator specification by the evaluation function selection unit 15 are conceivable, but in the case of the first embodiment, selection is made by a record number in the evaluation function set file 4 for simplification. When the evaluation function 4-1 is selected, the evaluation function 4-1 is specified as No. 1, the evaluation function 4-2 is specified as No. 2, and the evaluation function 4-3 is specified as No. 3. Both JOIN of evaluation function 4-2 and evaluation function 4-3
The evaluation function of “” is index information “DATA-TRAN”.
SFER ", which is a function that performs an evaluation using the data transfer amount between distributed computer sites as an index by this designation. Compared to the calculation processing performed at each computer site when performing a JOIN operation on the distributed database Such an index is selected because the amount of data transferred between the distributed computer sites determines the processing time.First, it is assumed that the operator explicitly selects the evaluation function 4-2. According to 4-2, "JOIN" and "D"
According to the keyword specification of ATA-TRANSFER ", the entire relation is transferred between computer sites to the partner computer site, the case where JOIN operation is performed is simulated, and the prediction transfer is performed when each relation is transferred to the partner computer site. (5) to (8) of the complex database operation processing obtained by decomposing in S2 are operation processing using a method of transferring the entire relation to the partner computer site. By comparing the estimated transfer amount evaluation information, the composite database operation with the smallest transfer amount is determined and executed from a plurality of sets of composite database operation processes.
The complex database calculation process determined by the evaluation unit 13 in S30 and S9 is executed by the calculation execution unit 14 as Expression 5-1 given by the operator 5 (S
9). In FIG. 6, the processes of S10 and S40 are shown after S9, but since these processes relate to the process of storing history information, a detailed description will be given in a later embodiment, and the description is omitted here. I do.

In particular, when the determined composite database operation process is complicated, the result of comparing the evaluation information may differ depending on the evaluation function applied to the composite database operation process obtained by decomposing and generating in S2. The optimum selection method in that case will be described below. First, the evaluation information is obtained by applying the 4-2 evaluation function to the JOIN operation of each database operation expression included in a plurality of sets of composite database operation processes obtained by decomposition and generation in S2. In performing this operation, the database 1-1 to the database 1-2 are used.
From database 1-2 to database 1-1, from database 1-1 to database 1-3, from database 1-3 to database 1-1, to database 1
-2 to database 1-3, and six patterns for transferring data from database 1-3 to database 1-2 are considered. According to the evaluation function 4-2, the JOI
Estimate the transfer amount as evaluation information in accordance with the above six data transfer patterns, assuming that all the relations necessary for performing the N operation are transferred to the partner of the JOIN and the partner performs the JOIN processing collectively. . FIG. 7 shows evaluation information obtained by executing the evaluation function 4-2. As shown in FIG. 7, the data transfer amount is calculated for each transfer pattern. Since each attribute length 20 and the number of records 21 are recognized in advance as shown in FIG. 1, the amount of transfer data is determined by the size of each relation. The transfer path 6 shown in FIG.
1 to 6-6 are the same as the transfer paths 6-1 to 6-6 shown in FIG. As for 6-5 and 6-6, since there is no JOIN operation mutually, the display of invalid "*" is described. Evaluation is performed for the other four patterns. Assuming that processing is performed from a pattern with a small transfer amount, the priority order is 6-3, 6-4, 6-2, and 6-1.
In accordance with the priority order of the evaluation function 4-2, the JOIN operation for performing the transfer to the transfer path 6-3 (site 3 → site 2) is given the highest priority, and the JOIN operation is executed in the order described above.

The above equation 5-1 is decomposed and generated into a set of (1) to (8), but if it is decomposed into only one composite database operation processing, the decomposed and generated operation processing is evaluated. Execute as high arithmetic processing. If the decomposition cannot be performed, the equation input by the operator is executed as it is. Also, the evaluation may be performed by including the expression input by the operator in the evaluation target to obtain the evaluation information.

