JP2006350741A - Database management device, distributed database management system, distributed database management method, and distributed database management program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a database management device, a distributed database management system, a distributed database management method, and a distributed database management program by which a data registration processing and a data retrieval processing can perform at high speed. <P>SOLUTION: The database management devices 2 is database devices which combine horizontal and vertical distributions to distributedly arrange data, and each database management device 2 includes; a hash function operation part 15e for returning a hash value resulting from combining an ID of a database management device where data specified by a key value and a column name is stored or should be stored, and a fractional record ID being a part of data, in response to input of the key value and the column name designating the data; and a distribution function operation part 15f which returns one or more database management device IDs where data designated by a column name is stored, in response to input of the column name. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a database management apparatus, a distributed database management system, a distributed database management method, and a distributed database management program in which a plurality of databases are distributed.

  When realizing a distributed database system, it is important how the definition information of the database constituting the distributed database is arranged on the site. The distributed database processing method differs depending on the difference in the arrangement information.

As a conventional distributed database technique, a database management apparatus using a record format based on a relational data model or an object-oriented data model is known (see Patent Document 1). In addition, the data is divided from the viewpoint of horizontal / vertical distribution, etc., and the global server that supervises the distributed database management device group receives the registered data and performs processing such as registering the data in the distributed database management device group A database system (centralized global DD / D management system) is known (see Patent Document 2). In the distributed global DD / D management system described further in Patent Document 2, each database management device that is distributed without using a global server autonomously processes a request from an application program.
Raghu Ramakrishnan, Johannes Gehrke, “Database Management Systems”, Second Edition, GcGraw-Hill, “record formats”, P221-224 Japanese Patent No. 002969627

  However, the database management apparatus disclosed in Patent Document 1 is based on a relational data model or an object-oriented data model, and when registering fragmented records, the records to be updated must be searched and updated. Therefore, there is a problem in that it takes time for processing. This has a problem that data cannot be written at high speed particularly in an application in which a large amount of fragment data such as sensed data is registered.

  Furthermore, in the distributed database system based on the centralized global DD / D management system disclosed in Patent Document 2, the data is stored in a database management apparatus group that divides and distributes data from the viewpoint of horizontal / vertical distribution as a method of storing data. However, when executing a search, the global server must accept the inquiry and execute the inquiry to the distributed database management apparatus group. Therefore, there is a problem that the global server becomes a performance bottleneck as the number of distributed servers increases. Further, in the distributed database system based on the distributed global DD / D management method, the application program freely designates the database management device to be accessed regardless of the load status of the database management device. There was a problem that performance could deteriorate without being able to disperse.

  In addition to this, in a distributed database device that distributes the database by combining horizontal and vertical distribution, inquiries are made for each individual fragment record included in the fragment record group obtained as an intermediate result of a query specifying a key value or query processing. When processing continues, the database is distributed according to the specified column value range (described as “range partitioning” in Patent Document 1) and column values (partitioning by hashing) in the normal horizontal distribution method. When processing a condition for a column that is not specified in the query, it is impossible to determine in which database management device the data to be accessed is distributed, so access to all database management devices occurs. There is a problem that the performance deteriorates. It was.

  In view of the above problems, the present invention quickly performs fragment record registration processing so as not to create a free storage area on the disk, and the global server becomes a bottleneck in performance as the number of distributed database management devices increases. Database management device that can quickly determine which database management device has data to be stored / accessed in data registration / inquiry processing, and can execute data registration / inquiry processing at high speed An object of the present invention is to provide a distributed database management system, a distributed database management method, and a distributed database management program.

  In order to achieve the above object, the first feature of the present invention is [a] a database device that distributes and arranges data by combining horizontal and vertical distributions, by inputting key values and column names for specifying data. A hash function calculation unit (15e) that returns a hash value in which data specified by a key value and a column name is stored or combined with a database management device ID to be stored and a fragment record ID that is a part of the data; , [B] A database device including a distributed function calculation unit (15f) that returns one or more database management device IDs in which data specified by a column name is stored by inputting a column name. To do.

