JP2017027471A - Distribution database system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique capable of improving search speed in a distribution database system and surely acquiring a search result.SOLUTION: A distribution database system according to the invention is configured to comprise plural computers comprising plural databases for holding different records respectively, and issue queries in which a search order is different each other to the respective databases to the respective computers in parallel.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a distributed database system.

  A distributed database system is known as one implementation method of a database that efficiently processes large-scale data. A distributed database system generally attempts to increase data processing efficiency by distributing records to a plurality of host computers and operating the host computers in parallel.

  The following Patent Document 1 describes a distributed database management system that executes data registration / inquiry processing at high speed. In this document, by determining in advance which database management device has data to be accessed, performance degradation caused by access to all database management devices is avoided (0006 of the same document). ~ 0007).

  Non-Patent Documents 1 and 2 below describe examples of distributed databases that use a large number of relatively inexpensive server computers.

JP 2006-350741 A

“Getting started with Cassandra-let's experience NoSQL by moving your hands”, URL: http://gihyo.jp/dev/serial/01/cassandra/0001 (October 07, 2015) “3rd distributed database that brings out the power of the cloud”, URL: http://thinkit.co.jp/story/2010/06/17/1612 (acquired on 07/07/2015)

  In the distributed database described in each of the above documents, the response speed for the search request varies depending on how much load is applied to the computer holding the data to be searched, and is not necessarily It is considered that a response is not always obtained at high speed.

  As one implementation example of a distributed database, it is conceivable to construct a distributed database at a low cost by using a large number of computers that are relatively turned off, such as personal computers. In this case, the terminal cost can be kept low, but the computer holding the data to be searched may not be operating. Even in such a case, it is desirable to ensure that the search target data can be acquired as reliably as possible.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a technique capable of improving the search speed in a distributed database system and reliably acquiring search results.

  The distributed database system according to the present invention has a plurality of computers having a plurality of databases each holding different records, and issues queries having different search orders for the respective databases to each computer in parallel.

  According to the distributed database system of the present invention, a query is issued to each computer in parallel, so that the search speed can be improved by preferentially adopting the one with fast response. Furthermore, since each computer holds the same record, search results can be obtained from other computers even if one of the computers is down.

1 is a configuration diagram of a distributed database system 1000 according to Embodiment 1. FIG. It is a figure explaining the procedure in which the search computer 300 issues a search query. It is a figure explaining the procedure in which the search computer 300 processes the response with respect to a search query. FIG. 10 is a diagram illustrating a procedure for a search computer 300 to issue a search query in the second embodiment. It is a figure explaining another procedure in which search computer 300 issues a search query in Embodiment 2. FIG.

<Embodiment 1>
FIG. 1 is a configuration diagram of a distributed database system 1000 according to the first embodiment of the present invention. The distributed database system 1000 is a database system configured by a plurality of DB (DataBase) computers. In FIG. 1, the example which has the three DB computers 100a-100c was shown. Since each DB computer has the same configuration, the DB computer 100a will be mainly described below. Alphabetic subscripts are used to distinguish each computer and its components.

  The DB computer 100a includes a CPU (Central Processing Unit) 110a and a DBMS (DataBase Management System) 111a.

  The CPU 110a is an arithmetic device that executes the DBMS 111a. The DBMS 111a is a program for mounting a database. The DBMS 111a implements functions that the database should have, such as record operations and responses to search queries. In FIG. 1, the DBMS 111 a provides three databases DB 210 to 230. The databases 210 to 230 hold different records. The DBs 210 to 230 that each DB computer has are copies of the DBs 210 to 230 that other DB computers have.

  The distributed database system 1000 further includes a search computer 300. The search computer 300 includes a CPU 310 and a search program 311. The CPU 310 executes a search program 311. The search program 311 implements processing for issuing a search query to each database constituting the distributed database system 1000. The operation of the search program 311 will be described later.

  In the following description, for convenience of description, each program may be described as an operation subject. However, it is noted that the program is actually executed by an arithmetic unit such as a CPU provided in a computer on which the program is installed.

  FIG. 2 is a diagram illustrating a procedure in which the search computer 300 issues a search query. The user instructs the search computer 300 to issue a search query after specifying a search keyword, for example. The search program 311 issues a search query according to the following procedure according to the instruction.

(FIG. 2: Step (1): Search for DB Computer 100a)
The search program 311 makes a search query requesting that the databases included in the DB computer 100a be searched in the order of (i) DB 210, (ii) DB 220, (iii) DB 230 according to the specified search keyword. Issued against.

