JP2013065113A - Load distribution system, data access device, and load distribution method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To distribute loads.SOLUTION: A load distribution system of an embodiment comprises a plurality of database servers and database access devices. Each of the plurality of database servers includes association storage means. The association storage means stores identification information and associated information associating with each other. The plurality of association storage means store the same identification information. The database access devices each include selection means, update request means, identification information storage means, and readout request means. The selection means selects a database server. The update request means requests the update of associated information having been associated with identification information and association storage means, to the selected database server. The identification information storage means stores identification information and database identification information associating with each other. The readout request means reads out the associated information that has been associated with the identification information by the association storage means from the database server identified with the database identification information that has been associated with the identification information by the identification information storage means.

Description

  Embodiments described herein relate generally to a load distribution system, a data access device, and a load distribution method.

  Conventionally, with the development of network and computer technology, services provided to users tend to become large-scale. With the increase in the scale of such services, various technologies have been proposed in order to provide stable services.

  With the increase in scale of this service, there is a problem that when a plurality of devices are connected, the load is concentrated on a certain device. In order to solve such a problem, various techniques for distributing the load have been proposed.

  For example, there is a technique for distributing and managing data with a plurality of hard disks using a technique such as RAID (Redundant Arrays of Inexpensive Disks). Further, a technique has been proposed that includes a plurality of database access servers for accessing a database and distributes the load between the database access servers so that the number of databases connected to the database is equal.

  Further, in the conventional technology, when a high transaction load is applied to the database system, a technique of reducing the load on the database server per unit is often adopted by distributing and arranging data accommodated by arranging a plurality of database servers.

JP 2008-234373 A

  However, in the system in which data is distributed, transactions are biased on the condition that an event occurs. As described above, when the load is concentrated on a part of the database servers, there has been a problem that the performance of the database is deteriorated or that the database system is provided with specifications more than necessary in advance.

  The load distribution system according to the embodiment includes a plurality of database servers and a database access device. The plurality of database servers include correspondence storage means. The correspondence storage means stores the identification information and the related information in association with each other. The plurality of correspondence storage means store the same identification information. The database access device includes selection means, update request means, identification information storage means, and read request means. The selection means selects a database server from a plurality of database servers. The update request unit requests the selected database server to update the related information associated with the identification information by the correspondence storage unit. The identification information storage means stores the identification information and the database identification information in association with each other. The read request unit reads the related information associated with the identification information by the correspondence storage unit from the database server identified by the database identification information associated with the identification information and the identification information storage unit.

FIG. 1 is a block diagram illustrating a configuration of a control system according to the first embodiment. FIG. 2 is a diagram illustrating an example of a network. FIG. 3 is a diagram showing an example of a user data storage method in a conventional DB system. FIG. 4 is a diagram illustrating an example of a user data storage method for a plurality of DB systems included in the control system according to the first embodiment. FIG. 5 is a block diagram illustrating a hardware configuration of the first SQL server and the second SQL server according to the first embodiment. FIG. 6 is a diagram illustrating a software configuration realized by the arithmetic unit of the first SQL server according to the first embodiment reading a program. FIG. 7 is a diagram illustrating an example of a table structure of the terminal correspondence table according to the first embodiment. FIG. 8 is a diagram illustrating an example of a table structure of the update table according to the first embodiment. FIG. 9 is a diagram illustrating an example of a table structure of the DB access table according to the first embodiment. FIG. 10 is a diagram illustrating a hardware configuration of the first DB system according to the first embodiment. FIG. 11 is a diagram illustrating a table structure of the main body DB according to the first embodiment. FIG. 12 is a diagram illustrating a hardware configuration of the first MP server according to the first embodiment. FIG. 13 is a flowchart illustrating a procedure of processing performed when a transaction is received in the SQL server according to the first embodiment. FIG. 14 is a sequence diagram illustrating backup processing in the control system according to the first embodiment. FIG. 15 is a diagram illustrating an example of a table structure of a DB access table held by the first to nth SQL servers according to the second embodiment. FIG. 16 is a flowchart illustrating a procedure of processing performed when a transaction is received in the SQL server according to the second embodiment. FIG. 17 is a diagram illustrating the concept of a DB system according to the third embodiment. FIG. 18 is a diagram illustrating an example of a table structure of a terminal correspondence table according to the third embodiment. FIG. 19 is a diagram illustrating an example of a table structure of a DB access table according to the third embodiment. FIG. 20 is a diagram illustrating an example of the DB access table according to the third embodiment when the first DB system and the fourth DB system belonging to the group 2 are backed up.

(First embodiment)
FIG. 1 is a block diagram illustrating a configuration of a control system 1 according to the first embodiment. A control system 1 illustrated in FIG. 1 is connected to terminal devices 180_1 to 180_k, an operator terminal 191, and a configuration management device 192 via a network. The control system 1 is assumed to process a transaction related to authentication and control from the external terminal devices 180_1 to 180_k and return a response.

  The terminal devices 180_1 to 180_k are PCs, telephone terminals, or mobile phone terminals that are connected via a network. Furthermore, an IP router or the like for connecting to another terminal device may be used. The terminal devices 180_1 to 180_k have one of the addresses as the connection destination among the first SQL server 100_1 to the nth SQL server 100_n.

  The terminal device 180_1 to the terminal device 180_k according to the present embodiment has addresses of a plurality of SQL servers among the above-described first SQL server 100_1 to the nth SQL server 100_n, so that the load is distributed. One SQL server is selected as a connection destination. Then, the terminal device 180_1 to the terminal device 180_k transmit a transaction to the selected connection destination.

  The operator terminal 191 and the configuration management device 192 are used for maintenance of the control system 1. The operator terminal 191 is a terminal for controlling the control system 1 and can be used as an interface when backing up the DB systems 120_1 to 120_m, for example.