As described above, in the first embodiment, in a distributed database processing apparatus that is combined with a decentralized database, the designated compound database operation is decomposed into a plurality of combinations of basic operations. An example of a distributed database operation processing device having functions and having a function of selecting an optimal operation processing method from a plurality of combinations of basic operation methods according to the distribution situation has been described. Further, in a compound distributed database operation processing device for a distributed database, a processing realizing method for decomposing a specified composite database operation into a plurality of basic operations of a combination and determining one to realize the operation Or, if there is a processing realization method that changes the order of the basic operations constituting the composite operation and realizes it, in the course of performing the composite operation processing on the database, a plurality of evaluations of the database combination operation processing capability by the composite operation processing method Of the evaluation function set file, the user explicitly selects an evaluation function based on the database distribution status and database distribution method, and selects the evaluation method according to the situation and environment. This evaluation method makes it possible to select the optimal complex operation processing method from multiple complex operation processing methods. For explaining an example of a distributed database processing unit for processing carried selects and.

Embodiment 2 FIG. In the second embodiment, a description will be given of an evaluation function having evaluation policy information for evaluating before execution, re-evaluating even during execution of an operation, and reviewing execution. Although "DYNAMIC" is described at the end as the evaluation type of the evaluation expression of the evaluation function 4-2 in FIG. 1, this is the transfer path 6-3 having the highest priority (FIG. 7).
), And then the transfer path 6-2
The evaluation policy information instructs to re-evaluate which one of the transfer path 6-1 and the transfer path 6-1 is to be executed first. If the evaluation policy information is described as “STATIC”, a pre-evaluation is performed before the execution of the transfer path 6-3, and the transfer path 6-
1. Evaluate which of the transfer paths 6-2 is to be executed first. Evaluating all execution orders in advance has an advantage that an optimum process as a whole through a processing sequence can be selected before execution. On the other hand, since the estimation is a pre-execution estimation, an order that is not optimal may be selected. At that point, the operator gives
In the evaluation function 4-2, "DYNAMIC" and the evaluation function 4-2
In "3", "STATIC" is designated. Evaluation function 4
In "-2", "DYNAMIC" is designated, so that the data transfer on the transfer path 6-3 is executed first, and the computer site 2
The JOIN operation is performed above. FIG. 4 shows a diagram after the execution result of the JOIN operation. 4, the field configuration of the database 2 is different from that of the database 2-1 of FIG. In FIG. 1, the fields of the database 2-1 are "SNNUM" and "SHIPNUM", but the fields of the database 2 of FIG.
M, SHIPNUM, and SHIPNAME
Has become one. "SHIPNAME" is the transfer path 6-
As a result of the execution of the step 3, the fields of the database 3-1 are transferred to the database 2 and are combined. After that, the number of records of the database 2-1 on the computer site 2 as the execution result becomes clear, and then the transfer path 6-1 and the transfer path 6-2 which are the next optimal data transfer.
Is evaluated, and one of the transfer paths is selected and executed. In this example, it is assumed that the number of SUPPLYRELEATION records in the database 2-1 has become 9000 due to the execution of the transfer path 6-3. In this case, the transfer amount of the transfer path 6-2 is 6 bytes (SHIPNUM (3B) +
SHIPNAME (3B) = 6 bytes) * 9,000 records = 54,000 bytes, which exceeds the transmission amount of the transfer path 6-1 of 48,000 bytes. Initial 5
000 records and transfer route 6-6 from transfer route 6-1
Although it is estimated from the evaluation information that the execution is performed with priority to the second, the trajectory correction is performed by the evaluation during the execution stage by designating “DYNAMIC”, and the transfer path 6-1 is transferred from the transfer path 6-2. Execute data transfer with priority.

As described above, by using the evaluation function specified by the user, the given composite database operation process is decomposed into a plurality of sets of composite database operation processes and evaluated to determine one set of composite database operation processes. Further, it is possible to determine the execution order of the database operation formulas of the determined composite database operation process and execute the optimum operation process on the distributed database.

As described above, in the second embodiment, in the process of performing the joint operation processing on the database, the evaluation function having the evaluation policy information for evaluating the database joint operation processing ability by the joint operation processing method is used. Evaluation policy information that evaluates before and determines the execution method, and evaluation policy information that evaluates before the operation and then re-evaluates even during the execution of the operation and reviews the execution, An example of a distributed database operation processing device that allows a user to select an evaluation function having evaluation policy information that is determined to be appropriate has been described.

Embodiment 3 Next, the operator performs an evaluation function 4-3 on the complex database operation processing of the formula 5-1.
Will be described, that is, the case where the number 3 is specified as the evaluation function.