(FIG. 2: Step (1): Search for DB Computer 100b)
The search program 311 issues a similar search query to the DB computer 100b in parallel with the search query issued to the DB computer 100a. However, the order of searching each database is different from the search query for the DB computer 100a. Here, the search is performed in the order of (i) DB 220, (ii) DB 230, and (iii) DB 210.

(FIG. 2: Step (1): Search for DB Computer 100c)
The search program 311 issues a similar search query for the DB computer 100c in parallel with the search query issued for the DB computers 100a and 100b. However, the search order of each database is different from the search query for the DB computer 100a, and is different from the search query for the DB computer 100b. Here, the search is performed in the order of (i) DB 230, (ii) DB 210, and (iii) DB 220.

(Figure 2: Step (1): Supplement)
The search program 311 does not necessarily strictly need to issue a search query simultaneously to each DB computer. That is, a search query may be issued to another DB computer before a search query response to a certain DB computer is completed.

  FIG. 3 is a diagram illustrating a procedure in which the search computer 300 processes a response to the search query. The search program 311 processes a response to the search query according to the following procedure.

(FIG. 3: Step (2): Search result for DB 210)
Each DB computer transmits a response to the search query to the search computer 300. Here, first, a response to the search query for the DB 210 will be described. Since each DB computer processes a search query according to its own computation load situation, the time required to complete a search for each database differs. In general, however, the fastest response to the database that originally received the search query is likely. In the example shown in FIGS. 2 to 3, since the DB computer 100a first receives a search query for the DB 210 among the DB computers 100a to 100c, the DB computer 100a first transmits a response to the search query for the DB 210. Assumed.

(Figure 3: Step (3): Adopt the result from 100a)
The search program 311 adopts the first received as a (valid) response of the search query to the DB 210 as the search query response to the DB. Here, the search result received from the DB computer 100a is adopted as the search result for the DB 210.

(Figure 3: Step (3): Supplement)
In this step, when a search result cannot be obtained from the response acquired from the DB computer 100a, it is not handled as a valid response. For example, a response indicating that the connection to the DB 210 could not be made does not include a search result for the DB 210 and is not adopted as a response to the search query.

(FIG. 3: Step (4): Cancel search for DB 210)
The search program 311 cancels a search query issued to the DB computers 100b and 100c for the DB 210. This is because the search query result for the DB 210 has already been obtained, and there is no need to further search a database having the same record.

(Figure 3: Step (4): Supplement 1)
By canceling the search query for the DB computers 100b and 100c in this step, the computational load on these computers can be suppressed. However, if the calculation load is not a problem, the search program 311 simply discards the duplicate search result after receiving it instead of issuing another command for canceling the search query. From the viewpoint of output, the search query can be treated as substantially cancelled.

(Figure 3: Step (4): Supplement 2)
The means for canceling a search query that has already been issued differs depending on the database management system. For example, when issuing a search query, a method is conceivable in which the ID of the query is received from the DBMS in advance, and the ID is specified and a cancel instruction is transmitted again.

(FIG. 3: Steps (2) to (4): Supplement)
The search program 311 processes search queries for the DBs 220 to 230 in the same manner as described above. That is, the first response received as a response to the search query is adopted, and the previously transmitted search query is canceled (or the search result is discarded) except for the DB computer that transmitted the response.

<Embodiment 1: Summary>
The distributed database system 1000 according to the first embodiment is configured such that each DB computer has a copy of the DB 210 to 230 holding different records, and searches the DB 210 to 230 in a different order for each DB computer. Issued search queries in parallel. As a result, even if any of the DB computers 100a to 100c goes down, the search result can be obtained from other DB computers.

  The distributed database system 1000 according to the first embodiment employs, as a search result for the DB, the first received search query for the DB possessed by each DB computer, and uses the search query for the same DB possessed by the other DB computer. Cancel. Thereby, while obtaining a search result as early as possible, it can suppress imposing excessive search load with respect to each DB computer.

<Embodiment 2>
In the first embodiment, for example, when the DB computer 100a first transmits a search query response to the DB 210 for a predetermined number of times or for a predetermined time, the DB computer 100a first transmits a search query response to the DB 210 first. It is supposed to be. In such a case, imposing a processing load for searching the DB 210 on the DB computers 100b and 100c is not desirable from the viewpoint of processing efficiency.