  The configuration management device 192 is a device that manages the configuration of the DB systems 120_1 to 120_m, and is used when constructing a database of these DB systems or pouring data.

  The control system 1 includes a first SQL server 100_1 to an nth SQL server 100_n, a first MP server 111, a second MP server 112, an L2SW 131, an L2SW 132, and a first DB system 120_1 to the first DB system 120_1. m DB systems 120_m.

  By the way, in a database system that immediately returns a response to a large number of large-scale transactions, the data structure is simplified in order to shorten the response time. That is, a single unique key is associated with each data, and a database for processing all transactions by only reading, updating, and adding the data is constructed. In the control system 1 according to the present embodiment, it is further possible to distribute the load when accessing the constructed first DB system 120_1 to the m-th DB system 120_m. Examples of the environment to which the control system 1 is applied include the following examples.

  FIG. 2 is a diagram illustrating an example of a network. In the network shown in FIG. 2, small ITX devices 251_1, 251_2, and 251_l are arranged around the IP network 250. There is a call control system 201 that controls incoming and outgoing calls of various terminal devices connected via the small ITX devices 251_1, 251_2, and 251_1.

  The call control system 201 identifies authentication and connection paths for hundreds and tens of millions of users, and notifies the caller. As described above, the call control system 201 is required to manage a large number of terminal devices, and to improve reliability and responsiveness of incoming and outgoing calls. For this reason, it is important to distribute the load. Therefore, the load can be distributed by applying the control system 1 shown in FIG.

  Further, the control system 1 in FIG. 1 can be applied to systems other than the call control system 201. For example, the present invention can also be applied to the short message service system 202 that manages transmission and reception of short messages in the PHS network 260 shown in FIG. For example, the short message service system 202 is considered to increase the load when a short message can be transmitted to a mobile phone terminal in addition to between PHS. Therefore, even in the short message service system 202, the load can be distributed by applying the control system 1 as shown in FIG.

  Note that the control system 1 shown in FIG. 1 is not limited to the call control system 201 or the short message service system 202, and can be applied to any system including a plurality of databases. Returning to FIG. 1, the configuration in the control system 1 will be described.

  The first SQL server 100_1 to the nth SQL server 100_n accept a transaction from the terminal device 180_1 to 180_k, issue a statement for accessing a database necessary for processing, and perform the first DB system 120_1 to mth. The DB system 120_m is accessed.

  The first SQL server 100_1 to the nth SQL server 100_n are composed of a plurality of servers as shown in FIG. Transactions from the terminal devices 180_1 to 180_k are transmitted by selecting one SQL server from the addresses of a plurality of SQL servers registered in each device. Thereby, the load from the terminal devices 180_1 to 180_k in the first SQL server 100_1 to the nth SQL server 100_n is distributed.

  When the first SQL server 100_1 to the nth SQL server 100_n receives a transaction from the terminal device 180_1 to 180_k, the first DB system 120_1 to the mth DB system 120_m are received from the key information included in the transaction request. Search individual data stored in. The individual data is information related to the terminal device or the subscriber associated with the terminal identification ID, and includes statistical information such as the number of accesses, for example. The first SQL server 100_1 to the nth SQL server 100_n update the statistical information each time transaction processing is performed.

  For this reason, the first SQL server 100_1 to the nth SQL server 100_n generate statements for searching for individual data and updating statistical information, and send the statements to the first DB system 120_1 to the mth DB system 120_m. To do.

  In the present embodiment, the example in which the transactions of the first SQL server 100_1 to the nth SQL server 100_n are distributed on the terminal device 180_1 to 180_k side has been described. However, the terminal devices 180_1 to 180_k may not have such a function. In such a case, for example, a similar function can be realized by arranging a load balancer on the control system 1 side.

  The first MP server 111 and the second MP server 112 perform processing according to an operation from the operator terminal 191. In addition, the first MP server 111 and the second MP server 112 accept data for adding, changing, or deleting individual data from the configuration management apparatus 192.

  The first MP server 111 and the second MP server 112 provide the operator terminal 191 with a function for maintaining the first DB system 120_1 to the m-th DB system 120_m. Further, the first MP server 111 and the second MP server 112 have a large capacity tape drive, and have a function of backing up data held by the first DB system 120_1 to the m-th DB system 120_m. Have. Furthermore, it has a function of collecting statistical data such as the performance of each of the first DB system 120_1 to the m-th DB system 120_m and storing them in the own apparatus.

  One of the first MP server 111 and the second MP server 112 operates as an active system and the other as a standby system, and the function is not lost even if one of them becomes a failure.

  The L2SW 131 and the L2SW 132 are connected between the first SQL server 100_1 to the nth SQL server 100_n, the first MP server 111, the second MP server 112, and the first DB system 120_1 to the mth DB system 120_m. A switch to be connected. The device is not limited to such a switch, and any device that can transfer data may be used. For example, an IP router may be used.

  The first DB system 120_1 to the m-th DB system 120_m each include a storage device that stores a terminal identification ID for identifying a terminal device and individual data related to the terminal in association with each other. The individual data includes individual data for each terminal device. The storage device may not be a single device, and may be, for example, RAID 1 + 0.

  The same terminal identification ID is stored in each of the storage devices (RAID 1 + 0) included in the first DB system 120_1 to the m-th DB system 120_m.

  Then, the first DB system 120_1 to the m-th DB system 120_m receive statements for accessing the database from the first SQL server 100_1 to the n-th SQL server 100_n, and add records to the database. , Reference or update.

  The first DB system 120_1 to the m-th DB system 120_m are a set of two (first 1-1 DB server 121_1 and 1-2 DB server 122_1, 2-1 DB server 121_2 and 2-2DB server 122_2, third -1 DB server 121_3, 3-2 DB server 122_3, m-1 DB server 121_m, and m-2 DB server 122_m). And it has a redundant structure by performing replication mutually.

