JPH05233390A - Data division management system for distributed data base - Google Patents

Abstract

PURPOSE:To generate a load division condition from the resource condition of each site in accordance with the division system inputted by a DBA. CONSTITUTION:Each site is provided with a computer resource management means 15 which stores and manages the computer resource condition, a division system selecting means 16 which selects the division system for data arrangement based on the computer resource condition by a user, a division condition generating means 17 which leads the division condition in accordance with the division system obtained from the division system selecting means 16 based on the computer resource condition obtained from the computer resource management means 15, and a division condition storage means 18, for which the data division condition led by the division condition generating mans 17 is stored, and the division system is only determined to efficiently distribute the load in accordance with the computer resource condition.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data division method applied when a load is distributed to a plurality of computer sites in a distributed database management system.

[0002]

2. Description of the Related Art In a distributed database system, there is an advantage that more sophisticated information can be obtained and a load of a computer can be distributed by cooperating and utilizing physically separated databases. .. By the method of dividing / arranging the data, it becomes possible to effectively distribute the load of the computer and make full use of the performance of the computer. An example of distributed arrangement of databases according to a conventional technique is shown. The data division is determined based on the index value and distributed based on the condition.

When the data includes information that depends on the site or the search method is determined, the divisional allocation method of the database can be determined so that the highly dependent data is allocated at each site.

For example, in a bank database or the like, data is arranged for an account based on the branch where the account is located. The division condition indicates a data allocation method at each site according to a branch (usually a branch code). In this case, if there is access to the data related to a certain account, it can be identified by the branch code of the branch of that account, and the site where the data is located can be known.

However, if the data has no site dependency, a database administrator (hereinafter referred to as DBA) determines the data division / arrangement conditions and stores them in a catalog or the like for management. The condition is also appropriately determined by the DBA using the index value and directly written in the catalog.

For example, existing data is arranged at three sites (A, B, C) mutually connected by a communication path as shown in FIG. The index most often used in the database is I (1 ≦ I ≦ 10 million). The DBA divides the data into equal parts based on the value of I, such as Site A: I = 1 to 33,300 Site B: I = 33,300 to 6,660,000 Site C: I = 6,660,000 to 10,000,000. Suppose you decide. The DBA registers this condition in the catalog and arranges the data so as to meet this condition.

Generally, when the number of sites is N, the DBA creates and stores a division condition based on the next index i so that the data is evenly arranged at each site. The data of the index i is stored in the mod (i / N) th site, and each time the input data is input, the site to be placed is calculated based on this division condition, and the data is stored in that site. Sent and stored. here,
An example of processing performance when data is equally divided at each site is shown below. In a distributed database consisting of three sites, assume that the main memory capacity of each computer is as follows. Site main memory amount Site A: 100 MB Site B: 10 MB Site C: 1000 MB Here, the processing speed increases as the main memory amount increases.

Therefore, if each of the above sites is subjected to the same process, the process performed at 10 msec at site A is
It will be performed at 100 msec at site B and 1 msec at site C.

Therefore, considering the case where access to data is uniform for each data, each site (A, B,
Since the processing amount of the database in C) is proportional to the data amount as described above, the processing load on the site B becomes large and efficient load distribution cannot be achieved.

[0010]

As described above, the conventional method has a drawback in that the DBA determines the data division conditions, which complicates the work of the DBA. Further, when the performance of the computer differs depending on the site, there is a problem that the conventional means for simply dividing the data into equal parts cannot efficiently distribute the load and use the computer.

[0011]

In the distributed database system according to the present invention, based on the computer resource management means for storing the computer resource status of all sites, and the computer resource status stored in this computer resource management means Partitioning method selecting means capable of selecting a partitioning method at the time of arranging data at a user level, and partitioning for guiding a partitioning condition according to the partitioning method obtained by the partitioning method selecting means from the computer resource status obtained from the computer resource managing means. The present invention is characterized by including condition generation means and division condition storage means for storing conditions for dividing the data introduced by the division condition generation means, and the system performs division / allocation management of data for load distribution according to the computer resource status. And

[0012]

In the above-described configuration of the present invention, the DBA inputs the partitioning method, and the system determines the partitioning condition by the partitioning condition generating means from the computer resource status of each site extracted from the computer resource managing means according to the partitioning method. It is generated and the division condition is stored in the division condition storage means.