  Furthermore, the first feature of the present invention is that [c] a data fragment registration unit (15g) for registering fragment records as data, [d] a database unit (3) for storing the registered fragment records by a write-once method, [ E) A record integration unit (15k) that integrates fragment records stored in the database unit (3) to form a record may be further added.

  A second feature of the present invention is [a] a distributed database management system for distributing and allocating data by combining horizontal and vertical distribution, specified by key value and column name by inputting the key value and column name. Hash function calculation unit (15e) that returns a hash value obtained by combining a database management device ID in which stored data is to be stored or a fragment record ID that is a part of the data, and a column name by inputting a column name A plurality of database devices having a distributed function calculation unit (15f) that returns one or more database management device IDs in which data specified by the database is stored; and [b] a load for collecting load information of the plurality of database management devices By statistically processing the load information in the information collection unit (25a) and load information collection unit (25a), the most A global server (1) having a low-load database management device and an access right management unit (25b) for determining an access right including an access period to the database management device, and [c] access to a plurality of database devices The gist of the invention is that it is a distributed database management system that is permitted based on the access right determined by the global server (1).

  Further, the second feature of the present invention is [D] When registering a fragment record in a database management device to register a fragment record among a plurality of database management devices, a key value and a column name held by the fragment record are input. A data registration processing unit that specifies an ID of a database management device in which data is to be stored using a hash function, transmits fragment data to the identified database management device, and [e] a plurality of database managements When searching for fragment records in the device, one or more column names described in the query statement for the search are specified, and the data necessary for the search is stored using a distributed function. A part of an inquiry statement that identifies one or more database management device IDs and that can be executed by the identified at least one database management device The first partial search unit (15h) that transmits and executes the search, [f] Necessary for the search using the hash function based on the key value and the column name of the data specified from the intermediate result of executing the search A second partial search unit (15i) that specifies the ID of the database management device that stores the data to be transmitted, transmits the partial query that can be executed by the specified database management device, and executes the search. Also good.

  A third feature of the present invention is [a] a distributed database management method for integrated management of a plurality of database management devices for distributing and arranging data by combining horizontal and vertical distributions, and a load on the plurality of database management devices. A step of collecting information by the load information collecting unit (25a), and [b] statistically processing the load information in the load information collecting unit (25a), so that the database management apparatus having the lowest load and its database management A step in which the access right management unit (25b) determines an access right including an access period to the device, [c] a step of permitting access to a plurality of database devices based on the access right, and [d] data distribution. When access is made to multiple database management devices and key values and column names are entered, the key values and column names are A step in which the hash function calculation unit (15e) returns a hash value obtained by combining the database management device ID in which the specified data is stored or to be stored and the fragment record ID that is a part of the data; The gist of the present invention is a distributed database management method including a step in which a distributed function calculation unit (15f) returns one or more database management device IDs in which data specified by a column name is stored by inputting a column name. .

  A fourth feature of the present invention is that [a] a computer for distributing and arranging data by combining horizontal and vertical distributions, load information of a plurality of database management devices for distributing and arranging data is collected by a load information collecting unit ( The data collected by the command 25a) and the load information in the load information collection unit 25a are statistically processed, thereby including the database management device with the lowest load and the access period to the database management device. An instruction for determining the access right by the access right management unit (25b), [c] an instruction for permitting access to a plurality of database devices based on the access right, and [d] a plurality of database management apparatuses for distributing and holding data When a key value and column name are entered, the data specified by the key value and column name is stored. Alternatively, the column name is determined by inputting a command in which the hash function calculation unit (15e) returns a hash value obtained by combining a database management device ID to be stored and a fragment record ID which is a part of data, and [e] column name. The gist of the present invention is a distributed database management program that executes an instruction to return one or more database management device IDs in which specified data is stored by the distributed function calculation unit (15f).

  According to the database management device, distributed database management system, distributed database management method, and distributed database management program of the present invention, the application program has the access right issued from the global server in advance. The load concentration on the server can be avoided. Further, since the global server issues an access right based on the load information of all the database management apparatuses, the load on the database management apparatus can be distributed.