  The first SQL server 100_1 to the nth SQL server 100_n have a function of selecting the first DB system 120_1 to the mth DB system 120_m. For example, when the first DB system 120_1 is selected, it can be used by accessing the 1-1 DB server 121_1 or the 1-2 DB server 122_1.

  A case will be described in which the first DB system 120_1 receives a statement for accessing the database from the first SQL server 100_1. In this case, when the statement is a read request, the first DB system 120_1 searches the database in the 1-1 DB server 121_1, and sends the search result to the first SQL server 100_1 that is the request source. Deliver. When the statement is an update request, the database in the 1-1st DB server 121_1 is updated. At this time, an update request is also sent to the first-side DB server 122_1 on the replication side. Then, after waiting for the reception of these responses, the first DB system 120_1 transmits an update completion response to the first SQL server 100_1 that is the request source. With this configuration, the control system 1 can continue to operate without any data loss even if either the 1-1 DB server 121_1 or the 1-2 DB server 122_1 fails.

  Although the first DB system 120_1 to the m-th DB system 120_m according to the present embodiment have a replication configuration of two units, the number of replications may be three or more.

  A conventional DB system will be described. FIG. 3 is a diagram showing an example of a user data storage method in a conventional DB system. In the example shown in FIG. 3, it is assumed that the entire system is constructed by the first to m-th DB systems. In such a case, user data is distributed and stored in each DB system every 1 / m. Then, by using a technique such as cluster or RAID, it is possible to operate these first to m-th DB systems as one large-scale DB system. However, since the user data stored in each DB system is different, for example, when reading and updating are concentrated on the user data stored in the third DB system, there is a problem that the load cannot be distributed. .

  FIG. 4 is a diagram illustrating an example of a user data storage technique for the DB systems 120_1 to 120_m included in the control system 1 according to the first embodiment. As in the example shown in FIG. 4, all user data is stored for each of the first to m-th DB systems. The latest data is held by any one of the first to m-th DB systems. When reading, refer to the DB system in which the latest user data is stored, and when updating, select one DB system from the first to mth DB systems using a rule that distributes the load. The user data stored in the selected DB system is updated. Thereby, the update is distributed to each DB system. In particular, when accessing a database, the load is higher in updating than in reading, and therefore the load can be distributed by performing this control. Next, each configuration for realizing the control will be described.

  FIG. 5 is a block diagram showing a hardware configuration of the first SQL server 100_1 and the second SQL server 100_2.

  The first SQL server 100_1 includes an external I / F 501, a server I / F 502, a calculation unit 503, a memory 504, a storage device 505, and a DB_I / F 506. The second SQL server 100_2 and other SQL servers not shown are also assumed to have the same configuration as the first SQL server 100_1, and the description thereof is omitted.

  The external I / F 501 receives a transaction such as an authentication request from the terminal devices 180_1 to 180_k.

  The server I / F 502 is an interface that performs communication with the first SQL server 100_1 to the nth SQL server 100_n.

  The computing unit 503 controls the entire first SQL server 100_1. The arithmetic unit 503 reads the program stored in the storage device 505 and implements various processes. For example, the arithmetic unit 503 performs processing by interpreting a transaction received by the first SQL server 100_1 and a request from another server.

  The memory 504 stores various data. For example, the memory 504 stores various data expanded by the processing of the calculation unit 503. For example, when the first SQL server 100_1 is activated, the terminal correspondence table 511, the update table 512, the DB access table 513, and the like are expanded in the memory 504.

  The storage device 505 stores programs and the like. The storage device 505 is used by the calculation unit 503 when starting up the server.

  The DB_I / F 506 is an interface that communicates with various DB servers.

  FIG. 6 is a diagram illustrating a software configuration realized by the calculation unit 503 of the first SQL server 100_1 reading a program. As illustrated in FIG. 6, the calculation unit 503 includes a terminal request reception unit 601, a selection unit 602, a read request unit 603, a server synchronization control unit 604, and an update request unit 605.

  The terminal correspondence table 511 holds a DB system in which the latest data is stored for each terminal device. FIG. 7 is a diagram illustrating an example of the table structure of the terminal correspondence table 511. As illustrated in FIG. 7, the terminal correspondence table 511 stores a terminal identification ID and a DB system number in association with each other.

  The terminal identification ID is an ID for identifying the terminal devices 180_1 to 180_k. The terminal identification ID is unique identification information as a storage target by the control system 1. As the terminal identification ID, for example, a telephone number or a subscriber ID for specifying a subscriber may be applied.

  The DB system number is a number for identifying a DB system in which the latest individual data associated with the terminal identification ID is stored. The latest individual data means the last updated individual data.

  The update table 512 is a table for managing update requests from other SQL servers. FIG. 8 is a diagram illustrating an example of the table structure of the update table 512. As shown in FIG. 8, the update table 512 stores SQL server identification information and command sequence numbers in association with each other.

  The SQL server identification information is information for identifying each SQL server. As information for identifying the SQL server, for example, an IP address may be used.

  The command sequence number is a number for performing synchronization with another SQL server 100_x (x is any numerical value from 2 to n). The command sequence number is assigned separately for each SQL server. The command sequence number is incremented by 1 each time an update request is received from the SQL server 100_x. If the command sequence number received by the server synchronization control unit 604 from the other SQL server 100_x together with the update request and the command sequence number stored in the update table 512 are not consecutive, the SQL server 100_1 It is assumed that the update request from the SQL server 100_x is missing.

  The DB access table 513 is a table for selecting an update destination DB system. FIG. 9 is a diagram showing an example of the table structure of the DB access table 513. As shown in FIG. 9, the DB access table 513 stores DB system identification information, an access ratio, and an “ON / OFF” flag in association with each other.

