JP2003141133A - Multi-dimensional database processing method and system for implementing the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique for reducing the number of communications to transmit data used for calculations and increase the speed of inquiry processing. SOLUTION: The multi-dimensional database processing method for transmitting data of a multi-dimensional database stored in a plurality of back end processing device and processing the data thus transmitted comprises a step of inputting inquiry range information indicating sets of members in an inquiry range, classifying the sets of members in the inquiry range every group to which the same calculation equation is applied, and extracting calculation use data range information corresponding to the information of the sets of the members to identify data used for the calculations, and step of achieving the data of the range indicated by the calculation use data range information every calculation equation group, and estimating the data concerned by the calculation equation to calculate the calculation value corresponding to each cross-point value of the inquiry.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multidimensional database processing system for processing data in a multidimensional database, and more particularly to transferring multidimensional database data stored in a plurality of backend processing devices to a frontend processing device. The present invention relates to a technique effectively applied to a multi-dimensional database processing system for processing.

[0002]

2. Description of the Related Art As a technique related to the present invention, a multidimensional database processing method, its parallelization, and a calculation function on a multidimensional database will be described.

First, a multidimensional database processing method and its parallelization will be described. A multidimensional database management system that aims to improve the efficiency of multidimensional data analysis is
It provides the function (slice function) of searching data specified by free combination of dimension members and the function of aggregating data at various levels of detail in various dimension directions as basic functions of data operation.

A typical speed-up method for data aggregation in a multidimensional database management system is a method of completing aggregation calculation when reading data into a database and storing (materializing) the result. When there is an aggregation request,
All that is required is to read the aggregated data that has been materialized. However, it is known that the total amount of aggregated data obtained by aggregating multidimensional raw data in various ways causes a combinatorial explosion with an increase in dimensions and an increase in the dimensional hierarchy that determines the level of detail of aggregation. There is a method to limit the aggregated data to be materialized while minimizing it.

Processing such as data retrieval and reading can be speeded up by applying parallel processing. Multidimensional data including aggregated data to be materialized is divided and stored in a parallel readable storage device, and parallel processing is applied to data retrieval, reading, and the like. An efficient division method of multidimensional data in that case is described in Japanese Patent Laid-Open No. 9-265479.

As described above, when the multidimensional data is divided and stored in a plurality of storage devices, the results from the processing devices for parallel database access as well as the processing devices for parallel database access are summarized. More than one processor is required. This is because processing such as search is not always closed in the processing device for accessing each parallel database. Hereinafter, the processing device for parallel access will be referred to as a back-end processing device, and the processing device that collects the results will be referred to as a front-end processing device. In some cases, one of the back-end processing devices also serves as the front-end processing device.

[0007]

In the conventional multidimensional database processing system described above, if the data required for calculation is not closed in one back-end processing device, it is necessary to collect the necessary data somewhere and perform the calculation. (However, in some cases, it is possible to perform partial calculation in each back-end processing device in a closed range, etc.), but in the query to multiple members from each dimension, the intersection point ( For each (combination of members from each dimension), the necessary data is transferred from the back-end processing device that has the data to the front-end processing device that performs calculations, so that communication between the back-end processing device and the front-end processing device is performed. This may be a performance bottleneck, and therefore it is necessary to reduce the number of communications between the back-end processing device and the front-end processing device.

An object of the present invention is to solve the above problems, to provide a technique capable of reducing the number of communications when transferring data used for calculation and speeding up inquiry processing.

[0009]

The present invention is used for calculation in a multi-dimensional database processing system for transferring data of a multi-dimensional database stored in a plurality of back-end processing devices to a front-end processing device for processing. The data to be obtained is acquired for each group of calculation formulas, and the calculated value at each intersection of the inquiry is obtained.

In the multidimensional database processing system of the present invention, when a query for a multidimensional database is received, query range information indicating a set of members of a query range is input from the query, and the set of those members is calculated by the same calculation. Classify by the group to which the formula applies.

When a calculated member is included in the set of classified members, the calculated member is decomposed into a calculation formula and a range of members to which the calculation formula is applied, and the decomposed calculation formula is applied. Members to identify the data to be used in the calculation by performing the process of classifying into the group that does not include it and the process that does not include the calculated member until the set of classified members does not include the calculated members. The calculation use data range information which is the information of the group is extracted for each group of calculation formulas.

Next, the data of the multidimensional database transmitted from the back-end processing device is requested by requesting the back-end processing device for each group of calculation formulas to transmit the data in the range indicated by the calculation use data range information. Obtained, the data is evaluated by the calculation formula of the group, and the calculated value which is the value at each intersection of the inquiry is obtained.

Here, when the data in the range indicated by the calculation use data range information is acquired for each group of the calculation formula, a request for transmitting data of a plurality of groups is sent to the back-end processing device at one time, The data of a plurality of groups may be collectively acquired, and if there is a dependency relationship between the calculation expressions of each group, the data of each group may be acquired in an order according to the dependency relationship.

As described above, according to the multidimensional database processing system of the present invention, the data used for calculation is obtained for each group of calculation formulas and the calculated value at each intersection of the inquiry is obtained. It is possible to reduce the number of times of communication when transferring a message and speed up inquiry processing.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION A multidimensional database processing system according to an embodiment for transferring data of a multidimensional database stored in a plurality of back end processing devices to a front end processing device for processing will be described below.

FIG. 1 is a diagram showing a schematic configuration of a parallel multidimensional database processing system of this embodiment. As shown in FIG. 1, the front-end processing device 13 of this embodiment includes a calculation use data range extraction processing unit 110 and a calculation evaluation processing unit 1.
11, an inquiry analysis processing unit 112, and an optimization processing unit 11
3, a code generation processing unit 114, a code distribution / execution request processing unit 115, and a data acquisition control processing unit 116.