An embodiment in which the evaluation function 4-3 of FIG. 2 is applied to a JOIN operation will be described below. In the evaluation function 4-3, “SEMIJOIN” is used for the operation, and “DATA TRA” is used for the index information.
NSFER "is specified. In this evaluation function 4-3, the JOIN operation is evaluated using SEMIJOIN in accordance with the specification. SEMIJOIN transmits only relevant attribute data to the partner computer site in advance before executing JOIN. In this example, an operation is performed in which irrelevant records are thinned out, and then only relevant records are sent again.

The SEMIJOIN operation will be described with reference to FIG. In FIG. 8, SEMIJOIN between relations A and B under the condition of attribute "A-1 = B-1"
Is executed. In this case, the entire relation A,
Alternatively, there is a method of transmitting the entire relation B to the partner computer site and performing a JOIN operation at the partner computer site. However, in SEMIJOIN, only relevant attributes are first transmitted to the partner computer site to perform a thinning process. In this example, computer site B with relation B
Transfer the related attribute B-1 to the computer site A (S51). Actually, the attribute B-1 is used for value comparison, so the attribute B-1
Medium, attributes with duplicate values need not be transferred. Therefore, only the unique attribute B-1 is transferred without duplication. For example, when 200 records are transferred as shown in FIG. 8, the transfer amount is 4 bytes * 200 records and the transfer amount is 8
00 bytes. In that case, of course, attribute B-
The transfer amount is smaller than that of all of them. Further, at the transfer destination computer site A, the value of the transferred attribute B-1 is compared with the attribute A-1 of the relation A,
Extract only records with equal attributes.
After the extraction, only the records extracted at the computer site A,
The data is transferred to the computer site B (S52). If the number of extracted records is 100 records, attribute A-
A length of 204 bytes * 100 records obtained by adding the length of the record of the attribute A-1 and the record of the attribute A-1 results in a transfer amount of 20400 bytes. The computer site B executes JOIN on the attribute A-1 of the relation extracted and transferred from the site A and the attribute B-1 of the relation B. In general, transmitting the attribute B-1 transferred from the computer site B to the computer site A and transmitting the relation A thinned out from the computer site A to the computer site B are more efficient than transmitting all the relations A. Transfer volume is small. The total transfer amount of S51 and S52 is 800 bytes + 20400 bytes = 22200 bytes, and the transfer amount when all the relations A are transmitted is 204 bytes * 200 records = 40800 bytes. It can be seen that the combined transfer amount is smaller. In the above-described processing, the operation of transferring the partially related attribute first, and then performing the thinning-out collation processing based on the attribute at the other end is performed by "S
EMIJOIN ”. Distributed database JOIN
By utilizing the SEMIJOIN operation in the operation,
It is effective in reducing the amount of data transfer. In the evaluation function 4-3, evaluation is performed by applying "SEMIJOIN" and optimizing the method of minimizing the data transfer amount. Such an evaluation is performed as an example in which the evaluation is applied to the database environment of FIG. When SEMIJOIN is used, 6-1, 6-2, 6-3, 6-4
There are four patterns: For each, SEMI
An evaluation is performed when a JOIN operation is performed.

The evaluation is calculated based on the sum of the attribute transfer amount for SEMIJOIN and the relation data amount after being thinned out by SEMIJOIN. In this example, all the relations that had to be originally transferred by JOIN are calculated by the relation amount thinned out by SEMIJOIN, and the estimated value is defined as the cost of the evaluation function 4-3. 6-1 to 6-
6, possible patterns 6-1, 6-2, 6-3, 6-
FIG. 9 shows the result of calculating the cost of each of No. 4 and FIG. According to FIG. 9, when a pattern with a low cost is selected, the SEM
When IJOIN is applied, 6-4, 6-2, 6-
3, 6-1. 6-4 and 6-3, 6-2 and 6
Since -1 overlaps, processing in the order of processing 6-4 and 6-2 is evaluated as the optimal method. As described above, the data transfer amount can be reduced by performing the evaluation based on the simple thinning-out operation by SEMIJOIN.

As described above, in the second and third embodiments, in a distributed database operation processing apparatus for a decentralized database, the operation is decomposed into a plurality of combinations of basic operations for a designated composite database operation process. When there is a processing realization method that is realized by changing the order of the basic operations that constitute the composite operation, or a processing realization method that is realized by changing the order of Multiple index information is held for the evaluation function that evaluates the processing capability, and from this evaluation index information, the index information that the user explicitly chooses as appropriate based on the database distribution status and the database distribution method is selected. Database processing that enables the user to select the evaluation method for each complex operation processing. For explaining an example of a processing device.