  When registering data, a database management apparatus in which the data is to be stored is specified at high speed using a hash function, and the data is written into the storage area by the appending method, so that data registration processing can be performed at high speed.

  At the time of query processing, it is possible to specify at high speed a database management apparatus that stores data necessary for processing a query using a hash function / distribution function.

  Hereinafter, a database management device, a distributed database management system, a distributed database management method, and a distributed database management program according to embodiments of the present invention will be described. In addition, the apparatus, method, etc. which are used in embodiment of this invention are examples, and of course this invention is not limited to these.

(Embodiment)
(Distributed database management system)
A distributed database management system 100 according to an embodiment of the present invention is a system for distributing a database by combining horizontal and vertical distribution. As shown in FIG. A plurality of database management devices 2 including the database 2, a client terminal 4, and the like are included.

  The global server 1 manages the access status of the distributed database management device group 5 including a plurality of database management devices 2, and the database accessed by the application program according to the access status so as not to put a load on the specific database management device 2. The management device 2 is specified and its access period is set. The client terminal 4 is a terminal used by the user, and includes an application program for performing data registration processing and search processing for the distributed database management device group 5. The client terminal 4 is an ordinary personal computer or the like that includes an input device, an output device, a communication device, a main storage device, a CPU, and the like.

(Database management device)
As shown in FIG. 2, the database management device 2 includes an input device 11, an output device 12, a communication control device 13, a communication interface 14, a central control device (hereinafter referred to as CPU) 15, a main storage device 16, and a database unit 3. A hash table storage unit 18, a function storage unit 19, a cache unit 20, and the like.

  The input device 11 is a keyboard, a mouse, an interface for receiving an input signal, or the like. It may be input via an external storage device such as a floppy (registered trademark) disk or hard disk. The output device 12 is a device for outputting a processing result or the like, and specifically indicates a printer, a liquid crystal display, a CRT display, or the like. The communication control device 13 generates a control signal for transmitting / receiving data to / from the global server 1, another database management device 2, and the client terminal 4. The communication interface 14 is a device for transmitting / receiving data to / from the global server 1, another database management device 2, and the client terminal 4.

  The CPU 15 performs arithmetic processing such as processes and instructions. The main storage device 16 temporarily stores a program describing a processing procedure and data to be processed as a main memory, and delivers a machine instruction and data of the program according to a request from the CPU 15. Data processed by the CPU 15 is written to the main storage device 16. The main storage device 16 and the CPU 15 are connected by an address bus, a data bus, a control signal, and the like. The cache unit 20 is an auxiliary memory and work memory of the main storage device 17.

  The database unit 3 additionally stores a fragment record composed of items such as ID, column name, column value, and time as shown in FIG. Fragment records with the same ID are integrated into a single integrated record 3a.

  For example, the integrated record 3a in FIG. 3 is obtained by integrating fragment record groups having an RFID of 212, and includes columns such as temperature and path (ID reader number). In the storage area of the database unit 3, only the start address of the free area is always managed, and in the additional recording of data, the second fragment record is written after the first fragment record one after another. Append. By writing data to the continuous area, the storage area is not wasted and writing can be performed at high speed.

  The hash table storage unit 18 is a storage area for storing a hash table. The function storage unit 19 is a storage area for storing a hash function and a dispersion function.

  As shown in FIG. 2, the CPU 15 includes a data registration processing unit 15a, a search processing unit 15b, a hash table management unit 15c, a function management unit 15d, a hash function calculation unit 15e, a distribution function calculation unit 15f, a data fragment registration unit 15g, A first partial search unit 15h, a second partial search unit 15i, a search result configuration unit 15j, a record integration unit 15k, a plan generation unit 151, and the like are provided.

  The data registration processing unit 15 a processes a data registration command from the application program of the client terminal 4. The search processing unit 15b processes a data search command from the application program. The hash table management unit 15 c manages the hash table storage unit 18. The hash table is registered as an entry when data is registered, and is used for entry search during data search. The function management unit 15d manages a hash function and a distribution function used at the time of data registration and retrieval.