The calculation use data range extraction processing unit 110 is
The inquiry range information 101 indicating the set of members of the inquiry range is input, the set of members of the inquiry range is classified for each group to which the same calculation formula is applied, and the set of members for identifying the data used for calculation is input. It is a processing unit that extracts the calculation use data range information 102, which is information, for each group of calculation formulas.

The calculation evaluation processing unit 111 acquires the data in the range indicated by the calculation use data range information 102 for each group of calculation formulas, evaluates the data according to the calculation formulas, and uses the value at each intersection of the inquiry. It is a processing unit for obtaining a certain calculated value 104.

The inquiry analysis processing unit 112 uses the inquiry sentence 10
5 is a processing unit that performs syntax analysis and semantic analysis. The optimization processing unit 113 is a processing unit that determines an optimum execution procedure based on the analysis result of the query analysis processing unit 112. The code generation processing unit 114 is a processing unit that generates the code 106 indicating the determined execution procedure.

The code distribution / execution request processing unit 115 is a processing unit that distributes the generated code 106 and requests its execution. The data acquisition control processing unit 116 is a processing unit that controls the acquisition of data from the back-end processing device 14.

The front-end processing device 13 includes a calculation use data range extraction processing unit 110, a calculation evaluation processing unit 111, a query analysis processing unit 112, an optimization processing unit 113, a code generation processing unit 114, and a code distribution / execution request processing unit 115. The program for functioning as the data acquisition control processing unit 116 is recorded in a recording medium such as a CD-ROM, stored in a magnetic disk or the like, and then loaded into a memory and executed. The recording medium for recording the program may be a recording medium other than the CD-ROM. The program may be installed in the information processing apparatus from the recording medium and used, or the recording medium may be accessed through the network to use the program.

The back-end processing device 14 has a data acquisition processing unit 117. The data acquisition processing unit 117
The data requested by the front-end processing device 13 is read from the storage device 17 and the front-end processing device 13 is read.
Is a processing unit for transmitting to.

The program for causing the back-end processing unit 14 to function as the data acquisition processing unit 117 is a CD-
After being recorded in a recording medium such as a ROM and stored in a magnetic disk or the like, it is loaded into a memory and executed.
The recording medium for recording the program is a CD-ROM.
Other recording media other than the above may be used. The program may be installed in the information processing apparatus from the recording medium and used, or the recording medium may be accessed through the network to use the program.

The parallel multidimensional database processing system 12 of this embodiment is a storage device 17 for storing multidimensional data.
A plurality of back-end processing devices 14 for parallel database access to a query, and a query statement 105 for multidimensional data
And one or more front-end processing devices 13 that play the role of a front-end that summarizes the results from the back-end processing devices 14, and between the front-end processing devices 13 and the respective back-end processing devices 14. , Are connected to each other by a high-speed network 18.

In this embodiment, the inquiry sentence 105 is
Any format may be used as long as it includes designation of the range of members to be inquired from each dimension. In the case of FIG. 1, one inquiry is issued and analyzed.
The range of members to be inquired from each dimension, even if the query is of a form that divides it into multiple query statements to form one query, such as sequentially specifying the range of members to inquire for each dimension. The present invention can be applied to any format that can finally analyze.

The front-end processing device 13 is a query sentence 1 for multidimensional data issued by the client 11.
Code 106 that analyzes 05 and represents the processing procedure of the analysis result
The back-end processing device 14 has an analysis processing unit 15 that generates a query and an execution processing unit 16c that executes an inquiry according to the content indicated by the code 106c. It has an execution processing unit 16. The code 106 may be in any format as long as the execution processing unit 16 can execute various processes according to the content thereof, and may be a code in the execution format or a format that is interpreted and executed by an interpreter.

The execution processing unit 16c of the front-end processing device 13 and the execution processing unit 16a of the back-end processing device 14
The execution processing unit 16b and the execution processing unit 16b include processes having different roles, and the execution processing unit 16a and the execution processing unit 16b of the back-end processing device 14 perform a process of accessing multidimensional data (data acquisition processing unit 117). Etc.),
The execution processing unit 16c of the front-end processing device 13 collects the results from the back-end processing device 14 and
It includes a process (a data acquisition control processing unit 116, a calculation evaluation processing unit 111, etc.) for performing calculation based on the acquired result.

In the analysis processing unit 15 of the front-end processing device 13, a query analysis processing unit 112 that performs processing such as syntax analysis and semantic analysis of the query sentence 105, and optimization that determines an optimum execution procedure based on the analysis result. Processing unit 113, code generation processing unit 11 that generates a code indicating the determined execution procedure
Includes 4.

The calculation use data range extraction processing section 110 showing the feature of the present embodiment exists in the analysis processing section 15. In addition, the calculation evaluation processing unit 11 that uses the calculation use data range information 102 created by the calculation use data range extraction processing unit 110 to efficiently execute an inquiry including a calculation.
1 exists in the execution processing unit 16c. In and after FIG. 3, the calculation use data range extraction processing unit 110 and the calculation evaluation processing unit 111, which characterize the present embodiment, will be mainly described.

The parallel multidimensional database processing system 12 of this embodiment does not have to have the system configuration shown in the example of FIG. 1, and the invention can be applied to other system configurations without departing from the gist of the invention. Of course, the present invention can be applied to any configuration in which data used for calculation is communicated from a process of analyzing a query including a calculation and accessing data according to the analyzed result to a process of performing calculation according to the analyzed result. It is possible to apply.