Embodiment 4 In the second and third embodiments, a case has been described in which each of the two evaluation functions is explicitly specified by an operator for evaluation. Next, a case where the operator does not explicitly specify the evaluation function will be described. In this case, a plurality of evaluation functions are executed in S8 of FIG. 6, and a total evaluation is performed with reference to each evaluation result to determine an operation execution order. Hereinafter, description will be given according to the example of FIG.

The database operation expression input from the operator 5 is the same as the expression 5-1. However, in this example, evaluation is shown in a case where the evaluation function 4-2 or 4-3 is not explicitly specified by the operator. In that case, JO
There are two evaluation functions 4-2 and 4-3 for the IN operation, and the two evaluation functions are executed before execution to calculate the evaluation result. In such a case, in 4-2, “DYNAMI
Although “C” is specified, in a plurality of evaluation function calculations, evaluation is implicitly performed for all “STATIC.” That is, since the order of all processes is determined before execution, “DYN” is used.
The AMIC "designation is ignored. First, the evaluation function 4-2 is calculated. The calculation result is the same as that in FIG.
The evaluation results are the same as in FIG. 7 and 9, the data transfer amount (the index information is DATA-TRANSFE
Since the evaluation is performed using the common index R), the evaluation results in which the evaluation information is not divided for each evaluation function are shown in FIG. As can be understood from FIG. 10, if the priority is set by paying attention to the transfer cost, the first place is the transfer by the transfer path 6-3 of the evaluation function 4-1 for executing the JOIN, and the second place is the evaluation function of the SEMIJOIN method. 4-2 transfer path 6-4
Will be transferred. In the following, the priority order is determined as shown in FIG. 10 from the third position onward.
Executing the transfer path 6-4 by the JOIN method is an optimal execution order to be processed. This evaluation makes it possible to execute a JOIN operation that is comprehensively suitable even if a plurality of evaluation methods exist.

As described above, according to the fourth embodiment, in a distributed distributed database operation processing apparatus for a decentralized database, the operation is decomposed into a plurality of combinations of basic operations for a specified composite database operation process. If there is a processing realization method that implements the multi-operation processing or a processing realization method that changes the order of the basic operations that compose the multi-operation, there are multiple Perform a comprehensive and unified evaluation under a common index using the evaluation function related to the arithmetic processing from among the evaluation functions that are not only a specific evaluation function but also a plurality of evaluation functions that are held. , An example of a distributed database operation processing apparatus having a method of selecting and executing an operation processing in an optimal order from a plurality of processing methods based on evaluation results It explained.

Embodiment 5 FIG. In the fifth embodiment, as another case where another compound operation is given, a case where an operator inputs a compound operation including not only a JOIN operation but also a UNION operation will be described. An evaluation method for determining the operation order of UNION and JOIN in the composite operation of FIG. 2 will be described with reference to the example of FIG. In this example, SUPPLIE
R type relation database is computer site 1
And SUPLIER1 and SUP in computer site 3 respectively.
It exists under the name PLIER2. In this situation, the operator sets the expression 8-1 “UNION (SUPPLIER) in FIG.
1, (JOIN (SUPPLIER1, SUPLLY,
SNUM = SNUM))) ”is assumed to be input. After inputting Expression 8-1, the decomposition generation unit 11 reads and references the evaluation function set file 4. In the evaluation function set file 4 in FIG. The only evaluation function that includes the union operation is the evaluation function 4-1. Therefore, whether the operator explicitly specifies the evaluation function 4-1 or not, the evaluation function 4
-1 is selected. In Equation 8-1, the UNION operation is J
Although it is specified to be executed prior to the OIN operation, the UNION operation and the JOIN operation are performed according to the evaluation function 4-1.
Performs conversion to change the order of operations. In this example, “UNI
ON (SUPPLIER1, (JOIN (SUPPLIER
ER1, SUPLY, SNUM = SNUM))) is replaced by “JOIN ((UNION (SUPPLIE)
R1, SUPPLIER2), SUPLY, SNUM
= SNUM)) ”and converted to UNIO
The N operation is first performed by SUPPLIER1 and SUPPLIER2.
And then perform a JOIN operation. JOIN
As shown in the second and third embodiments, the processing of the operation is the specified evaluation function when the evaluation function for the JOIN operation is specified, and all the evaluation functions 4- when the evaluation function is not specified. The evaluation information is acquired by applying 2, 4-3 and executed. The execution procedure is as described above. If this evaluation function 4-1 is not registered, the input arithmetic expression is executed with the input expression.