  In such a case, it may be desirable not to transmit a query for searching the DB 210 to the DB computers 100b and 100c, or at least to advance the order of searching the DB 210 from the current state. In the second embodiment of the present invention, one example of the operation will be described.

  FIG. 4 is a diagram illustrating a procedure for the search computer 300 to issue a search query in the second embodiment. As a premise of FIG. 4, it is assumed that the DB computer 100a first transmits a search query response to the DB 210 for a predetermined number of times or for a predetermined time as described above. For example, it is assumed that the DB computer 100a first returns a search query response to the DB 210 continuously for 10 times or continuously for one hour.

  The search program 311 initially searches the DB 210 last in a search query issued to the DB computer 100b. However, considering that there is a high possibility that the DB computer 100a will respond first, it is considered that there is a high probability that even if a query that searches the DB 210 last is issued, it will be canceled later. Accordingly, the search program 311 advances the search order of the DB 210 in a search query issued to the DB computer 100b thereafter.

  In the example shown in FIG. 2, the DB computer 100b searches the DB 210 third, but in FIG. 4, the search order is changed so as to search second. Thereby, for example, when the load on the DB computer 100a increases and the response performance decreases, the possibility that the DB computer 100b first transmits a search result related to the DB 210 increases. Therefore, requesting the DB computer 100b to search the DB 210 does not necessarily impose a useless load, but is useful as a process for improving the search efficiency of the entire distributed database system 1000.

  Since the DB computer 100c is searching the DB 210 second than before, there is no need to change the search order for the DB 210.

  FIG. 5 is a diagram illustrating another procedure in which the search computer 300 issues a search query in the second embodiment. In FIG. 4, the DB computer 100b searches the DB 210 second, but if it is almost certain that the DB computer 100a can obtain a search result related to the DB 210, the DB computer 100b may not search the DB 210. . Therefore, in FIG. 5, the DB computer 100 b specifies to perform the search in the order of DB 220 to DB 230.

  There is also a possibility that the load on the DB computer 100a increases rapidly or goes down. Therefore, in the example shown in FIG. 5, the DB computer 100c performs the search in the same order as before.

  In the above example, the DB computer 100a first transmits a search result related to the DB 210. However, the search order of other DB computers and other DBs can be changed according to the same criteria.

<Embodiment 2: Summary>
In the distributed database system 1000 according to the second embodiment, when a specific DB computer first transmits a search result related to any DB, the order in which the other DB computer searches the DB is changed. Thereby, it is possible to reduce an excessive processing load for searching the DB.

<Modification of the present invention>
The present invention is not limited to the above-described embodiments, and includes various modifications. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of a certain embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of a certain embodiment. In addition, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

  In the above embodiment, the distributed database system 1000 has three DB computers each having three DBs, but the number of DB computers and the number of DBs are not limited to these. The present invention can be applied to any distributed database system including at least two DB computers and two DBs.

  In the above embodiment, each DB constituting the distributed database system 1000 may be a relational database or a key-value store database. Other types of databases (for example, object-oriented databases) may be used.

  Since each DB computer holds a copy of the same database, the present invention can be applied, so the types of DBMSs do not necessarily have to be the same. For example, the DB 210 may be a relational database, and the DB 220 may be a key / value store database.

  100a to 100c: DB computer, 300: search computer, 311: search program, 1000: distributed database system.

Claims (2)

A distributed database system composed of a plurality of databases each holding different records,
A first computer having a first database and having a second database holding records different from those held by the first database;
A second computer having a copy of each of the first and second databases;
A search computer that issues a search query to records held by the distributed database system;
Have
The search computer issues the search query to the first computer in a first order configured to search the second database after searching the first database;
The search computer is further configured to search for a copy of each of the first and second databases in an order different from the first order in parallel with issuing the search query to the first computer. Issued the search query in the second order
The search computer employs the first response received from the search query response to the first database or a copy thereof, and the search query for the first database or the copy transmitted to a source other than the transmission source. And the search query for the second database or its copy sent to other than the sender is adopted, and the first received response of the search query to the second database or its copy is adopted. Cancel the
Obtaining the search query results for the first and second databases by combining the search query response for the first database or a copy thereof and the search query response for the second database or a copy thereof. A distributed database system characterized by The search computer issues the search query to the second computer thereafter when the first response to the search query for the first database is received from the first computer for a predetermined number of times or for a predetermined time. 2. The distributed database system according to claim 1, wherein the order of searching for the copy of the second database is made earlier than the order in the second order, or the copy of the second database is not searched. .

Images