  The DB system identification information is information for identifying one of the first DB system 120_1 to the m-th DB system 120_m. When the DB system identification information is stored in the DB access table 513, the DB system identified by the DB system identification information is to be updated.

  The access ratio indicates the rate at which the DB system indicated by the DB system identification information is updated. In the control system 1 according to the present embodiment, the DB system to be updated by the SQL servers 100_1 to 100_m is selected at random, but the ratio to be selected is stored in the DB access table 513.

  The “ON / OFF” flag is a flag for switching whether to perform load distribution by switching the DB system to be updated.

  When “OFF” is set in the “ON / OFF” flag, switching of the DB system storing the latest individual data is suppressed. When switching of the DB system storing the latest individual data is inhibited, it is possible to back up each DB server with a normal backup function provided in the database management system.

  The terminal request receiving unit 601 receives transaction requests from the terminal devices 180_1 to 180_k. And the terminal request reception part 601 performs an authentication process with respect to the terminal device 180_1-180_k which performed the transaction request | requirement as needed.

  The selection unit 602 selects a database server to be updated from the first DB system 120_1 to the m-th DB system 120_m according to a rule determined so that the load is distributed to each database server. In the present embodiment, an example of selecting a database at random will be described as an example of the rule. The present embodiment is not limited at random. For example, the first DB system 120_1 to the m-th DB system 120_m may be selected according to a predetermined order.

  The selection unit 602 selects a database server to be updated based on the access ratio associated with each database identification information stored in the DB access table 513. The selection unit 602 according to the present embodiment selects a DB system in accordance with the access ratio every time a transaction is accepted. For example, when the access ratio of the DB system number 1 is 20%, the selection unit 602 selects the DB system number 1 when 1 to 20 occurs in the random numbers that take a numerical value of 1 to 100.

  By the way, in the method of randomly selecting a DB system that performs data update, the load due to data update is distributed, but if many references are made, the load may be concentrated on a specific DB system. Accordingly, when the access concentration is detected by measuring the number of accesses for each DB system in the first SQL server 100_1 to the nth SQL server 100_n or measuring the response time of the DB system, the access is concentrated. By setting the SQL server access ratio to be low, the load on the DB system can be immediately made uniform.

  As an example different from the present embodiment, when the selection unit 602 of the SQL server holds a predetermined selection ratio (a ratio for selecting a read-source DB system) and receives a transaction, an update destination DB It may be selected whether the system is a read-out DB system or is randomly selected from all DB systems. As described above, for some transactions, the number of updates of the terminal correspondence table 511 can be reduced by selecting the DB system from which the individual data is read out as an update target. Furthermore, even when the transaction per unit time becomes high, the number of updates of the terminal correspondence table 511 performed between the SQL servers can be suppressed. In this case, the data update load is not distributed in the authentication process corresponding to the selection ratio, but since the individual data of the terminal device with high access frequency is gradually moved to another DB system, the load is evenly distributed over time. Is done.

  Returning to the first embodiment, the read request unit 603 reads the latest individual data from the first DB system 120_1 to the m-th DB system 120_m in response to transactions from the terminal devices 180_1 to 180_k. The read request unit 603 according to the present embodiment specifies a DB system identified by the DB identification number associated with the terminal identification ID that identifies the requesting terminal device and the terminal correspondence table 511 as the read destination. Then, the read request unit 603 reads individual data associated with the terminal identification ID in the identified DB system.

  Thus, the latest update data is stored in any of the first DB system 120_1 to the m-th DB system 120m. Therefore, the first SQL server 100_1 according to the present embodiment can recognize which DB system stores the latest individual data by referring to the terminal correspondence table 511, and therefore accesses the DB system. . In the present embodiment, an example in which the terminal correspondence table 511 is referred to has been described. However, other methods may be used as a detection method of the DB system in which the latest individual data is stored.

  The update request unit 605 is an update request for individual data associated with a terminal identification ID for identifying a terminal device that has made an update request, with a database access statement, for the database in the DB system selected by the selection unit 602. I do. Further, the update request unit 605 may request addition of data when the corresponding record does not exist.

  When the update request unit 605 makes an update request to a DB system different from the DB system that stores the latest data, the terminal identification ID indicating the update request source is sent to the terminal correspondence table 511, and The terminal correspondence table 511 is updated in association with the update destination DB system number. Thereby, the terminal correspondence table 511 can always hold the DB system in which the latest individual data is stored.

  In this way, by updating the randomly selected DB system, the load generated by the update process can be distributed to all DB systems.

  The server synchronization control unit 604 synchronizes between the update performed by the first SQL server 100_1 and the updates performed by the other SQL servers 100_2 to 100_n.

  For example, when the DB system is updated by the update request unit 605, the server synchronization control unit 604 includes SQL server identification information for identifying the own device (first SQL server 100_1) that has performed the update, and a sequence number. The updated terminal identification ID and the DB system number indicating the update destination are associated with each other and notified to the other SQL servers 100_2 to 100-n via the server I / F 502.

  Further, the server synchronization control unit 604, via the server I / F 502, SQL server identification information for identifying other SQL servers, a command sequence number, an updated terminal identification ID, and a DB system number indicating an update destination, , The terminal correspondence table 517 is updated with the received information (after associating the terminal identification ID with the DB system number indicating the update destination). Furthermore, the server synchronization control unit 604 determines whether the received command sequence number and the command sequence number stored in the update table 512 are consecutive. If it is continuous, it is assumed that the process is normally performed, and “1” is added to the command sequence number in the update table 512, and the synchronization control is terminated. If not consecutive, the server synchronization control unit 604 requests the other SQL server to retransmit an update request regarding the missing command sequence number. Thereafter, the server synchronization control unit 604 updates the command sequence number in the update table 512.