As described above, in the fifth embodiment, in a distributed distributed database operation processing apparatus for a decentralized database, the operation is decomposed into a plurality of combinations of basic operations for a specified composite database operation process. In the case where there is a processing realizing method that determines and executes one from a plurality of combinations, or a processing realizing method or a processing realizing method that realizes by changing the order of the basic operations constituting the composite operation, It has a plurality of evaluation functions for evaluating the arithmetic processing capability for the arithmetic processing realization method. It is possible to select a specific evaluation function according to the situation by the user's selection from among them, and this evaluation result Describes an example of a distributed database operation processing device that automatically selects an optimal processing method from among the operation processing realization methods and performs complex operation processing. It was.

Embodiment 6 FIG. Since the expression of the evaluation function shown in FIG. 2 has an expression format that can be understood by humans, the user can add / change / delete the expression. Additional changes/
FIG. 12 shows the configuration of a distributed database operation processing device that enables deletion. In FIG. 12, reference numeral 16 denotes an evaluation function update unit, which performs addition / change / deletion of an evaluation function stored in the evaluation function set file 4 according to an operator's instruction. Other symbols are the same as those in FIG. FIG. 13 shows an example of a grammar of the evaluation function expressing means of the present invention. The description method shown in FIG. 13 complies with the Bacchus description method. Hereinafter, the text example of FIG. 13 will be described in the order of description. The first line indicates that the evaluation function is described in three forms of (Equation 1), (Equation 2), and (Equation 3). (Equation 1) shows the order of processing of each operation. The operation specified first ("FIRST"
To the operation described after FIRST (“F
(Right side of IRST "). Normally, in a composite operation including a plurality of basic operations, the order of the basic operations can be changed depending on the expansion method. The evaluation function in FIG. 4-1 corresponds to this expression 1. The basic operations that can be specified by “operation” are “UNION” and “JOI”.
N "," PROJECTION "," SELEC
T "," INSERT ", and" DELETE ".

In the case where there is a designation in (Equation 1), in a complex operation in which the order can be changed, the arithmetic processing is performed in the designated order.

(Equation 2) shows the execution procedure and evaluation index of the JOIN process. The evaluation methods are “JOIN” and “S
EMIJOIN ”can be selected.“ JOIN ”directly executes the JOIN operation, and if the relation of the JOIN processing is a distributed database, the data is transferred to one computer site of the operation.
IJOIN "performs the SEMIJOIN operation, transfers the thinned relation to the other computer site, and J
OIN processing is performed. As the index information, in addition to the data transfer amount indicated by DATA-TRANSFER shown in FIG. 1, the execution time indicated by “TIME”, and “CPU-
It is possible to select a computer processing time of each computer site indicated by “TIME” as an index.
“CPU-TIME” also has its own index calculation method. In addition, "STATI"
C and DYNAMIC can be selected.
If "TIC" is selected, the execution order of all basic operations is determined before execution of the composite operation, and is not changed during execution.
When "YNAMIC" is selected, the execution order of the basic operation is determined before the execution of the composite operation. However, the evaluation is performed every time the basic operation is completed, and the processing order is changed even during the execution.
“STATIC” has a lower system load for evaluation, but “DYNAMIC” has a chance to review it in the middle and change to the optimal basic operation order. May be selectable.