FIG. 2 is a view showing the schematic arrangement of the front end processing apparatus 13 of this embodiment. As shown in FIG. 2, the front-end processing unit 13 of this embodiment includes a display 201, a keyboard 202, and a central processing unit 203 (CP).
U), flexible disk drive 204, main memory 205, magnetic disk device 206, communication control device 207
It can be realized by a general computer including, and a special hardware device only for applying the present embodiment is not necessary.

The display 201 is used to display the execution status of the database management system. The keyboard 202 is used to input a command instructing execution of the database management system.

The central processing unit 203 executes various programs including a program for realizing the present embodiment. The flexible disk drive 204 is used for reading / writing data from / to the flexible disk 209.

The main memory 205 is used to hold various programs including programs for implementing the present embodiment and temporary data associated with program execution and the like.
The magnetic disk device 206 is used for storing a database (multidimensional data), storing temporary data generated during an inquiry operation to the database, and the like.

The communication control device 207 is used for communication via the network 210 such as exchanging requests and data with other database systems having the same configuration in the parallel multidimensional database processing system 12. To be done. The system bus 208 is used to connect these various devices.

In the main memory 205, an operating system 211, a work area 212, a database server processing unit 213, etc. are held.

The operating system 211 provides basic functions for executing various programs including a program for realizing the present embodiment, such as inputting / outputting data to / from a peripheral device. The work area 212 is used to store data that is temporarily needed when the program is executed.

The database server processing unit 213 analyzes the query sentence 105 for the database and executes the code 10
6 is a processing unit that creates 6 or executes an inquiry based on the code 106. In addition to the above functions, the definition of multidimensional data is analyzed and the work area 212 and the magnetic disk device are used as dictionary information. It has a function of registering in 206 or the like. Database server processing unit 21
3 has an analysis processing unit 15 and an execution processing unit 16, and the database management system control unit 214 controls these functions.

The present embodiment relates to analysis and execution of a query of multidimensional data, and these processes of this embodiment are performed by the analysis processing unit 15 and the execution processing unit 16. It is assumed that the work area for execution and the dictionary information are present in the work area 212. The dictionary information may be placed in the magnetic disk device 206 or may be placed in the work area 212 as cache information.

In FIG. 2, the analysis processing unit 15 and the execution processing unit 16 are clearly separated, and the database management system control unit 2
Although it is written under the control of 14, the query analysis is performed even if there is a processing unit commonly executed for each function or if the respective functions are mixed without going through the database management system control unit 214. It does not matter as long as it can execute or execute.

A program for realizing this embodiment is
It is set in the main memory 205 via the flexible disk 209 or the network 210. In the present embodiment, the flexible disk drive 204 is used for exchanging programs and data with the outside. However, another type of portable medium such as a magneto-optical disk or a write-once optical disk is used. You can also

The outline of the multidimensional database processing system of this embodiment will be described below with reference to a simple schematic diagram.

FIG. 16 shows a client in the case where a calculated member is included in the combination of inquired members of this embodiment.
It is a figure which shows an example of the inquiry content from AP. Generally, in a query using a calculation formula for a member, different calculation results are obtained because the applied member and the calculation formula are different depending on the intersection (combination of members from each dimension) related to the calculation formula. Therefore, the calculation formula acquires necessary data for each relevant intersection and evaluates them independently.

In the prior art, in an inquiry for a plurality of members from each dimension as shown in FIG. 16, necessary data is stored for each intersection (combination of members from each dimension) including each calculated member. Since the data is transferred from the back-end processing device 14 to the front-end processing device 13 that performs calculation, the communication between the back-end processing device 14 and the front-end processing device 13 may become a performance bottleneck.

In the present embodiment, in order to solve this, the analysis processing unit 15 obtains data to be used for calculation, and the execution processing unit 16 causes the back-end processing unit 14 to
Consider collecting data in a batch of communication. (It may not be necessary to complete once depending on the size of the communication buffer and the amount of data used for calculation. Can be expected to reduce).

However, it is inefficient to independently obtain and merge a set of members corresponding to the data required for calculation for each intersection (combination of members from each dimension). The parts other than merge will overlap with the processing performed at the time of execution, so the following is done to efficiently find the range of data used for calculation.

Grouping of ranges to which the same formula is applied.-For each grouped range, a set of members for identifying data used in the formula is extracted. This is an element that uses multidimensional data in the formula. If any of the following (1) to (3), the set of members that identify the data to be used should be obtained collectively for the set of members of the query range (range grouped). Because it is possible. However, it is assumed that the calculation formula does not include a conditional formula. If conditional expressions are included, it will be difficult to collect data to be used collectively and accurately.

(1) Element identified by a set of members from any dimension (2) Current member from any dimension to which 0 or more member value expressions are applied (3) For data in the specified range Numerical function of a type that performs calculations other than this, as long as it is an element that can collectively obtain data to be used, it can be applied to the present embodiment without problems.

In this embodiment, the inquiry range (hereinafter referred to as QR
, And the range of data used in the elements in the calculation formula (hereinafter referred to as UA) is a set of the range of the members from each dimension, and thus is expressed as follows. In the present embodiment, for simplification, only when the range of members is designated independently for each dimension will be considered.

[0050]

[Equation 1] QR: = <p, (D1, D2,…, Dp), (QMR1, QMR2,
…, QMRp)> UA: = <p, (D1, D2,…, Dp), (UMR1, UMR2,…, UMR
p)> where p is the number of dimensions, Di is the dimension ID (as long as the information can uniquely identify the dimension), and QMRi and UMRi indicate the range of members from the dimension Di. The member range is the member ID (as long as the information can uniquely identify the member)
Can be represented by a set or the like.

If the range of members is not specified independently for each dimension, the process of obtaining the range of data to be used is
Although it is a little more complicated than that described below, it is possible to extract the members to be used and it is possible to apply this embodiment.

