WO2012081131A1

WO2012081131A1 - Key-value index store type database system

Info

Publication number: WO2012081131A1
Application number: PCT/JP2010/073265
Authority: WO
Inventors: 健人吉田; 裕介大久保; 圭隆坂本
Original assignee: 株式会社東雲
Priority date: 2010-12-16
Filing date: 2010-12-16
Publication date: 2012-06-21

Abstract

[Problem] To realize a computer system with the high operability and adding functions of a RDBMS, simplify and clarify the specifications of the complex client-server software application, and provide an efficient computer system development environment and maintenance environment in client-server computer system database management, while ensuring the high throughput, scalability, and distributed processing function of a KVS. [Solution] Provided is a key-value index store type database system in which, during communication between a client and a server according to the KVS method, key information is included in a value and, when storing data, pre-defined elements in a value are indexed on the server end, thereby providing high operability and completing the system only through simple communications.

Description

Key-value index store database system

The present technology relates to a client-server type computer system in which data required by a client computer is supplied by a server computer, and more particularly to a method for retaining data in a database management system of the server computer.

Conventionally, RDBMS (Relational DataBase Management System) and KVS (Key Value Store) exist as main methods for handling large-scale data aggregates (databases) of client-server type computer systems.

RDBMS is a method for managing database information, which is called a relational database, associated with each other using a language called SQL. In a relational database, data is represented as a set (record) of a plurality of fields, and the set of records is represented by a table called a table. Each record uses key data such as an ID number and a name, and data can be combined and extracted relatively easily in the SQL language. This is a widely used database.

On the other hand, KVS is a simple database expressed only with a key and a value (value). When a key is given, the value associated with the key can be acquired and stored. For this reason, it is difficult to express a complicated data structure, but the processing is fast because of the simple data structure. Since distributed processing is also possible, it is possible to ensure extremely high throughput and scalability as compared to SQL of a relational database. However, in KVS, it is difficult to perform simple search from value, so it is necessary to perform data analysis of value represented by Map-Reduce simultaneously or frequently and maintain index information separately from the database itself. When is implemented, the system configuration becomes complicated. In addition, it is difficult to maintain the immediacy of data.

US7496589 Republication 07-063945

The problem to be solved by the present invention is a computer having a high KVS throughput and scalability and a distributed processing function in a database management of a client-server computer system, and a high operability and additional function of an RDBMS.・ The system is to be realized. Furthermore, the specification of a complex client / server type software application is clarified to provide an efficient computer system development environment and maintenance environment.

In the database management of the client server type computer system of the present invention, means for converting a collection (object) of information existing in the main memory of the client computer into text information and sequentially storing it in the server computer Take.

At this time, the KVS system is constructed in such a manner that the server computer adds key information for individually recognizing the data and connects the key and the object. In this method, the server does not need to recognize the contents of the object, so there is no need to configure a field for storing individual data in the object, and a text object can be stored in one field. Therefore, the structure of the database is simplified. In other words, a system is constructed in which the server side does not recognize the data contents, but recognizes the data contents only on the client side.

At this time, as shown in claim 1 of the present invention, by including the key generated on the server side by the object itself, there is some overhead, but the object itself can be self-contained. In other words, data sent to the server without a key is treated as new data and inserted into the database, and data sent with a key is stored as existing data read from the server. Updates a record with the same key. At this time, the key generated on the server side is also stored in the primary key field defined as another field of the record.

When the stored data is read from the server side, since the KVS is constructed, it is possible to obtain object information including the key by referring to the primary key.
Here, as a problem of KVS, it is necessary to consider a method in the case of selectively extracting a specific record group. In order to realize this function, some items of a predetermined object can be transferred to the field of the record. That is, data predetermined by the system is extracted from the object at the time of data writing and written in a predetermined field. This state is the state of the database table in claim 2. Furthermore, if an index is provided for a field defined as shown in claim 3, a high-speed search is possible.

As described above, the client server type computer system of the present invention has the database table structure described in claims 1 to 3 in order to directly communicate the object information held by the client with the server in the database management. Have realized its function.

In the client server type computer system of the present invention, information existing in the main memory of the client computer is communicated directly with the server as an object, and there is no need to define its structure on the server side. This is possible and has the effect of reducing the load of database management on the server side.

In the conventional RDBMS, the database table is a two-dimensional table composed of columns and rows, and basically there are physical length and type restrictions for each field, and an array or multi-dimensional array is stored. Although it was difficult, if the database management of the present invention is used, the object information is stored as it is, so there is no particular limitation. For this reason, in the conventional RDBMS, it is necessary to join (JOIN) a plurality of tables in order to generate the object structure required by the client, and there is an overhead in data generation. There is no limit. In addition, when a new data item is added to an object, only the definition on the client side is modified and the server side is not affected at all.