  The hash function calculation unit 15e is a database management system in which data is stored or should be stored by inputting a key value (record ID) and column name pair in a distributed database in which the database is distributed by combining horizontal and vertical distribution. A function that returns a hash value combining the device ID and the data fragment ID is calculated.

  The distribution function calculation unit 15f inputs a column name and calculates a distribution function that returns a set of database management device IDs in which the data is stored or should be stored.

  The data fragment registration unit 15g stores a part of the record (fragment record) in the storage area of the database unit 3 by the appending method. The database management device 2 to store data is specified using the above hash function.

  The first partial search unit 15h is used in the processing of the WHERE clause at the time of inquiry processing, and specifies one or more database management device IDs that store data necessary for the search using a distributed function. Then, the database management device 2 transmits a portion of an inquiry that can be executed and executes a search.

  The second partial search unit 15i is used in the processing of the WHERE clause at the time of inquiry processing, and uses a key value (record ID) of the data specified from the intermediate result of the search, a column name pair, and a hash function, One or more database management device IDs storing data necessary for the search are specified. Further, the database management apparatus 2 transmits a query sentence (phrase) that can be executed to execute a search.

  The search result configuration unit 15j configures a final search result by processing the SELECT phrase during the inquiry process. The record integration unit 15k restores the record from the fragment record and executes inquiry processing. The plan generation unit 151 generates a plan for processing the WHERE clause.

(Global server)
As shown in FIG. 4, the global server 1 includes an input device 21, an output device 22, a communication control device 23, a communication interface 24, a CPU 25, a main storage device 26, a load information storage unit 27, an access right information storage unit 28, and the like. I have.

  The load information storage unit 27 stores load information such as CPU cost and IO cost. The access right information storage unit 28 stores access right information such as the database management apparatus 2 that can be used by the application program and an access permission period for the database management apparatus 2.

  The CPU 25 includes a load information collection unit 25a and an access right management unit 25b. The load information collection unit 25 a periodically collects load information such as CPU cost and IO cost of all the database management devices 2 to be distributed and writes it in the load information storage unit 27.

  The access right management unit 25b controls the application program to use a database management device with a low load in order to prevent concentration of access to the global server at the time of inquiry or data registration, and distributes load on inquiry processing and search processing. To do. The access right management unit 25b includes a statistical processing unit 25c and an access right calculation unit 25d.

  The statistical processing unit 25c collects load information serving as a reference for determining the database management device 2 having the lowest load, and statistically processes the data.

  In response to a request from the application program, the access right calculation unit 25d calculates the database management apparatus 2 that can be accessed by the application program, and the period during which access is permitted, and returns it to the application program. As the accessible database management device 2, the database management device 2 having a light load at that time is selected and determined based on the collected load information.

  The other devices are the same as the database management device 2 and will not be described.

(Distributed database management system operation)
As main operations of the distributed database management system 100,
1. 1. Initialization processing of the database management device 2 2. Information collection processing of the global server 1 3. Access right issue processing of the global server 1 4. Data registration processing of the database management device 2 5. Inquiry processing of the database management device 2 6. WHERE phrase processing of the database management device 2 7. Conditional expression processing in the WHERE clause of the database management device 2 A SELECT phrase process of the database management apparatus 2 is exemplified. Hereinafter, each of these operations will be described with reference to the drawings.

(Initialization of database management device)
Hereinafter, the initialization process of the database management apparatus 2 will be described with reference to the flowchart of FIG.

(A) First, in step S11, the hash function calculation unit 15e of one database management device 2 calculates a hash function used for data distribution processing. In the calculation of the hash function, the function h1 is determined by using statistical information such as a histogram of the values of the fragment records that have already been input, so that a set of key values of the fragment records will be uniformly hashed in the future. For how to determine the hash function, see introduction to algorithms (author: Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, Clifford Stein, publisher: The MIT Press)
), And SHA-1, MD5, etc. are widely known.