In the present embodiment, when the element (hereinafter, CE) in the calculation formula of the inquiry range QR is the above-mentioned (1) to (3), the range of data used by the element is as follows. Ask.

(1) In the case of an element identified by a set of members from an arbitrary dimension In this case, the element CE in the calculation formula is expressed as follows. Where k is the number of dimensions in the formula (k ≦ p), Dmj is the dimension ID of the jth member (1 ≦ j ≦ k), and Mmj is the ID of the jth member.
(1 ≦ j ≦ k).

[0054]

[Equation 2] CE: = <k, (Dm1, Dm2,…, Dmk), (Mm1, Mm2,…, Mmk)> Range UMRi of each member in the data range UA used by element CE in this formula Is calculated as follows.

That is, the dimension Di (1 ≦ i in the inquiry range QR
≤p) and the dimension CE of the element CE in the formula Dmj (1 ≤ j ≤ k)
If a dimension matching the dimension Di in the query range QR exists in the dimensions Dm1 to Dmk of the element CE in the calculation formula, the member Mmj of the matching dimension Dmj is compared with , Set to member UMRi of dimension Di of data range UA, and query range QR
The dimension Dm of the element CE in the formula is the dimension that matches the dimension Di in
If it does not exist in 1 to Dmk, the member QMRi of the dimension Di in the query range QR is set to the member UMRi of the dimension Di in the data range UA.

(2) Including a current member to which zero or more member value expressions are applied from any dimension In this case, the element CE in the calculation expression is expressed as follows. Here, k is the number of dimensions (k ≦ p), Dmj is the dimension ID of the j-th member (1 ≦ j ≦ k), Mmj is the ID of the j-th member (1 ≦ j ≦ k),
VELmj is a list of member value expressions applied to the jth member (1
≦ j ≦ k). However, in the case of the current member, the member ID is NULL because no member is specified.

[0057]

[Equation 3] CE: = <k, (Dm1, Dm2,…, Dmk), (Mm1, Mm2,
…, Mmk), (VELm1, VELm2,…, VELmk)> The range UMRi of each member in the data range UA used in the element CE in this calculation formula is calculated as follows.

That is, the dimension Di (1 ≦ i in the inquiry range QR
≤p) and the dimension CE of the element CE in the formula Dmj (1 ≤ j ≤ k)
And a dimension matching the dimension Di in the query range QR exists in the dimensions Dm1 to Dmk of the element CE in the calculation formula,
If the member Mmj of the matching dimension Dmj is not NULL (current member), set the member Mmj to the data range U.
Set to member UMRi of dimension Di of A, and if member Mmj of matching dimension Dmj is NULL (current member), apply member value expression VELmj to member QMRi of dimension Di in query range QR This is set in the member UMRi of the dimension Di of the data range UA.

If a dimension matching the dimension Di in the query range QR does not exist in the dimensions Dm1 to Dmk of the element CE in the calculation formula, the member QMRi of the dimension Di in the query range QR.
Is set to member UMRi of dimension Di of data range UA.

(3) In case of a numerical function of a type that operates on data in a specified range A set of members for identifying data used in the numerical function is expressed as follows. For simplification, we will consider only when the range of members is specified independently for each dimension. For simplification, the current member is not included in the member range specified in the numerical function.

In this case, the element CE in the calculation formula is expressed as follows. Where k is the number of dimensions of the member range specified in the numerical function (k ≤ p), Dmj is the dimension ID of the jth member range specified in the numerical function (1 ≤ j ≤ k), and NMRmj is the numerical function It is the specified j-th member range (1 ≤ j ≤ k).

[0062]

[Equation 4] CE: = <k, (Dm1, Dm2,…, Dmk), (NMRm1, NMRm
2, ..., NMRmk)> The range UMRi of each member in the data range UA used by the element CE in this calculation formula is calculated as follows.

That is, the dimension Di (1 ≦ i in the inquiry range QR
≤p) and the dimension CE of the element CE in the formula Dmj (1 ≤ j ≤ k)
, And if a dimension matching the dimension Di in the query range QR exists in the dimensions Dm1 to Dmk of the element CE in the calculation formula, the member range NMRmj of the matching dimension Dmj.
To the member UMRi of the dimension Di of the data range UA, and the dimension matching the dimension Di in the query range QR is the element CE in the calculation formula.
If it does not exist in the dimensions Dm1 to Dmk of, the member QMRi of the dimension Di in the query range QR is set to the dimension of the data range UA.
Set to member UMRi of Di.

Even if the member range is not specified independently for each dimension or the current member is included in the member range specified in the numerical function, the processing becomes slightly more complicated than that described above. It is possible to extract the members to be applied, and it is possible to apply this embodiment.

Further, as in the case where the above-mentioned conditional expression is included, it may be difficult to collectively and accurately obtain the data to be used. In this case, a conventional method such as acquiring data to be used for each intersection from the back-end processing device 14 is
It may be used in combination with this embodiment.

In FIG. 16, the groups of the range to which the same calculation formula is applied are shown visually so as to be easy to understand, and the same English characters are put at the intersections of the same groups so that the range to be a group can be understood. Of course, what the client AP actually displays is generally data (including calculated values) corresponding to the set of members at the intersection.

For example, the intersection point in which C is entered is the moving average calculation of the number of sales which is a calculation member in the major dimension, and the other dimensions are members stored in the database not related to the calculation formula (hereinafter referred to as basic member). As a result, it is a group to which the formula for calculating the sales number moving average is applied.

Further, the intersection point with E is the moving average calculation of the number of sales which is the calculation member in the major dimension, the total of Shikoku which is the calculation member in the area dimension, and the other dimension is the group which is the basic member. This is a group to which the formula for calculating the moving average of sales and the formula for calculating Shikoku total are applied.