As a result, the data communication efficiency between the client and the server is dramatically improved, and in particular, the throughput when acquiring data can be expected to be improved by about 10 times. In addition, changes in specifications have little effect on the database structure, so specifications can be easily added and maintainability is excellent.
Furthermore, in the client server type computer system of the present invention, since the usage amount of the CPU and the main memory on the server side is limited, an efficient server operation can be realized.

It is a conceptual diagram of the computer system using this invention. It is an example of the access of the database by this invention. 1 is an automatic database construction system using a database according to the present invention.

In the client-server type computer system, the best mode for carrying out the present invention is that the data communication speed between the client and server is sufficiently high, and the client computer has sufficient processing capability for the server computer. This is the case. Recent improvements in network technology and computer performance sufficiently satisfy this situation.

Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a conceptual diagram of a client server type computer system using the present invention. Reference numeral 1 denotes a database table according to the present invention, which stores data in KVS. Reference numeral 2 denotes a server computer, which is equipped with the database system indicated by reference numeral 1. Reference numeral 3 denotes a hash table developed on the main memory of the client, and reference numeral 4 denotes a client computer. Here, the hash table indicated by reference numeral 3 has a reference relationship indicated by reference numeral 5. Each hash table is paired with the corresponding database table on the server, and is connected by a network connection denoted by reference numeral 6 to read and write data necessary and sufficient for operations on the client computer. It can be done synchronously or asynchronously.

At this time, the communication between the client computer and the server computer has a very simple configuration such as INDEX (key acquisition), GET (read), PUT (write), and SELECT (selection), and can easily realize high throughput. . In the client computer, once acquired data is stored in a hash table using the key, and the data once called is reused. Especially for tables with many references, it is possible to maintain the data structure without increasing the amount of memory used.

In this embodiment, the client system operates with JavaScript, and the server side operates with PHP. In this case, the client performs AJAX (abbreviation of Asynchronous Javascript XML) communication with the server. Object data developed in the main memory of the client is converted into text efficiently by converting it into JSON (Java Script Object Notation). The text information is transferred to the server by AJAX communication. The server generates a centralized key, adds it to the object, stores it in the database, and converts the object including the obtained key into text and passes it to the client. The client re-objects the received data through the EVAL () function and expands it in the main memory via a hash table. By repeating this, it is possible to create the same situation as if the data having the reference structure is expanded on the main memory of the client computer.

FIG. 2 is an example of database access according to the present invention. Reference numerals 7, 9, and 11 indicate requests (requests) from the client computer and their responses (responses) divided by arrows (→), and

reference numerals

8, 10, and 12 indicate the database on the server side corresponding thereto. Indicates the contents of the table.

Reference numerals 7 and 8 indicate basic KVS access. It is assumed that _KEY_, which is the name of the field, is a field for storing a primary key that guarantees unity, and _VALUE_ is a field for storing a value (value), and this name is already defined as a system. Three lines beginning with PUT of reference numeral 7 indicate writing, and three different new data are inserted into the database. A unique key indicated by _KEY_ is assigned before being inserted into the database, and the value is also added to the value. The response including PUT returns the value including the key. In the last three rows starting with GET of reference numeral 7, the value is read using the value of _KEY_. This example is a very simple example, but in practice, it is possible to have complex information such as an array or binary data as an object. At this time, the database table has the structure claimed in claim 1 of the present invention.

Next, reference numerals 9 and 10 are examples in which “name” is added as an index to a simple KVS. In this case, it is assumed that “name” is an index for the corresponding table. The three lines starting with PUT of reference numeral 9 are the same as the request of reference numeral 7, and the response is also the same. However, “name” extracted from the value is added as an index on the database. As a result, the SEL request on the fourth line of reference numeral 9 enables selection of records whose “name” starts with “A”, and in this example, two records are acquired. In other words, even if the function of KVS is expanded by having an index other than the key and value, if the procedure that “name” is an index is processed in advance, the client side request and response are particularly There is no change. At this time, the database table has the structure claimed in

claims

2 and 3 of the present invention. By arranging a plurality of indexes as necessary, a function such as a complicated selection rearrangement is realized.

Further,

reference numerals

11 and 12 are examples in which additional database access is made to reference numerals 9 and 10. The first two lines starting with PUT of reference numeral 11 are examples of record update. Since _KEY_ exists in data sent from the client side to the server, the record in the database is overwritten. In the specification of this embodiment, since PUT is performed by adding an item to existing _VALUE_, the elements “prop1” and “prop2” in the existing object are not deleted. Since the third row does not have _KEY_, it is treated as new data and inserted into the database.