As described above, since the division condition is generated from the computer resource status, the load distribution is balanced according to the computer resource status, and the performance of each site can be fully utilized.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a system configuration according to an embodiment of the present invention. In FIG. 1, 11 is a computer site, and 12 is a communication path connecting the computer sites to each other.

The computer site 11 is provided with a CPU 13 for controlling the system, a database 14, and each of the components 15 to 18 for implementing the present invention.

That is, in the computer site 11, computer resource management means 15 for storing and managing the computer resource status of all the computer sites, and data allocation based on the computer resource status obtained from this computer resource management means 15. A partitioning method selecting unit 16 for selecting a partitioning method at this time from the user, and a partitioning method that derives a partitioning condition according to the partitioning method obtained by the partitioning method selecting unit 16 from the computer resource status obtained by the computer resource managing unit 15. The condition generation means 17 and the division condition storage means 18 for storing the conditions for dividing the data, which are guided by the division condition generation means 17, are provided.

FIG. 2 shows an example of the configuration of the distributed database system of the present invention. Here, three computer sites (A, B, C) 20a, 20b, 20c are connected to the communication path 22 (of FIG. 1). (Corresponding to the communication path 12) and a configuration in which they are mutually connected by a communication network 21. Each computer site (A, B, C) 20a, 20
b and 20c respectively correspond to the computer site 11 shown in FIG. FIG. 3 is a flow chart for explaining the processing operation of the above embodiment. FIG. 4 is a diagram for explaining the flow of processing operations when load distribution is performed according to the size of the main memory in the above embodiment. The operation of the embodiment of the present invention will be described with reference to the drawings.
First, the flow of processing for generating division conditions will be described using the flowchart of FIG. This division condition generation process is
It is realized by the division condition generating means 17 under the control of the CPU 13. The computer resource management means 15 stores and manages the computer resource status of all computer sites.

In the division condition generation process, first, the user selects and inputs a division method that determines what kind of computer resource status value is used to divide the data by the division method selection means 16 (see FIG. 3 step S1).

When a partition method is selected here, the resource value which is the parameter of the selected partition method is obtained from the catalog information of the computer resource management means 15 (step S2 in FIG. 3).

Based on the resource value obtained from the computer resource managing means 15 and the division method obtained from the division method selecting means 16, the division condition generating means 17 generates a division condition (step S3 in FIG. 3). ). The division condition generated by the division condition generation means 17 is stored in the division condition storage means 18 (step S4 in FIG. 3). In accordance with the division condition stored in the division condition storage means 18, data division processing is performed to achieve load distribution.

Next, a method of generating a division condition will be described with reference to FIGS. As shown in FIG. 2, the distributed database used here includes a communication path 21 (corresponding to the communication path 12 in FIG. 1) and three computer sites (A, B, C) mutually connected by the communication network 21. I shall.

In a system that mainly performs database processing among the processing performed by a computer, a system having a larger main memory tends to have higher database processing performance. The main storage amount here refers to the size of the area such as the buffer or data cache used for database processing in the main storage amount of the computer. Therefore,
If the user selects to allocate all the data in proportion to the main memory as a partitioning method, the computer main memory 15 is the site main memory: Site A: 100 MB Site B: 10 MB Site C: 1000 MB. I understand. The ratio of the main memory amount for each site is A: B: C = 10: 1: 100, and the data is distributed at this ratio. Extract frequently used index I, and the value of I is 1 ≦ I ≦ 10
Using the above ratio, the division condition generation means 17 determines that the site A: I = 1 to 900,000 sites B: I = 900 to 1 million sites C: I = 1 to 10 million. In addition, the partitioning conditions for the placement that can distribute the load of database processing are determined. This division condition is stored in the division condition storage means 18. Since the amount of data and the amount of database processing are proportional, the ratio of the amount of database processing for each site is also A: B: C = 10: 1: 100. At each site, the processing time when a certain processing is performed is inversely proportional to the main memory amount, so that A: B: C = 10: 100: 1.

If 100 processings are performed on site A at an average of 10 msec within a certain period of time, 10 processings are performed on site B at an average of 100 msec, and 10 processings are performed at site C at 10 msec.
The processing of 00 cases will be performed in an average of 1 msec. Therefore, the time required for database processing at each site is about 1 second, and it can be said that the load is efficiently distributed.

There are various other means for generating the division condition from the resource status of the computer, including the following methods, and the DBA inputs these methods according to the characteristics of the data. (1). Predict the places that are likely to become hotspots (where they are accessed more frequently than other data) and assign them to sites with large main memory. By processing hotspot data at high speed, the load is distributed and the risk of system down is reduced.