In (Equation 3), two values, ON and OFF, can be set for “HISTORY”. If it is OFF, no particular processing is performed. The specification of “HISTORY” will be described with reference to the flowcharts of FIGS. Note that "H
A history storage unit used when "ISTORY" is ON is shown in Fig. 14. In Fig. 14, reference numeral 17 denotes a history storage unit, and other symbols are the same as those in Fig. 3. "HIST"
If ON is specified for "ORY" (S4), the history stored in the history storage unit 17 after the composite database operation specified by the operator is decomposed and generated into a combination of a plurality of basic operations. Refer to the information (S2
0), it is checked whether there is history information that matches the combination of the basic operations after decomposition and generation (S21). If there is matching history information, the processing capability has already been evaluated because the processing capability has already been evaluated. This formula is executed without the evaluation by the function (S22, S9). If there is no matching combination expression in the history storage unit 17, an optimum combination is selected from a plurality of combinations of basic operations using an evaluation function and executed (S5). After the execution, the combination formula is stored in the history storage unit 17 (S10, S40).
When the same composite database operation processing is specified next time, the combination of the basic operations stored in the history storage unit is selected, so that the evaluation by the evaluation function can be omitted. Note that the notation example in FIG. 13 shows a text-based expression method, and the user can easily register / add / change / delete the evaluation function definition using an ordinary text editor. However, the internal format for storing the evaluation function does not need to be text, but provides an editor that has a graphical user interface or a user interface in a menu format, and implements evaluation function definition registration / addition / change / deletion. You may.

In the sixth embodiment, in a distributed distributed database operation processing apparatus for a distributed database, processing is realized by decomposing a specified composite database operation into a plurality of basic operations of a combination. If there is a method or a processing realization method that realizes by changing the order of the basic operations constituting the composite operation, in a system having a function of incorporating a function for evaluating the arithmetic processing capability for each operation realization method, In this method, the user can appropriately add / change / evaluate the evaluation function according to the status of the database, the technical environment such as the system environment and the algorithm, and the like. An example of the distributed database operation processing device capable of being deleted has been described. Also, in a compound distributed database operation processing device for a decentralized database, a processing realizing method for decomposing a specified composite database operation into a plurality of basic operations of a combination, or a compound operation When there is a processing realization method that is realized by changing the order of the basic operations to be configured, the processing is selectively executed in a database operation processing system having a function of performing an arithmetic processing by selecting from the plurality of operation realization processing methods. By storing the database composite operation, even when a similar composite operation process is specified by the user, the matching composite database is referred to by referring to the processing method stored as the optimal method before executing the evaluation. An example of a distributed database operation processing device having a function of selecting an operation process has been described.

As described above, according to the distributed database arithmetic processing apparatus described in the first to sixth embodiments, the distributed database arithmetic processing system is integrated with the existing distributed database joint arithmetic processing system so as to meet the joint arithmetic processing conditions and the current database configuration. It is possible to select the most appropriate join operation processing method to realize the database effect operation with the maximum effect. Also, every time the database configuration is changed, the user can explicitly specify or change the join operation processing algorithm, and the database operation processing device allows the database operation processing device to comprehensively select based on the evaluation information. It is also possible to reduce complexity.

[0041]

As described above, an example of the distributed database operation processing device according to the present invention has been described.
The decomposition generation unit generates another plurality of composite database operation processes from the composite database operation process input by the operator, and the generated other plurality of composite database operation processes are stored in, for example, a database by the evaluation information acquisition unit. The evaluation information is obtained by simulating based on the dispersion state of the information that is obtained, the obtained evaluation information is compared, the operation processing having a higher evaluation than the other plurality of composite database operation processing is determined, and the determined operation processing is performed. Is executed by the arithmetic execution unit. For this reason, there is an effect that it is possible to realize arithmetic processing in accordance with the distribution of information stored in a plurality of databases.

Further, the evaluation function storage unit stores an evaluation function for evaluating the processing capability of the plurality of other complex database arithmetic processings based on the distribution status of the database, and uses the stored evaluation function to store the evaluation information. The evaluation unit is configured to acquire the evaluation information. For this reason, since the evaluation function applied regardless of the complex database operation processing input from the operator is the evaluation function stored in the evaluation information storage unit, there is an effect that the environment for acquiring the evaluation information can be unified.

The evaluation function is an evaluation function for instructing to change at least the execution order of executing the database operation expression. For this reason, the execution order of the database operation formula can be changed according to the distribution state of the information stored in the database, and there is an effect of providing a flexible distributed database operation processing environment.

Further, since an evaluation function updating unit is provided, the evaluation function stored in the evaluation function storage unit can be easily added / changed /
The deletion can be updated, and there is an effect that it is possible to cope with a change in the distribution state of the information stored in the database. In addition, the evaluation function can be updated in accordance with a change in the computer system that manages each of the distributed databases, and there is an effect that the system configuration can be flexibly changed.

Further, by providing the history storage unit, it is possible to reuse the evaluation result of the composite database arithmetic processing executed in the past, so that there is an effect that the composite database arithmetic processing can be performed efficiently.