  As a result, the terminal correspondence tables 511 included in the SQL servers 100_1 to 100_n are synchronized and always store the same data. That is, in the present embodiment, since the synchronization processing needs to be performed quickly, there is no room for checking whether or not the update is appropriately performed. Therefore, the server synchronization control unit 604 according to the present embodiment requests retransmission for an update request regarding a command sequence number that is missing from the update table 512. Thereby, reliable synchronization in the terminal correspondence table 511 provided in each of the SQL servers 100_1 to 100_n can be realized.

  Next, the DB system will be described. FIG. 10 is a diagram illustrating a hardware configuration of the first DB system 120. The 1-1 DB server 121_1 and the 1-2 DB server 122_1 have the same configuration. The first DB system 120_1 is composed of two servers, a 1-1 DB server 121_1 and a 1-2 DB server 122_1. When one DB server stops, replication synchronization is taken to continue operation on the other.

  As illustrated in FIG. 10, the first-first DB server 121_1 includes an SQL_I / F 1001, an arithmetic unit 1002, a memory 1003, a storage device 1004, and a DB_I / F 1005.

  The SQL_I / F 1001 receives a statement for accessing a database from the first SQL server 100_1 to the nth SQL server 100_n.

  The DB_I / F 1005 is an interface for performing database replication synchronization between two DB servers.

  The calculation unit 1002 performs processing when a statement for accessing the database is received.

  The memory 1003 stores various data. A database cache or the like is expanded in the memory 1003 according to the present embodiment.

  The storage device 1004 stores data. The storage device 1004 according to the present embodiment is constructed with RAID 1 + 0. In the present embodiment, a plurality of disk drives are physically used across one database for RAID 1 + 0. Thereby, the load of each disk drive can be distributed. Furthermore, the storage device 1004 manages the database to be stored by dividing it into a plurality of databases. The first SQL server 100_1 to the nth SQL server 100_n refer to, update, and add data in a manner similar to that in which a plurality of DB systems exist. As a result, it is possible to prevent access to a specific disk drive from being biased.

  Each database in the storage device 1004 includes a transaction log storage unit 1011 and a main body DB 1012. By using a transaction log that can be written at high speed and stored in the transaction log storage unit 1011, replication synchronization between the 1-1 DB server 121_1 and the 1-2 DB server 121_2 becomes possible.

  Each database in the storage device 1004 has a roll forward function. Thus, even if a failure occurs when the data table is not updated, the data table can be recovered from the transaction log after recovery. Therefore, high-speed processing is possible while ensuring fault tolerance.

  FIG. 11 is a diagram showing a table structure of the main body DB 1012. In the example illustrated in FIG. 11, the main body DB stores a terminal identification ID, terminal related information, and statistical information in association with each other.

  The terminal related information is information related to the terminal and is universal information with a low update frequency. The terminal related information includes, for example, the name and address of the user who owns the terminal device, and the model of the terminal device. As for the terminal-related information, the same data is registered in all of the first DB system 120_1 to the m-th DB system 120_m.

  The statistical information is assumed to be transient information that is updated each time the terminal device accesses the control system 1. As the statistical information, in the example illustrated in FIG. As for the statistical information, the latest data is registered in any one of the first DB system 120_1 to the m-th DB system 120_m.

  Next, the MP servers 111 and 112 will be described. FIG. 12 is a diagram illustrating a hardware configuration of the first MP server 111. As shown in FIG. 12, the first MP server 111 includes an operator I / F 1101, an external I / F 1102, a calculation unit 1103, a memory 1104, an HDD 1105, a tape drive 1106, and an internal I / F 1107. Prepare. Note that the second MP server 112 has the same configuration as the first MP server 111, and a description thereof will be omitted.

  One of the first MP server 111 and the second MP server 112 is activated, and the other one is deactivated. When one unit does not start due to a failure or the like, the other unit starts.

  The operator I / F 1101 is an interface that communicates with the operator terminal 191. The external I / F 1102 is an interface that communicates with the configuration data management server.

  The internal I / F 1107 is an interface that communicates with the first SQL server 100_1 to the nth SQL server 100_n, the first DB system 120_1 to the mth DB system 120_m, and the second MP server 112.

  The memory 1104 is used as a work area for the arithmetic unit 1103 or a work area for backup.

  The calculation unit 1103 controls the entire MP server 111. The arithmetic unit 1103 implements a function for performing maintenance of the control system 1 by executing a program stored in the HDD 1105. For example, the calculation unit 1103 processes according to an instruction from the operator terminal 191 or processes data from the configuration management device 192.

  In addition, the calculation unit 1103 includes a terminal identification ID stored in association with the main body DB in each DB server of the first DB system 120_1 to the m-th DB system 120_m, individual data, HDD 1105 and tape in association with each other. Back up to drive 1106.

  The first MP server 111 and the second MP server 112 add and update the data of the DB server in the DB system based on the data received from the configuration management device 192. For this reason, the first MP server 111 and the second MP server 112 hold the terminal correspondence table 511, the update table 512, and the DB access table 513 on the memory 1104, similarly to the SQL server.

  The HDD 1105 stores statistical data of the first SQL server 100_1 to the nth SQL server 100_n, the first DB system 120_1 to the mth DB system 120_m, and the first MP server 111 itself.

  The tape drive 1106 reads out various data stored in the HDD 1105 and performs backup. Note that the arithmetic unit 1103 performs encryption processing on the data before backing up the data in the tape drive 1106.

  When the first MP server 111 or the second MP server 112 backs up the first DB system 120_1 to the m-th DB system 120_m, the first SQL server 100_1 to the n-th SQL. The server 100_n temporarily stops the selection of the DB system to be updated by the selection unit 602. Thus, when the first MP server 111 or the second MP server 112 backs up the data held in the first DB system 120_1 to the m-th DB system 120_m, which DB server has the latest data. Can be backed up while maintaining consistency between the data indicating whether the data is stored in the database and the data actually stored in the DB server.