FIG. 3 is a schematic diagram of a database automatic generation system using the present invention. By using the present invention, database access becomes very simple, and everything including table generation can be automated. First, an abstract database access routine (PDO) 14 is constructed for the database 13 according to the present invention, and CREATE (table generation), DROP (table deletion), INDEX (key acquisition), GET Implements basic functions such as (data reading), PUT (data writing), SEL (data selection acquisition), and DEL (data deletion). At this time, the index element and the key and value field names in each table are described in the definition file 17. On the client side, an AJAX interface 15 is constructed on the client side for the PDO 14, and a Javascript program 16 can be executed. In this way, all AJAX-type programs can be operated according to instructions from the client side, and by describing the detailed information of the elements of the objects of each table in the definition file 17, the database An automatic generation system is constructed. Furthermore, if an access routine indicated by reference numeral 19 is also prepared for the PHP program indicated by reference numeral 18 that operates as a batch system on the server side, the definition file indicated by reference numeral 17 is shared so that the client-side AJAX system and the server-side batch can be shared. The system can operate in an equivalent environment.

The present invention can be used for a wide range of applications such as business management software such as sales management, financial accounting, and payroll personnel, which conventionally used RDBMS. Compared with a system using conventional RDBMS, Execution speed can be improved while maintaining maintenance performance. Further, it can be used for all or part of a general client-server type computer system or a WEB-based system represented by ASP (Application Service Provider), and its application range is wide.

A KVS database is easy to encapsulate and a discrete database can be constructed. A database with high update capacity and a large capacity can be installed on a cloud computer, and a database containing highly confidential and important information can be installed on an in-house server so that the system can be operated discretely. This solves a problem that is difficult for a conventional RDBMS that requires data combination (JOIN), and is an invention that can spur the application of cloud computing to applications in the future.

1 Database table 2 of the present invention A server computer, which is equipped with a database system 1.
3 Hash table expanded on the main memory of the client 4 Client computer 5 Reference relationship of the hash table of code 3 6 Network connection between the hash table and the corresponding database on the server 7 Request from the client computer and its response ①
8 Indicates the state of the database on the server side for reference numeral 7.
9 Requests and responses from client computers ②
10 Indicates the state of the database on the server side for symbol 9.
11 Request from client computer and its response ③
12 Indicates the state of the database on the server side for reference numeral 11.
13 Database and database table having the structure of the present invention 14 Abstract Access routine to database of the present invention 15 AJAX interface for client computer 16 Program 17 operating in client computer described in JavaScript 17 Each database table Definition file describing index information 18 PHP program that operates as a batch on server computer 19 I / O access routine of server computer

Claims

In a database table composed of records having a plurality of fields, each record is composed of at least two or more fields A and B, and field A is composed of at least one key and a reference value corresponding thereto. A database table described in a value type reference structure and having the information of field B, which is the primary key of the previous record, in the reference structure of field A described above.
In a database table composed of records having a plurality of fields, each record is composed of at least three fields A, B, and C, and one field A includes at least one key and its reference value. The information of the field B that is described in the key-value type reference structure composed of the above and is the primary key of the previous record, and the information of at least one field C group other than the fields A and B is described above. A database table characterized by having a reference structure of field A.
In a database table composed of records having a plurality of fields, each record is composed of at least three fields A, B, and C, and field A is composed of at least one key and its reference value. The information of field B, which is described in the key-value type reference structure and is the primary key of the previous record, and the information of field C group having at least one index information other than field A and field B, is described above. A database table characterized by having a reference structure of field A.
In a database management system having a plurality of database tables, the database table of claim 2 or claim 3 is included, and when the record is written, the information of the field B as the primary key and the field A from the information of the field A described above are written. -The information of the entire record is constructed by extracting the information of at least one field C group other than the field B, and the information of the field A by the information of the field C group having index information when selecting the record A database management system characterized by selectively acquiring.
5. A client server type computer system in which a server computer supplies data required by a client computer, wherein the server computer has the database management system according to claim 4, and Communication with the server computer is carried out by (1) transmitting the field A information of claim 3 when writing, and (2) transmitting the field B information of claim 3 when reading. (3) In the case of selective reading, it is performed by transmitting information necessary for selecting the field C group of claim 3.
6. In the client server type computer system according to claim 5, the value information for the field A prior to claim 3 is sent to the server when a data write request is made from the client computer. If there is no primary key information to be written in the previous field B, the server computer generates an exclusive key and adds it to the value information. A client-server computer system characterized by writing the information.