Further, the evaluation function has at least index information as an index for evaluating the processing performance of the complex database operation processing. For this reason, there is an effect that fair evaluation can be performed under a certain evaluation index.

Further, the provision of the evaluation function selection section allows the user to evaluate the composite database operation processing using an arbitrary evaluation function. Conventionally, evaluation is performed based on a predetermined evaluation function. However, since the user can arbitrarily select an evaluation function, there is an effect that a flexible distributed database operation processing device can be realized.

Further, the evaluation function selection section can specify and select an evaluation function having predetermined index information. For this reason, for example, a complex database operation process that can obtain the same result although the operation procedures in the middle are different because they are different operators such as “JOIN operation” and “SEMIJOIN operation” in which the index information of the evaluation function is the same. With respect to the above, there is an effect that a highly evaluated operation process can be determined based on the same index information, and the operation process can be executed efficiently.

The evaluation information includes at least evaluation policy information for instructing to evaluate an arithmetic expression to be executed in the course of the arithmetic processing and to review the execution order. For this reason, when using the evaluation function having the evaluation policy information, the evaluation information acquisition unit may be in the middle of calculation,
The evaluation unit obtains evaluation information of an arithmetic expression to be executed from now on, and the evaluation unit re-evaluates the arithmetic expression based on the evaluation information and determines an arithmetic expression to be executed next. As a result, in the composite database operation processing having a database operation expression that results in an evaluation result different from the initial evaluation result, there is an effect that the operation processing can be realized flexibly and efficiently.

Further, the evaluation function is described by an expression having a predetermined description grammar. For this reason, if the user understands the description grammar, there is an effect that anyone can easily perform the updating / addition / change / deletion of the evaluation function.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an expression of a composite database operation and a distributed database according to the first and second embodiments.

FIG. 2 is a diagram showing an example of the contents of an evaluation function set file according to the first to sixth embodiments of the present invention.

FIG. 3 is a diagram showing a configuration of a distributed database operation processing device according to the first embodiment of the present invention.

FIG. 4 is a diagram illustrating an intermediate stage of a composite database operation according to the first and second embodiments of the present invention.

FIG. 5 is a flowchart showing the operation of the distributed database operation processing device according to the first to sixth embodiments of the present invention.

FIG. 6 is a flowchart showing the operation of the distributed database operation processing device according to the first to sixth embodiments of the present invention.

FIG. 7 is an approximate calculation diagram of an evaluation function 4-2 for a JOIN operation according to the second embodiment of the present invention.

FIG. 8 is SEMIJ in Embodiment 3 of the present invention.
FIG. 1 is a configuration diagram of a distributed database operation processing device illustrating the principle of OIN.

FIG. 9 shows SEMIJ according to Embodiment 3 of the present invention.
FIG. 4 is an approximate calculation diagram of an evaluation function 4-3 of the OIN operation.

FIG. 10 is an approximate calculation diagram showing a joint evaluation result of evaluation functions 4-2 and 4-3 according to the fourth embodiment of the present invention.

FIG. 11 shows a UNIO according to a fifth embodiment of the present invention.
The figure explaining the evaluation function with respect to the mixed operation of N and JOIN.

FIG. 12 is a configuration diagram of a distributed database processing device according to a sixth embodiment of the present invention.

FIG. 13 is a diagram illustrating a description grammar of an evaluation function of an operation according to the sixth embodiment of the present invention.

FIG. 14 is a configuration diagram of a distributed database operation processing device according to a sixth embodiment of the present invention.

[Explanation of symbols]

1,2,3 Computer site, 1-1,2-1,3-1
Database, 1-1-1, 1-1-2, 2-1-1,
2-1, 2, 3-1-1, 3-1-2 Attributes, 4 Evaluation Function Set File, 4-1, 4-2, 4-
3 evaluation function, 5 operator, 5-1 complex database operation, 6-1, 6-2, 6-3, 6-4, 6-5, 6
-6 transfer path, 8-1 complex database operation, 10
Distributed database arithmetic processing unit, 11 decomposition generation unit, 1
2 evaluation information acquisition unit, 13 evaluation unit, 14 operation execution unit, 15 evaluation function selection unit, 16 evaluation function update unit, 1
7 history storage unit, 20 data size, 21 number of records.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 19/00 110 G06F 19/00 110 F Term (Reference) 5B045 DD18 5B049 DD05 EE03 EE05 EE41 FF09 GG07 5B075 KK03 KK13 KK33 KK40 ND03 ND34 QR01 QR02 QR03 QR04 QT06 5B082 GA03