  Next, the processing of the SQL server that is performed when a transaction from the terminal device is received will be described. FIG. 13 is a flowchart showing a procedure of the above-described processing in the SQL server according to the present embodiment. In the example illustrated in FIG. 13, the first SQL server 100 </ b> _ <b> 1 is used for description. However, any SQL server performs the same process. The terminal devices 180_1 to 180_k hold the IP addresses of the first SQL server 100_1 to the nth SQL server 100_n, and select one SQL server in a round robin method and transmit a transaction at the time of authentication or the like. To do. Thereby, the transaction processing loads of the first SQL server 100_1 to the nth SQL server 100_n are equalized. In this flowchart, an example in which the first SQL server 100_1 performs processing will be described.

  First, the terminal request receiving unit 601 receives a transaction from a terminal device (for example, the terminal device 180_1) (step S1301). Next, the terminal request reception unit 601 authenticates the terminal device 180_1 that has transmitted the transaction (step S1302).

  When the authentication is properly performed, the terminal request reception unit 601 of the first SQL server 100_1 refers to the terminal correspondence table 511 in the own apparatus and is associated with the terminal identification ID for identifying the transmission source of the transaction. The existing DB system number is read (step S1303).

  Then, the read request unit 603 reads individual data associated with the terminal identification ID from the DB system identified by the read DB system number (for example, the first DB system 120_1), and stores the individual data Reading is locked (step S1304).

  Thereafter, each time the selection unit 602 receives an update request, the random number is used to randomly select one DB system to be updated from the first DB system 120_1 to the m-th DB system 120_m (step S1305). ).

  Then, the update request unit 605 updates the individual data associated with the terminal identification ID that identifies the transmission source of the transaction for the selected DB system (step S1306).

  Next, the update request unit 605 determines whether the DB system that updated the individual data is the same as the DB system that read the individual data (step S1307). If it is determined that they are the same (step S1307: YES), the process proceeds to step S1310.

  On the other hand, when it is determined that the update request units 605 are not identical (step S1307: No), the terminal identification ID for identifying the transmission source of the transaction and the DB system number for identifying the updated DB system with respect to the terminal correspondence table 511 Are updated in association with each other (step S1308).

  Furthermore, the server synchronization control unit 604 associates the updated terminal identification ID with the DB system number and notifies the other SQL server and MP server (step S1309). As a result, the terminal correspondence table 511 provided in other SQL servers and MP servers is updated.

  Thereafter, the terminal request receiving unit 601 releases the lock of the individual data that has been prohibited from being read, and returns a response to the terminal device that is the transaction transmission source (step S1310). This completes a series of transaction processing.

  Next, backup processing performed in the control system 1 will be described. FIG. 14 is a sequence diagram illustrating the above-described processing in the control system 1 according to the present embodiment. This sequence diagram is an example in which the first MP server 111 backs up.

  First, the first MP server 111 sets the “ON / OFF” flag in the DB access table 513 to “OFF” for the first SQL server 100_1 to the nth SQL server 100_n in the control system 1. (Step S1401).

  Accordingly, the first SQL server 100_1 to the nth SQL server 100_n update the “ON / OFF” flag in the DB access table 513 to “OFF” (step S1402). Then, the first SQL server 100_1 to the nth SQL server 100_n notify the first MP server 111 of the update (OK) (step S1403).

  As a result, since the switching of the DB server storing the latest data is stopped, the first MP server 111 requests the first DB system 120_1 to the m-th DB system 120_m to perform backup (step). S1404). Thereafter, the first DB system 120_1 to the m-th DB system 120_m read data from the database in the storage device 1004 (step S1405). Then, the first DB system 120_1 to the m-th DB system 120_m transfer the read data to the first MP server 111 (step S1406). The backup procedure may be the same as the conventional procedure.

  Then, the first MP server 111 accumulates the transferred data in the HDD 1105 (step S1407).

  After the transfer of all data is completed, the first MP server 111 sets the “ON / OFF” flag in the DB access table 513 to the first SQL server 100_1 to the nth SQL server 100_n, “ It is requested to turn “ON” (step S1408).

  Accordingly, the first SQL server 100_1 to the nth SQL server 100_n update the “ON / OFF” flag in the DB access table 513 to “ON” (step S1409). Then, the first SQL server 100_1 to the nth SQL server 100_n notify the first MP server 111 of the update (OK) (step S1410).

  Thereafter, the calculation unit 1103 in the first MP server 111 performs an encryption process on the data stored in the HDD 1105 (step S1411). Thereafter, the arithmetic unit 1103 stores the data after the encryption processing in the tape drive 1106 (step S1412).

  Backup is performed according to the processing procedure described above. By performing the above-described processing procedure at the time of this backup, data indicating which DB server stores the latest data held by the first SQL server 100_1 to the nth SQL server 100_n, and the DB server Consistency with stored data can be achieved.

(Modification 1 of the first embodiment)
In the first embodiment, the example in which the access ratio is determined when the selection unit 602 selects is described. The access ratio is not limited to a predetermined one, and the access ratio of each DB system may be changed according to the load of each DB system. The selection unit 602 measures the access ratio by measuring the number of times that the first SQL server 100_1 to the nth SQL server 100_n have referred, added, and updated each DB system, and weights corresponding to the load of each process After multiplying by the constant, the load of each DB server is calculated by adding the total number of times after multiplying by the weighting constant. Then, the selection unit 602 changes the access ratio when detecting a load deviation of the DB system. Thereby, the load of each DB system can be made close to uniform.