When performing this grouping, any method may be used as a method of determining which of the calculation formula of the sales number moving average calculation formula and the calculation formula of the Shikoku total is applied first. Assuming the formula for moving average sales calculation has a higher priority than the formula for total Shikoku, and the formula for moving average sales calculation is used, the same formula as in C is applied. , And C and E may be considered to be in the same group, but in general, in the case of E, the result of extracting the set of members identifying the data used in the formula for calculating the moving average of sales is Shikoku. It will include the total calculated members.
In that case, a set of members that identify the data used in the Shikoku total calculation formula is extracted by recursive processing or the like. As a result, C and E differ in the process of obtaining the set of members that identify the data to be used, and it is not possible to obtain them collectively by one procedure.

Generally, the process of obtaining the set of members for identifying the data to be used differs depending on the number and type of calculation members related to the intersection, and the process is divided into 8 groups of A to G and blanks. As a procedure, first, the members projected for each dimension are divided into a basic member group and each calculation member, and a group is formed by each combination for all dimensions of the query.

The number of groups can be calculated by the following formula. Here, NCMi is the number of calculation members projected from the dimension Di (1 ≦ i ≦ the total number of dimensions).

[0072]

[Equation 5] π (NCMi + 1) In the case of the example in FIG. 16, since there is one calculation member for each dimension projection and there are three dimensions, the following is obtained.

[0073]

## EQU00006 ## (1 + 1) * (1 + 1) * (1 + 1) = 8 FIG. 17 is a diagram showing an example of acquisition of data collected for each group at the time of execution of this embodiment. In FIG. 17, based on the example shown in FIG. 16, acquisition of data collected for each group at the time of execution is visually represented in an easy-to-understand manner. As in Figure 16,
This kind of display is not actually displayed on the client AP. In reality, the database system performs the process of acquiring the data identified by the set of members at each intersection.

At the time of execution, the data identified by the set of members at each intersection is acquired or calculated in an arbitrary order. Figure 1
In the case of 7, data is acquired or calculated in the order of arrows,
Assume that you have reached the intersection at the beginning of the arrow (1996, Tokyo, moving average of sales).

First, the group for the intersection is selected. This is a group indicated by C in the case of the example in FIG.

Since it is the evaluation of the first intersection in group C,
The data used in the range C enclosed by the two dotted ellipses is collectively acquired from the back-end processing device 14 in one communication. (There may be cases where it is not necessary to do it only once depending on the size of the communication buffer and the amount of data used for calculation, but even in that case, a drastic reduction in the number of times of communication can be expected). However, as described in the description of FIG. 16, when the data to be used cannot be collectively acquired in association with the conditional expression and the like, the data is acquired at each intersection. If not, the obtained result is stored in a buffer or the like.

In the following, at the time of evaluating the other 11 intersections of C, the data to be used can be taken out from the buffer, not from the backend processing unit 14. As a result, in the conventional case, at least 12 times communication is required for the group C, but only once. With respect to the whole, where there are 42 intersections and communication is required at least 42 times, the communication with the number of groups of 8 is sufficient, and the purpose of reducing the number of communication can be achieved.

In the case of the example in FIG. 17, the data used for each group is acquired, but at the time of analysis, the set of members for identifying the used data of a plurality of groups are merged and acquired collectively at the time of execution. Is also good. Further, the data of the calculation formula applied to the group may be collectively acquired within the range divided for each element to be used. This should be decided in relation to the size of the buffer that can be prepared at the time of execution.

When there is a dependency relationship between the calculation expressions of each group, the data of each group may be acquired in the order according to the dependency relationship. For example, if it is necessary to calculate the average before 1995 and the Shikoku total before calculating the moving average of sales, the data of the corresponding group may be acquired first.

The outline of the present embodiment has been described above. Next, the calculation use data range extraction processing unit 110 and the calculation evaluation processing unit 111 which characterize the present embodiment will be mainly described.

FIG. 3 is a diagram showing the processing contents of the calculation and use data range extraction processing unit 110 of this embodiment. As shown in FIG. 3, the calculation use data range extraction processing unit 110 receives the inquiry range information 101, which is the information of the set of members of the inquiry range, and calculates the set of information that identifies the data used for the calculation. The used data range information 102 is output.

The calculation use data range extraction processing unit 110 of this embodiment includes a calculation application grouping process 301, a group range conversion process 302, and a group calculation use data range extraction process 303.

The calculation application grouping processing 301 is shown in FIG.
16 is a process corresponding to the grouping of the ranges to which the same calculation formula is applied, which is shown in the outline related to, and the range conversion process 302 for each group is performed for each grouped range shown in the outline related to FIG. This is a process corresponding to extraction of a set of members that identify data used in a calculation formula.

Further, the group-based calculation use data range extraction process 303 recursively recursively when a calculation member is included in the set of members identifying the data to be used extracted for one calculation formula by the above process. This is a process of applying the calculation use data range extraction processing unit 110 to obtain a set of members for identifying the data to be used. It may be realized not by recursion processing but by iterative processing.

FIG. 4 shows the calculation evaluation processing section 111 of this embodiment.
It is a figure which shows the processing content of. As shown in FIG. 4, the calculation evaluation processing unit 111 includes a data member information 103 that is information of a set of members at each intersection of a query and a calculation use data range that is information of a set of members that identifies data used for calculation. The information 102 is input, and the calculated value 104, which is the value at each intersection of the inquiry, is output.

The calculation evaluation processing unit 111 of this embodiment includes a calculation applicable group range information selection process 401, a calculation use data acquisition process 402, and a calculation formula evaluation process 403.