For example, in the distributed database having the configuration shown in FIG. 2, it is assumed that the index I sets the following division conditions and the number of accesses within a certain time is as follows. Site A: I = 2 to 5 million: about 2000 times Site B: I = 5 to 10 million: about 200 times Site C: I = 1 to 2 million: about 20000 times In this case, the ratio of access times is A : B: C = 10: 1: 100, which is the same as the ratio of the main memory amount, so it can be said that the load is distributed. (2). Main memory, external memory, or semiconductor memory,
The storage device is classified into the other storage devices, and the data distribution is determined based on the ratio. It is assumed that the main memory amount is the same as the above example, and the external memory amount is as follows. Site external storage amount Site A: 20 GB Site B: 10 GB Site C: 50 GB Therefore, the ratio of external storage amount is A: B: C = 2: 1: 5 When the data is divided according to the ratio of main storage amount to external storage amount Then, from the respective ratios, the data ratio is set as A: B: C = 5: 1: 20. At that time, based on the index I, division conditions such as site A: I = 1 to 2 million sites B: I = 2 million to 2.5 million sites C: I = 2.5 million to 10 million can be generated. (3). Also consider the CPU and communication performance. The amount of data that can be processed also depends on the CPU and communication performance. The load can be distributed by considering these.
For example, the CPU performance ratio of each site is taken, and the data is distributed according to the ratio. It is assumed that the MIPS value of each site is as follows. Site MIPS value Site A: 30 MIPS Site B: 10 MIPS Site C: 60 MIPS This ratio is A: B: C = 3: 1: 6, and the data is distributed based on this ratio, and Site A: I = 1 to 3 million. Site B: I = 3,000,000 to 4,000,000 Site C: I = 4,000,000 to 10,000,000 can be generated. Similarly, by distributing the data in accordance with the communication performance ratio, it is possible to perform load distribution of the machine which is likely to become a communication bottleneck.

As described above, according to the embodiment of the present invention,
In a distributed database system in which data is distributed and distributed to multiple sites connected by communication channels, the DBA determines the partitioning method, not the method relying only on the DBA's experience and intuition, but on the basis of computer resource status. It is possible to efficiently distribute the load by performing division / arrangement management of.

[0027]

As described in detail above, according to the present invention, computer resource management means for storing the computer resource status of all the sites constituting the distributed database system,
Partitioning method selecting means capable of selecting at a user level a partitioning method when arranging data based on the computer resource status stored in the computer resource managing means, and the above-mentioned partitioning based on the computer resource status obtained from the computer resource managing means. The system is provided with a division condition generating means for guiding a division condition according to the division method obtained by the method selecting means, and a division condition storing means for storing a condition for dividing the data guided by the division condition generating means, and the system is provided with a computer resource status. In the distributed database system in which data is distributed / distributed to a plurality of sites connected by communication paths, the DBA simply determines the partitioning method to configure the partitioning / arrangement of data for load distribution according to Therefore, the load can be efficiently distributed.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration example of a distributed database system targeted by the present invention.

FIG. 3 is a flowchart illustrating a processing operation of the above embodiment.

FIG. 4 is a diagram for explaining the flow of processing when load distribution is performed according to the size of main memory in the above embodiment.

[Explanation of symbols]

11 ... Computer site, 12, 22 ... Communication path, 13 ... CP
U, 14 ... Database, 15 ... Computer resource management means, 16 ... Dividing method selecting means, 17 ... Dividing condition generating means, 18 ... Dividing condition storing means, 20a ... Computer site A, 20b ... Computer site B, 20c ... Computer site C, 21 ... Communication network.

Claims (1)

[Claims] 1. A division condition storage means, which is constructed on a plurality of computer sites mutually connected by a communication path, and stores a condition for dividing data, is provided according to the conditions stored in the division condition storage means. In a distributed database system in which data is divided and allocated to each computer site, computer resource management means for storing the computer resource status of each computer site, and data based on the computer resource status stored in the computer resource management means Partitioning method selecting means for selecting a partitioning method when arranging; partitioning condition generating means for deriving a partitioning condition according to the partitioning method obtained by the partitioning method selecting means from the computer resource status stored in the computer resource managing means; And generate partitioning conditions from the computer resource status according to the selected partitioning method. A distributed data management method for distributed databases.