Claims (10)

[Claims] 1. A distributed database operation processing apparatus for executing a composite database operation process having two or more basic operators of a database by using information distributed and stored in a plurality of databases. A decomposition generating unit that inputs and analyzes the execution request of the above and generates another plurality of composite database operation processes that can obtain the same operation result as the operation result obtained by executing the above composite database operation; An evaluation information acquisition unit that simulates another plurality of composite database operation processes generated by the unit based on the distribution status of the plurality of databases, and acquires evaluation information for evaluating the composite database operation process; Compares the evaluation information of multiple complex database operations acquired by the acquisition unit An evaluation unit that determines a highly evaluated operation process from among the plurality of composite database operation processes; and an operation execution unit that executes the composite database operation process determined by the evaluation unit as the input composite database operation process And a distributed database operation processing device. 2. The distributed database processing device according to claim 1,
Further, an evaluation function storage unit that stores an evaluation function for evaluating the processing performance of the plurality of database operation processes based on the distribution status of the plurality of databases, wherein the evaluation information acquisition unit includes the evaluation function storage unit The distributed database operation processing device according to claim 1, wherein the evaluation information is obtained by simulating a plurality of composite database operation processes generated by the decomposition generation unit based on the stored evaluation function. 3. The plurality of composite database operation processes decomposed by the decomposition generation unit each include one or more database operation expressions, and the evaluation function stored in the evaluation function storage unit includes the plurality of composite function units. 3. The distributed database operation processing device according to claim 2, wherein the evaluation function is an evaluation function for instructing to change an execution order of executing a database operation expression included in the database operation process. 4. The distributed database processing device according to claim 1,
3. The distributed database operation processing device according to claim 2, further comprising an evaluation function updating unit that changes an evaluation function stored in the evaluation function storage unit. 5. The distributed database operation processing device according to claim 1,
Further, a history storage unit that stores the composite database operation process determined by the evaluation unit as history information, wherein the evaluation unit is configured to store a plurality of composite database operation processes generated by the decomposition generation unit in the history storage unit. When there is a matching composite database operation with reference to the history information stored in the database, the matching composite database operation is determined as the highly evaluated composite database operation. 2. The distributed database operation processing device according to 1. 6. The evaluation function has at least index information serving as an index for evaluating the processing capability, and the evaluation function storage unit stores index information indicating an operation formula related to the evaluation function in the evaluation function. The evaluation information acquisition unit refers to the index information stored in the evaluation function storage unit, acquires an evaluation function related to each of the plurality of complex database operation processes, and acquires the acquired evaluation. 3. The distributed database operation processing device according to claim 2, wherein evaluation information based on index information of the evaluation function is obtained from the function. 7. The distributed database operation processing device,
The evaluation information storage unit further includes an evaluation function selection unit configured to select a predetermined evaluation function from the evaluation functions stored in the evaluation information storage unit. The evaluation information acquisition unit is configured to perform the evaluation function selection based on the evaluation function selected by the evaluation function selection unit. 3. The distributed database operation processing device according to claim 2, wherein the evaluation information on the plurality of composite database operation processes generated by the decomposition generation unit is acquired. 8. The distributed database operation processing device according to claim 7, wherein the evaluation function selection unit selects an evaluation function having predetermined index information stored in the evaluation function storage unit. 9. The multiple composite database operation process generated by the decomposition generation unit has one or more database operation expressions, and the evaluation function is in the process of executing the database operation expression by the operation execution unit. In this, the database has at least evaluation policy information for instructing to review the execution order of the database arithmetic expressions to be executed, and the evaluation information acquisition unit is configured to execute a database to be executed according to the evaluation policy information of the evaluation function. The evaluation unit acquires evaluation information of the arithmetic expression, and the evaluation unit determines an arithmetic expression having a higher evaluation than a database arithmetic expression to be executed based on the evaluation information acquired by the evaluation information acquisition unit. Executing the database operation formula determined by the evaluation unit. Scan processing unit. 10. The distributed database operation processing device according to claim 2, wherein said evaluation function is described by an expression having a predetermined description grammar.