Calculation applicable group range information selection processing 401
Is a process corresponding to the selection of a group for an intersection (a set of members) to be evaluated shown in the outline related to FIG. 17, and a calculation use data acquisition process 402 is a group shown in the outline related to FIG. This is a process corresponding to the collective acquisition of the data used in the calculation for the converted range.
The calculation formula evaluation process 403 is a process of performing a calculation using the data used for the calculation acquired by the above process to obtain the calculated value 104 for the intersection.

FIG. 5 shows the inquiry range information 101 of this embodiment.
It is a figure which shows the example of. Inquiry range information 1 as shown in FIG.
01 is the information of the set of members of the inquiry range, and is represented by the range of the inquired member for each dimension. For simplicity,
Consider only when the range of members is specified independently for each dimension.

In the inquiry range information 101 of this embodiment,
The range information is divided into dimension-based range information 501, and has a member ID list 502 indicating the range of members for each dimension. Even if the information is not in this format, the present invention can be applied without any problem as long as the information can acquire the range of the member to be inquired for each dimension.

FIG. 6 is a diagram showing an example of the calculation use data range information 102 of this embodiment. As shown in FIG. 6, the calculation use data range information 102 is information of a set of members for identifying data used for calculation.

Calculation use data range information 10 of this embodiment
In FIG. 2, the range information is divided into the calculation-applied group-specific range information 601 for each group formed in the calculation-applied grouping process 301, and the group-specific range conversion process 302 and the group-specific calculation use data range extraction process 303 perform calculation. Use data range information 604 corresponding to a set of members for identifying data to be used is created.

The group range information 602 is information indicating a set of member ranges from each dimension as a result of dividing the inquiry range information 101 into calculation applicable groups, and can be expressed in the same format as the inquiry range information 101. Group ID 60
3 is information for selecting a group corresponding to the set of members at the intersection according to the data member information 103 at the time of execution, and is set in the calculation use data range extraction processing unit 110 at the time of analysis.

It is to be noted that, even if it is not in this format, any information can be used as long as it can collectively pass a set of members for identifying data used for calculation to the calculation evaluation processing unit 111 at the time of execution. In the case of the example in FIG. 6, group range information 6 for each group
Although 02 is divided, the ranges of a plurality of groups may be merged so that the ranges can be collectively acquired at the time of execution.

FIG. 7 is a diagram showing an example of the group ID 603 of this embodiment. As shown in FIG. 7, the group ID 603 is
This is information for selecting the calculation application group range information 601 corresponding to the set of members at the intersection according to the data member information 103 at the time of execution.

As described in the outline relating to FIG. 16, the range information 601 by calculation application group is obtained by dividing the members projected for each dimension into a basic member group and each calculation member and combining them in all dimensions. Made Therefore, the range information 601 for each calculation applicable group can be identified by the set of IDs of the calculation members of each dimension. However, the dimension identified by the basic member group is identified by a value not used for the calculated member ID such as NULL.

The group ID 603 is the group range information 6
02, and the group range information 602 can be used as a substitute. If the calculation applicable group-specific range information 601 can be uniquely determined by comparing it with the set of intersection members according to the data member information 103, it may be used as the group ID 603.

FIG. 8 shows the used data range information 6 of this embodiment.
It is a figure which shows the example of 04. As shown in FIG. 8, in the used data range information 604, the calculation application group range information 601 is included.
In the calculation formula applied to, the use data range information 801 of the set of members for identifying the data to be used is created for each element using the multidimensional data.

The element range information of the used data range information 801 is created by converting the group range information 602 for each group in the group range conversion process 302.
This element range information is the range of members for each dimension,
It can be expressed in the same format as the inquiry range information 101.

In the case of a group including a plurality of calculated members, there is a high possibility that there are still calculated members in the element range information. In such a case, the element range information is queried in the group-specific calculation use data range extraction processing 303. The calculation use data range extraction processing unit 110 is applied as a range by recursive processing or repetitive processing to create calculation use data range information 802.

FIG. 9 shows the data member information 10 of this embodiment.
It is a figure which shows the example of 3. As shown in FIG. 9, the data member information 103 is information on the set of members at each intersection of the inquiry. As described in the outline related to FIG. 17, at the time of execution, the data identified by the set of members at each intersection is acquired or calculated in an arbitrary order. At this time, each intersection can be identified by a set of IDs of members of each dimension as shown in FIG.

FIG. 10 is a flow chart showing the processing procedure of the calculation application grouping processing 301 of this embodiment. As described in the outline related to FIG. 16, first, the ranges to which the same calculation formula is applied are grouped.

As described in the outline relating to FIG. 16, this is achieved by dividing the members projected for each dimension into a basic member group and calculation members (step 1001) and combining them in all dimensions (steps). 1002), groups can be created for each combination. The calculation-applied group range information 601 is created for each group (step 1003), and the information (group range information 602, group ID 603, etc.) necessary for the subsequent processing is created and set (steps 1004 and 1005).

FIG. 11 is a flow chart showing a processing procedure of processing for each calculation application group in the group range conversion processing 302 of the present embodiment. As described in the outline relating to FIG. 16, a set of members for identifying the data used in the calculation formula is extracted for each grouped range.

Since there may be a group associated with a plurality of calculation members, first, the calculation formula to be applied is uniquely determined (step 1101). The method of determination may be any method.

With respect to the element using each multidimensional data in the applied formula, a set of members for identifying the data to be used sequentially is extracted (step 1102).

In step 1103, step 1102
It is checked whether or not the data use element has been acquired in step S1, and if there is the acquired data use element, the process proceeds to step 1104.

Since it may be difficult to collectively obtain a set of members that identify the data to be used by a simple operation such as in the case of being related to a conditional expression, it is judged whether conversion processing is possible (step 1104), the conversion is not performed for the elements that are not possible, and at the time of execution, the data to be used is collectively given up from the back-end processing device 14 and the conventional method is used together.