(Modification 2 of the first embodiment)
Further, the access ratio changing method is not limited to the first modification. In the second modification, an access ratio changing method different from the first modification will be described. The selection unit 602 according to the modification 2 measures the response time from when the statement for accessing the database is transmitted until the response is returned for each of the first DB system 120_1 to the m-th DB system 120_m. . When a bias is detected in the response time for each DB system, the access ratio is changed so that the response times are uniform. Thereby, it is possible to suppress a bias in response time by making the load of each DB system close to uniform.

(Modification 3 of the first embodiment)
In Modification 3, an access ratio changing method different from Modifications 1 and 2 will be described. In the third modification, the first SQL server 100_1 to the nth SQL server 100_n are loaded with loads other than normal transaction processing such as DB server failure, database replication resynchronization, and hard disk RAID reconstruction from the DB system. When the notification of the occurrence is received, the selection unit 602 sets an access ratio predetermined for each event for the notification source DB system with the notification of these events as a trigger. As a result, the impact of the event on transaction processing can be reduced, and the load on the DB system can be reduced.

(Second Embodiment)
In the first embodiment, the example in which the SQL server can be updated for any DB system has been described. However, when a plurality of SQL servers try to update information related to the same terminal device, a situation occurs in which the same record of the same DB system is updated at almost the same timing. In such a case, there is a possibility that proper synchronization cannot be performed when synchronizing between SQL servers.

  Therefore, in the second embodiment, an example will be described in which SQL servers that perform updates for each DB system are limited in advance. It is assumed that access is restricted only to updates, and references can be requested from any SQL server.

  In order to realize this, the first SQL server 100_1 to the nth SQL server 100_n need to hold a correspondence relationship between the update destination DB system and the SQL server that requests the update in the case of the DB system. There is.

  FIG. 15 is a diagram illustrating an example of a table structure of the DB access table 513 held by the first SQL server 100_1 to the nth SQL server 100_n according to the second embodiment. As shown in FIG. 15, the DB access table 513 according to the second embodiment stores DB system identification information, SQL server identification information, an access ratio, and an “ON / OFF” flag in association with each other. ing.

  As shown in FIG. 15, the DB access table 513 of the second embodiment is different from the first embodiment in that DB system identification information and SQL server identification information are associated with each other. Then, when the update destination DB system is selected by the selection unit 602, the update request unit 605 notifies the update request to another SQL server associated with the selected DB system.

  Next, the processing of the SQL server that is performed when a transaction is received from the terminal device according to the present embodiment will be described. FIG. 16 is a flowchart showing a procedure of the above-described processing in the SQL server according to the present embodiment. In the example illustrated in FIG. 16, the description will be made using the first SQL server 100_1, but it is assumed that any SQL server performs the same process.

  The processing procedure shown in FIG. 16 is the update destination in the same procedure as steps S1301 to S1305 in FIG. 13 of the first embodiment (except for the individual data reading lock performed in step S1304). The selection is performed up to the selection of the DB system (steps S1601 to S1605).

  Thereafter, the update request unit 605 refers to the DB access table 513 to determine whether or not the selected DB system is associated with the own apparatus, that is, whether or not the selected DB system can be updated ( Step S1606). If it is determined that updating is possible (step S1606: Yes), the same processing as steps S1306 to S1309 of the first embodiment is performed.

  On the other hand, when the update request unit 605 determines that the selected DB system cannot be updated (step S1606: No), the selected DB system and the SQL server associated with the DB access table 513 (the selected DB An update request is notified to an SQL server that can update the system (step S1607). As a result, the requested SQL server performs the same processing as steps S1608 to S1611.

  Thereafter, the terminal request receiving unit 601 returns a response to the terminal device that is the transaction transmission source (step S1612). This completes a series of transaction processing.

  In the present embodiment, by limiting the SQL servers that perform updates to the DB system, it is possible to improve stability and reliability when performing synchronization between SQL servers.

(Third embodiment)
In the first embodiment, when backing up the database, it is necessary to set all the “ON / OFF” flags of the DB access table 513 to “OFF” to stop the switching of the DB system to be updated. . Since it is necessary to stop the switching for all the DB systems, there is a possibility that the processing load is biased between the DB systems during the backup as long as the switching is stopped. Therefore, in the third embodiment, a technique for reducing the bias in processing load during backup will be described.

  In the third embodiment, individual user data is managed by being divided into a plurality of groups. FIG. 17 is a diagram showing the concept of the DB system according to the present embodiment.

  As shown in FIG. 17, a group is composed of a plurality of DB systems. In the example illustrated in FIG. 17, it is assumed that storage devices 1701 and 1702 of RAID 1 + 0 are provided. The group 1 includes a second DB system 1712 and a third DB system 1713. The group 2 includes a first DB system 1711 and a fourth DB system 1714.

  Thus, the storage devices 1701 and 1702 are divided into groups. In the DB system in the same group, individual data associated with the same terminal identification ID is stored as in the above-described embodiment. On the other hand, the DB system between different groups does not include the same terminal identification ID, but stores different terminal identification IDs. That is, the terminal identification ID is managed for each group.

  FIG. 18 is a diagram illustrating an example of a table structure of the terminal correspondence table 511 according to the third embodiment. As illustrated in FIG. 18, the terminal correspondence table 511 stores terminal identification IDs, groups, and DB system numbers in association with each other. The group is the identification information of the group in which the terminal identification ID is stored.

  Then, the load is distributed within the same group. For example, when updating the individual data related to the terminal identification ID managed in the group 1, the load is distributed between the second DB system 1712 and the third DB system 1713. On the other hand, when updating the individual data related to the terminal identification ID managed in the group 2, the load is distributed between the first DB system 1711 and the fourth DB system 1714. The storage devices 1701 and 1702 are connected to other storage devices (not shown), and replication is executed in the same manner as in the above-described embodiment.