Further, the dispersion function calculation unit 15f calculates the dispersion function h2. The hash function h1 and the distribution function h2 can be implemented by the function storage unit 19 of each database management device 2 holding at least the following input-output correspondence table. That is, h1 and h2 are
h1 (input: record key value, column name) → (output: database management device ID, record key value hash value)
The function is a set of h2 (input: column name) → (output: database management device ID). However, the database management device ID obtained by h1 (key value, column name) is included in the database management device ID group obtained by h2 (column name). h1 receives the key value of the record and the column name (or table name and column name) in the database schema, and calculates a hash value. The hash value is a composite hash value, which is the ID of the distributed database management device 2 and the hash value of the input record.

  h2 receives a column name in the schema of the database unit 3 or a table name and a column name as input and returns a set of IDs of the database management devices 2 to be distributed.

(B) In step S12, the determined hash function is transmitted to all the database management apparatuses 2 (distributed database management apparatus group 5) via the communication network.

(C) In step S <b> 13, the function management unit 15 d of all the database management apparatuses 2 receives the same hash function and stores it in the function storage unit 19. Further, the hash table management unit 15c creates a hash table based on this hash function. The created hash table is stored in each hash table storage unit 18.

(Global server information collection processing)
Next, processing in which the global server 1 collects load information from the distributed database management device group 5 will be described with reference to the flowchart of FIG.

(A) First, in step S21, the administrator of the global server 1 inputs a load information collection opportunity. The collection trigger is determined by a method of making a regular time interval, a method of triggering when the database management device 2 is below a certain load, or the like. In addition, there are a method of using those standards that are unified in the distributed database management device group 5 and a method of using a method that is different for each individual database management device 2.

(B) In step S22, the load information collection unit 25a in FIG. 4 collects load information such as IO cost and CPU cost from the distributed database management device group 5 according to the designated collection opportunity, and the global server 1 It is written in the load information storage unit 27 of FIG.

(C) The process in step S22 is executed unless the administrator of the global server 1 designates stop in step S23. If stop is designated, the load information collection flow is terminated.

(Global server access right issuance processing)
Next, processing in which the global server 1 issues an access right in response to a request from the application program of the client terminal 4 will be described with reference to the flowchart of FIG.

(A) First, in step S31, the access right management unit 25b accepts an access right issuance request from the application program of the client terminal 4.

(B) In step S32, using the CPU / IO load information of all the database management devices 2 collected by the statistical processing unit 25c, the access right calculation unit 25d uses as the access right the application program that should most be accessed. The database management device 2 with a low load and the access period to the database management device 2 are calculated. The calculated access right is stored in the access right information storage unit 28.

  Regarding the access period, the CPU / IO load information of the past database management apparatus and the access right already assigned to the application program are referred to, and the change of the load of each database management apparatus 2 is predicted. Thereby, an access period with a low load is determined.

(C) In step S33, the database management device 2 to which these accesses are permitted and its access period are returned to the application program as access rights via the communication network.

(Data registration processing of database management device)
Next, an operation when the database management apparatus 2 that receives a registration request from the application program of the client terminal 4 and is determined by the access right registers data will be described with reference to the flowchart of FIG.

(A) In step S41, the data registration processing unit 15a that receives the data registration request from the client terminal 4 inputs the key value and column name of the input fragment record into the hash function h1, and obtains the database management device ID. Using this database management device ID, the database management device 2 to register the registered fragment record is specified.

  For example, in the first fragment record of FIG. 3, the key value is 212 and the column name is temperature. These are input to the hash function h1.

(B) In step S42, it is determined whether the database management device 2 to be registered is its own device or another device. In the case of the own device, the data fragment registration unit 15g registers the fragment record in the database unit 3 of the own device. Also, the hash table management unit 15c adds this data registration to the hash table storage unit 18 as a hash table entry. The location where the entry is added in the hash table is determined using the hash value of the key value of the record, which is another result obtained from the hash function h1.

(C) In the case of another device, the fragment record requested for registration is transmitted to the designated other device, and a request is made to register it in the database unit 3 of the other device.

(Inquiry processing of database management device)
Next, the operation when the database management apparatus 2 that receives the inquiry from the application program of the client terminal 4 processes the inquiry will be described with reference to the flowchart of FIG.