If conversion processing is possible, step 1
Proceeding to step 105, as described in the outline relating to FIG. 16, the elements using the multidimensional data include the elements identified by the set of members, the current member, the numerical function and the like. A simple conversion process as shown in the outline of FIG. 16 is performed on these elements to collectively obtain a set of members for identifying the data to be used.

FIG. 12 is a flow chart showing the processing procedure of the group-based calculation use data range extraction processing 303 of this embodiment. In step 1201, the use data ranges for each group and each element extracted as a result of the range conversion process for each group 302 are sequentially acquired. Step 1202
Then, it is checked whether or not the used data range is acquired in step 1201, and if there is the acquired used data range, the process proceeds to step 1203.

In step 1203, it is checked whether or not the calculated member is still included in the set of members for identifying the data used in the acquired used data range. If the calculated member is included, the process proceeds to step 1204. In step 1204, the calculation and use data range extraction processing unit 110 is recursively applied to obtain a set of members for identifying the data to be used.

FIG. 13 is a flow chart showing the processing procedure of the calculation applicable group range information selection processing 401 of this embodiment. As mentioned in the overview related to Figure 17, at run time,
The data identified by the set of members at each intersection is acquired or calculated in any order. To calculate the data identified by the set of members at each intersection, first select a group for the intersection.

In step 1301, range information 601 for each calculation application group, which is group information, is sequentially acquired from the calculation use data range information 102, and step 1302 is executed.
Then, the group of the data member information 103 which is the information of the intersection and the obtained range information 601 for each calculation applicable group
Compare the ID 603. Then, in step 1303, the range information 601 for each calculation application group having the same composition of the calculation members is selected.

FIG. 14 is a flow chart showing the processing procedure of the calculation and use data acquisition processing 402 of this embodiment. As described in the outline related to FIG. 17, in each group, the data used in the range of the entire group at the time of evaluation of the first intersection is set from the backend processing device 14 to 1
Collected in a single communication, associated with a group, and stored in a buffer, etc. (It may be possible to do it only once depending on the size of the communication buffer and the amount of data used for calculation, but even in that case, the number of times of communication is greatly reduced Can be expected).

First, the range information 601 for each calculation application group selected in the calculation application group range information selection processing 401.
It is checked whether or not there is a buffer associated with the group identified by (step 1401 and step 1402).

If there is no related buffer in the processing of step 1402, the data of the range is multidimensionally calculated by the use data range information 604 under the calculation application group range information 601 when evaluating the first intersection of the group. The data is collectively obtained from the data stored in the database (step 1403) and stored in the buffer in association with the group (step 1404).

If there is a related buffer in the processing of step 1402, the data used for the buffer has already been acquired in the evaluation of the second and subsequent intersections of the group, so there is no need to go to the stored data, and the number of communication Can be reduced.

To associate a group with a buffer,
Group ID 603, range information 601 by calculation application group
An address that can uniquely identify the group may be used. In addition to the group ID 603, there is also a method of assigning a serial number ID to the calculation applicable group-specific range information 601.

In the example of FIG. 14, the data in the range of each used data range information 604 under the group is managed by the buffer associated with the group, but the used data range information 6
For each 04, it may be managed by a buffer associated with an element or the like in the calculation formula.

Also, a group ID is a group of IDs.
The data in the merged range may be managed in a buffer associated with the ID, and the data used for the calculation can be obtained in the grouped range, and if the association method can reduce the number of communications, It can be applied to this embodiment.

FIG. 15 shows the calculation formula evaluation processing 40 of this embodiment.
It is a flowchart which shows the processing procedure of 3. As shown in FIG. 15, in the calculation formula evaluation processing 403, calculation is performed using the data acquired in the calculation use data acquisition processing 402, and the calculated value 104 for the intersection is obtained.

In step 1501, the elements forming the calculation formula are sequentially acquired in accordance with the calculation rule. Except for parts that are unique to multidimensional, such as elements that require multidimensional data,
Calculations can be performed according to the general rules and methods used for evaluation of calculation formulas, such as four arithmetic operators, parentheses, and conditional expressions.

Relevant to the present embodiment is the evaluation part of an element that requires multidimensional data, and step 150
In 2, it is checked whether the evaluated element is an element identified by a set of members, a current member, an element that uses multidimensional data, such as a numerical function, and the number of data use elements that use multidimensional data In that case, the process proceeds to step 1503.

In step 1503, it may be difficult to collectively obtain a set of members for identifying data to be used by a simple operation, such as in the case of being related to a conditional expression. Therefore, it is determined whether conversion processing is possible. If it is a data use element capable of range conversion, the process proceeds to step 1504, and the data to be used is acquired from the buffer. If the element is a data use element whose range cannot be converted, step 1505
Proceed to and obtain the data to be used from the stored data of the multidimensional database using the conventional method.

In step 1506, the data which is obtained from the buffer or the stored data is used to evaluate the elements constituting the calculation formula. If the calculation formula is not completed, the process returns from step 1507 to step 1501 to proceed with the calculation. To go.

The processing described above is performed by the analysis processing unit 15 of the database server processing unit 213 shown as an example in FIG.
And the execution processing unit 16. However, the program is not limited to the one stored in the external storage device physically connected directly to the computer system as in the example of FIG. 2, and is read and written by a computer such as a magnetic disk device or a flexible disk device. It can be stored in another recording medium.

As described above, according to the multidimensional database processing system of this embodiment, the data used for the calculation is acquired for each group of the calculation formulas and the calculated value at each intersection of the inquiry is obtained. It is possible to speed up the inquiry process by reducing the number of communications when transferring the data to be transmitted.

[0127]

According to the present invention, since the data used for calculation is acquired for each group of calculation formulas and the calculated value at each intersection of the inquiry is obtained, the number of communications when transferring the data used for calculation is reduced. It is possible to speed up the inquiry processing.

[Brief description of drawings]

FIG. 1 is a diagram showing a schematic configuration of a parallel multidimensional database processing system of the present embodiment.

FIG. 2 is a diagram showing a schematic configuration of a front-end processing device 13 of this embodiment.

FIG. 3 is a calculation use data range extraction processing unit 1 of the present embodiment.
It is a figure which shows the processing content of 10.

FIG. 4 is a diagram showing processing contents of a calculation evaluation processing unit 111 of the present embodiment.

FIG. 5 is a diagram showing an example of inquiry range information 101 according to the present embodiment.

FIG. 6 is a diagram showing an example of calculation use data range information 102 of the present embodiment.

FIG. 7 is a diagram showing an example of a group ID 603 of this embodiment.

FIG. 8 is a diagram showing an example of used data range information 604 of the present embodiment.

FIG. 9 is a diagram showing an example of data member information 103 of the present embodiment.

FIG. 10 is a calculation application grouping process 301 of the present embodiment.
5 is a flowchart showing the processing procedure of step S1.

FIG. 11 is a range conversion process 302 for each group according to the present embodiment.
10 is a flowchart showing a processing procedure of processing performed for each calculation application group in FIG.

FIG. 12 is a flowchart showing a processing procedure of group-specific calculation and use data range extraction processing 303 of the present embodiment.

FIG. 13 is a flowchart showing a processing procedure of calculation applicable group range information selection processing 401 according to the present embodiment.

FIG. 14 is a calculation use data acquisition process 402 of the present embodiment.
5 is a flowchart showing the processing procedure of step S1.

FIG. 15 is a flowchart showing a processing procedure of a calculation formula evaluation processing 403 of this embodiment.

FIG. 16 is a diagram showing an example of inquiry contents from a client AP when a combination of inquired members includes a calculated member in the present embodiment.

FIG. 17 is a diagram showing an example of acquisition of data collected for each group at the time of execution of this embodiment.

[Explanation of symbols]

11 ... Client, 12 ... Parallel multi-dimensional database processing system, 13 ... Front-end processing device, 14 ... Back-end processing device, 15 ... Analysis processing unit, 16 ... Execution processing unit, 17 ... Storage device, 18 ... Network, 101 ...
Inquiry range information, 102 ... Calculation use data range information, 1
03 ... Data member information, 104 ... Calculated value, 105 ... Query sentence, 106 ... Code, 110 ... Calculation use data range extraction processing unit, 111 ... Calculation evaluation processing unit, 112 ... Query analysis processing unit, 113 ... Optimization processing unit , 114 ... Code generation processing unit, 115 ... Code distribution / execution request processing unit, 11
6 ... Data acquisition control processing unit, 117 ... Data acquisition processing unit, 201 ... Display, 202 ... Keyboard, 20
3 ... Central processing unit, 204 ... Flexible disk drive, 205 ... Main memory, 206 ... Magnetic disk unit,
207 ... Communication control device, 208 ... System bus, 209
… Flexible disk, 210… Network, 21
DESCRIPTION OF SYMBOLS 1 ... Operating system, 212 ... Work area, 213 ... Database server processing part, 214 ... Database management system control part, 301 ... Computation application grouping process, 302 ... Group range conversion process, 303 ...
Calculation-use data range extraction process by group, 401 ... Calculation-applied group range information selection process, 402 ... Calculation-use data acquisition process, 403 ... Calculation formula evaluation process, 501 ... Dimension range information by dimension, 502 ... Member ID list, 601 ... Calculation Range information by application group, 602 ... Group range information, 60
3 ... Group ID, 604 ... Used data range information, 801
... use data range information, 802 ... calculation use data range information.

Claims (5)

[Claims] 1. A multi-dimensional database processing method for transferring data of a multi-dimensional database stored in a plurality of back-end processing devices to a front-end processing device for processing, and query range information indicating a set of members of a query range. Enter the value, classify the set of members of the query range for each group to which the same calculation formula is applied, and set the calculation use data range information, which is the information of the set of members for identifying the data used for calculation, to the calculation formula. The step of extracting for each group, the data in the range indicated by the calculation use data range information is acquired for each group of the calculation formula, the data is evaluated by the calculation formula, and the calculation is the value at each intersection of the inquiry. And a step of obtaining a value. 2. When a calculated member is included in the set of classified members, the calculated member is decomposed into a calculation formula and a range of members to which the calculation formula is applied, and the calculation use data range information is extracted. The multidimensional database processing method according to claim 1, characterized in that. 3. The data of a plurality of groups are collectively acquired when the data of the range indicated by the calculation use data range information is acquired for each group of the calculation formula. 2. The multidimensional database processing method described in any one of 2. 4. When acquiring the data of the range indicated by the calculation use data range information for each group of calculation formulas, the data is acquired in an order according to the dependency relation between the calculation formulas. The multidimensional database processing method according to any one of claims 1 to 3. 5. In a multidimensional database processing system for transferring data of a multidimensional database stored in a plurality of back-end processing devices to a front-end processing device for processing, query range information indicating a set of members of a query range. Enter the value, classify the set of members of the query range for each group to which the same calculation formula is applied, and set the calculation use data range information, which is the information of the set of members for identifying the data used for calculation, to the calculation formula. A calculation use data range extraction processing unit that extracts for each group, and obtains data in the range indicated by the calculation use data range information for each group of the calculation formula, evaluates the data by the calculation formula, and inquires A multidimensional database processing system, comprising: a calculation evaluation processing unit that obtains a calculation value that is a value of an intersection.