(A) In step S51, the received inquiry is analyzed and a parse tree of the inquiry is created.

(B) In step S52, a subroutine for processing the WHERE clause of the inquiry is executed.

(C) In step S53, a subroutine for processing the SELECT phrase is executed.

(WHERE phrase processing of database management device)
Next, a subroutine for processing the WHERE phrase in step S52 will be described with reference to the flowchart of FIG.

(A) In step S61, the plan generation unit 151 creates a plan for executing the WHERE clause. In the execution plan, an execution plan is created by combining the processing of the minimum conditional expression obtained by dividing the WHERE clause by AND / OR / parentheses in parallel or in series. For example, in the case of a WHERE clause of a = 5 AND b> 3 AND c = 10, the minimum conditional expressions are a = 5, b> 3, and c = 10.

  In this example, possible execution plans are a completely serial plan 1 shown in FIG. 11 (a), a plan 2 that executes only one conditional expression shown in FIG. 11 (b) in parallel, and a plan 2 shown in FIG. 11 (c9). This is plan 3 for executing two conditions in parallel.

  Since there are three conditional expressions, there are three types of plan 2, but FIG. 11B describes a plan for executing only b> 3 in parallel. When there are three or more conditional expressions, there is a plan that sets three or more conditional expressions that can be executed in parallel. Using a generally known distributed query optimization technique, an execution plan that can be processed at the highest speed is selected and the plan is executed.

(B) In step S62, the conditional expression processing in the WHERE phrase is performed by the first partial search unit 15h and the second partial search unit 15i. Details will be explained in the next section.

(Conditional expression processing in the WHERE clause of the database management device)
Next, a conditional expression processing method performed by the first partial search unit 15h and the second partial search unit 15i will be described with reference to the flowchart of FIG.

(A) First, in step S71, when the key value is not specified in the WHERE clause, the database management device 2 in which the fragment record to be searched is registered using the first partial search unit 15h. If a key value is specified in the WHERE clause, the second partial search unit 15i is used to determine the database management device 2 in which the fragment record to be searched is registered.

(B) In step S72, it is determined whether the identified database management device 2 is the self device or another device. If it is the self device, the processing of the conditional expression described later is performed in step S73.

(C) If it is determined that the device is another device, the conditional expression is transmitted to the other device (database management device 2) identified in step S74, and a request is made to transmit the conditional expression result value. In step S75, a conditional expression result value is received from another device.

  The processing of the conditional expression will be described for each plan. In the plan 1, first, a = 5 is processed to specify a fragment record group that satisfies the condition. In the processing of the conditional expression in the WHERE clause described later, the column name a is input to the distribution function h2, and as a result, one or more database management devices 2 that manage the data of the column name a are specified. Then, a = 2 is processed for each identified database management apparatus 2, and a fragment record group is obtained as a result.

  Next, b> 3 is processed for each fragment record. In the processing of the conditional expression in the WHERE clause described later, one or more database management apparatuses 2 that store the column fragments of b of the corresponding record are specified by inputting the key value and the column name b to the hash function h1. Then, b> 3 is processed for each specified database management apparatus 2, and a fragment record group is obtained as a result.

  Finally, c = 10 is processed for each fragment record. This is to specify one or more database management devices 2 that store the c column fragment of the corresponding record by inputting the key value and the column name c to the hash function h1 in the processing of the conditional expression in the WHERE clause described later. . Then, c = 10 is processed for each specified database management apparatus 2, and a fragment record group is obtained as a result.

  In plan 2, the process of a = 5 and the process of b> 3 are executed in parallel. Next, c = 10 is processed for each fragment record obtained as a result of a = 5. Finally, a product set of the fragment record group obtained as a result of b> 3 and the fragment record obtained by processing c = 10 is calculated.

  In plan 3, all the conditions are executed in parallel, and the product set of the obtained fragment record groups is calculated.

(SELECT phrase processing of database management device)
Next, a subroutine in which the search result configuration unit 15j processes the SELECT phrase in step S53 will be described with reference to the flowchart of FIG.

(A) First, in step S81, the search result configuration unit 15j uses the hash function h1 for each fragment record calculated by the WHERE clause, and inputs the corresponding key value and the column name specified by the SELECT clause. One or more database management devices that store the specified column fragment are identified.

(B) In step S82, it is determined whether the specified database management device 2 is its own device or other device. If it is the own device, the fragment record to be searched is specified from the hash table in step S83, and the column value is acquired. To do.

(C) If it is determined as another device, the other device (database management device 2) identified in step S84 is requested to identify the fragment record to be searched from the hash table and acquire the column value. In step S85, a column value is received from another device. Finally, the record integration unit 15k restores the record from the fragment record stored in the database unit 3, and executes the inquiry process.

  Therefore, it is possible to add data to the continuous area while enabling normal inquiry processing, and the registration processing is performed at high speed.

(Example)
Next, an example of the operation of the distributed database management system 100 will be described with reference to FIG. FIG. 14 shows an application example in supply chain management. The client terminal 4 uses an application program to “examine the section (ID reader number) where room temperature is 30 ° C. or higher in the transportation route of goods 212”. Executing a query. This is SQL, “SELECT
ID reader number FROM distribution goods WHERE RFID = 212 AND temperature> 30 ”.

  If this inquiry is executed by the database management apparatus 2, first, the SQL is converted into an inquiry tree by parsing. Next, the WHERE clause is processed. In this example, in order to process the conditional expression of temperature> 30, the database management device 2 storing the fragment record whose key value is 212 and the column name is temperature is specified using the hash function h1, and 4 is obtained as a result. .

  Then, a conditional expression of temperature> 30 is transmitted to the database management apparatus 2 specified in 4 to identify and return a fragment record group (key value is 212) that satisfies the condition.

  Finally, the SELECT phrase is processed. The database management device 2 that stores the fragment record having the key value 212 and the column name ID reader number is specified using the hash function h1, and 2 is obtained as a result. Then, the database management apparatus 2 specified in 2 identifies the fragment record corresponding to the key value 212, and returns the column value.

  As described above, according to the present invention, it is possible to solve the problem that the registration process of the fragment record is slow and the problem that the storage area on the disk is vacant.

  The problem that the global server becomes a performance bottleneck as the number of distributed database management apparatuses increases can be solved.

  Data registration / inquiry processing can be performed at high speed by determining which database management device has data to be stored / accessed data in data registration / inquiry processing at high speed.

It is a figure which shows the structure of the distributed database management system which concerns on embodiment of this embodiment. It is a figure which shows the structure of the database management apparatus which concerns on embodiment of this embodiment. It is a figure which shows the internal structure of the data of a database part. 2 is a diagram showing a configuration of a global server 1. FIG. 4 is a flowchart showing an initialization process of the database management device 2. 3 is a flowchart showing information collection processing of the global server 1. 4 is a flowchart showing access right issue processing of the global server 1. 4 is a flowchart showing a data registration process of the database management device 2. 4 is a flowchart showing an inquiry process of the database management device 2. 10 is a flowchart showing a WHERE phrase process of the database management device 2. It is a figure which shows the conditional expression of each plan. It is a flowchart which shows the conditional expression process in the WHERE clause of the database management apparatus 2. It is a flowchart which shows the SELECT phrase process of the database management apparatus 2. It is a figure which shows the Example of embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Global server 2 ... Database management apparatus 3 ... Database part 3a ... Integrated record 4 ... Client terminal 5 ... Distributed database management apparatus group 11 ... Input apparatus 12 ... Output apparatus 13 ... Communication control apparatus 14 ... Communication interface 15 ... CPU
15a ... Data registration processing unit 15b ... Search processing unit 15c ... Hash table management unit 15d ... Function management unit 15e ... Hash function calculation unit 15f ... Distributed function calculation unit 15g ... Data fragment registration unit 15h ... First partial search unit 15i ... First 2-part search unit 15j ... search result configuration unit 15k ... record integration unit 15l ... plan generation unit 16 ... main storage device 17 ... main storage device 18 ... hash table storage unit 19 ... function storage unit 20 ... cache unit 21 ... input device 22 ... Output device 23 ... Communication control device 24 ... Communication interface 25 ... CPU
25a ... Load information collection unit 25b ... Access right management unit 25c ... Statistics processing unit 25d ... Access right calculation unit 26 ... Main storage device 27 ... Load information storage unit 28 ... Access right information storage unit 100 ... Distributed database management system

Claims (6)

A database device that distributes and arranges data by combining horizontal and vertical distribution,
By inputting a key value and a column name for designating the data, a database management device ID in which data specified by the key value and the column name is stored or to be stored, and a fragment record ID which is a part of the data are A hash function computing unit that returns a combined hash value;
A dispersion function calculation unit that returns one or more database management device IDs in which data specified by the column name is stored by inputting the column name;
A database management device comprising: A data fragment registration unit for registering the fragment record as data;
A database part that stores the registered fragment records in the appending method;
The database management apparatus according to claim 1, further comprising: a record integration unit configured to integrate the fragment records stored in the database unit to form a record. A distributed database management system for distributing data by combining horizontal and vertical distribution,
By inputting the key value and the column name, a hash value obtained by combining the data specified by the key value and the column name or the database management device ID to be stored and the fragment record ID which is a part of the data A plurality of database devices having a hash function operation unit to be returned and a distributed function operation unit for returning one or more database management device IDs in which data specified by the column name is stored by inputting the column name; ,
A load information collection unit that collects load information of the plurality of database management devices;
A global server having a database management device having the lowest load and an access right management unit that determines an access right including an access period to the database management device by statistically processing the load information in the load information collection unit And
Access to the plurality of database devices is permitted based on the access right determined by the global server. When registering the fragment record in the database management device to which the fragment record is to be registered among the plurality of database management devices, the key value and the column name held by the fragment record are input, and the hash function is used. A data registration processing unit that identifies an ID of a database management device to store data, and transmits and registers the fragment data to the identified database management device;
When searching the fragment records in the plurality of database management devices, specify one or more column names described in the query statement for the search, and use the distribution function to search for the fragment records Identify at least one database management device ID storing data to be transmitted, send a part of the query statement that can be executed by the identified at least one database management device, and execute a search A first partial search unit to perform,
Based on the key value and column name of the data specified from the intermediate result of executing the search, the hash function is used to identify and specify the ID of the database management device that stores the data required for the search The distributed database management system according to claim 3, further comprising: a second partial search unit that transmits a partial query that can be executed by the database management apparatus, and executes a search. A distributed database management method for integrated management of a plurality of database management devices for distributing and arranging data by combining horizontal and vertical distributions,
A load information collection unit collecting load information of the plurality of database management devices;
By statistically processing the load information in the load information collection unit, the access right management unit determines an access right including a database management device having the lowest load and an access period to the database management device;
Allowing access to the plurality of database devices based on the access rights;
When a plurality of database management devices that distribute and hold the data are accessed and a key value and a column name are input, data specified by the key value and the column name is stored or database management to be stored A step in which the hash function calculation unit returns a hash value obtained by combining the device ID and the fragment record ID that is a part of the data;
A distributed database management method comprising: a distributed function computing unit returning one or more database management device IDs in which data specified by the column name is stored by inputting the column name. In a computer that distributes data by combining horizontal and vertical distribution,
An instruction for a load information collection unit to collect load information of a plurality of database management devices for distributing and arranging the data;
By statistically processing the load information in the load information collection unit, an instruction for the access right management unit to determine an access right including a database management device having the lowest load and an access period to the database management device;
An instruction for permitting access to the plurality of database devices based on the access right;
When a plurality of database management devices that distribute and hold the data are accessed and a key value and a column name are input, data specified by the key value and the column name is stored or database management to be stored An instruction for the hash function computing unit to return a hash value obtained by combining the device ID and the fragment record ID that is a part of the data;
A distributed database management program that, upon input of the column name, causes a distributed function calculation unit to execute one or more database management device IDs in which data specified by the column name is stored.

Images