  The DB access table 513 holds the correspondence between this group and the DB access. FIG. 19 is a diagram illustrating an example of the table structure of the DB access table 513. As illustrated in FIG. 19, the DB access table 513 stores DB system identification information, a group, an access ratio, and an “ON / OFF” flag in association with each other. Thus, in this embodiment, the DB system and the group are stored in association with each other. Then, the selection unit 602 according to the present embodiment selects a DB system to be updated from the group to which the terminal identification ID belongs and the associated DB system.

  As shown in FIG. 17, since the DB system of the same group is built in different storage devices (RAID1 + 0) 1701, 1702, not only the load is distributed within the group but also the storage devices (RAID1 + 0) 1701, 1702. The load is also distributed within.

  The storage device (RAID1 + 0) 1701 is managed by a certain DB server, and the storage device 1702 is managed by a different DB server. Thus, in this embodiment, the DB server manages a plurality of DBs. Then, by updating each DB system from the SQL server, load distribution is performed according to the number of hard disk drives installed in the DB server and the RAID configuration without any restrictions on the scale or number of DB servers. Can be done efficiently.

  The control system 1 switches the “ON / OFF” flag for each group in the DB access table 513 when performing backup. FIG. 20 is a diagram illustrating an example of the DB access table 513 when the first DB system 1711 and the fourth DB system 1714 belonging to the group 2 are backed up.

  As shown in FIG. 20, “ON / OFF” of the first and fourth DB systems to which the group 2 belongs becomes “OFF”, and “ON / OFF” of the second and third DB systems to which the group 1 belongs. "Is" ON ". As a result, when selecting the update target, the selection unit 602 suppresses the selection of the DB system to be updated with respect to the terminal identification ID belonging to the backup target group, and selects the update target group. A DB system to be updated is selected for a terminal identification ID that does not belong.

  In the first and fourth DB systems, switching is stopped for backup, whereas in the second and third DB systems, load distribution can be realized. Thus, by backing up in units of groups, it is possible to prevent the load distribution from being completely stopped, and to reduce the load bias in the database system.

  In the embodiment and the modification described above, the processing load among a plurality of databases can be made uniform by distributing updates with a higher load than the database reference. As described above, by making the processing load uniform, it is possible to suppress the performance degradation of the database. Furthermore, it is possible to eliminate the need for performance that is higher than necessary.

  A program executed by the first SQL server 100_1 to the nth SQL server 100_n according to the above-described embodiment and modification is a file in an installable format or an executable format, and is a CD-ROM or a flexible disk (FD). , CD-R, DVD (Digital Versatile Disk), and the like recorded on a computer-readable recording medium.

  Further, the programs executed by the first SQL server 100_1 to the nth SQL server 100_n according to the above-described embodiment and modification are stored on a computer connected to a network such as the Internet and downloaded via the network. You may comprise so that it may provide. Further, the program executed by the first SQL server 100_1 to the nth SQL server 100_n according to the above-described embodiments and modifications may be provided or distributed via a network such as the Internet.

  Further, the programs executed by the first SQL server 100_1 to the nth SQL server 100_n according to the above-described embodiments and modifications may be provided by being incorporated in advance in a ROM or the like.

  The programs executed in the first SQL server 100_1 to the nth SQL server 100_n according to the embodiment and the modification described above are the above-described units (terminal request reception unit, selection unit, read request unit, server synchronization control unit, Update request unit) is included in the module configuration, and as the actual hardware, the arithmetic unit 503 reads the program from the storage medium and executes the program so that each unit is loaded on the memory 504, and the terminal request reception unit, A selection unit, a read request unit, a server synchronization control unit, and an update request unit are generated on the memory 504.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 1 ... Control system, 100_1-100_n ... SQL server, 111, 112 ... MP server, 120_1-120m ... DB system, 121_1-121_4, 122_1-122_4 ... DB server, 180_1-180_k ... Terminal device, 191 ... Operator terminal, 192 DESCRIPTION OF SYMBOLS ... Configuration management apparatus 201 ... Call control system 202 ... Short message service center 250 ... IP network 251_1, 251_2, 251_l ... Small ITX device 260 ... PHS network 501 ... External I / F 502 ... Server I / F, 503 ... arithmetic unit, 504 ... memory, 505 ... storage device, 511 ... terminal correspondence table, 512 ... update table, 513 ... DB access table, 601 ... terminal request reception unit, 602 ... selection unit, 603 ... read request unit 604: Server synchronous control , 605 ... Update request unit, 1001 ... SQL_I / F, 1002 ... Calculation unit, 1003 ... Memory, 1004 ... Storage device, 1005 ... DB_I / F, 1011 ... Transaction log storage unit, 1101 ... Operator I / F, 1102 ... External I / F, 1103 ... arithmetic unit, 1104 ... memory, 1105 ... HDD, 1106 ... tape drive, 1107 ... internal I / F, 1701, 1702 ... storage device, 1711 to 1714 ... DB system

Claims (12)

In a load balancing system comprising a plurality of database servers and a database access device that accesses a database held by the plurality of database servers,
Multiple database servers
Corresponding storage means for storing unique identification information as a storage target of the load balancing system and related information associated with the identification information in association with each other,
The same identification information is stored in the correspondence storage means provided in the plurality of database servers,
The database access device is:
Selecting means for selecting a database server to be updated according to a rule determined so that a load is distributed to each database server from the plurality of database servers;
Update request means for making an update request for the related information associated with the identification information and the correspondence storage means for the database server selected by the selection means;
Identification information storage means for storing the identification information and the database identification information for identifying the database server that made the update request last for the related information associated with the identification information by the update request means; ,
Request the database server identified by the database identification information associated with the identification information and the identification information storage means to read out the related information associated with the identification information and the correspondence storage means. Reading request means,
Load balancing system. The selection means stops the selection of the database server according to the rule when performing backup for the plurality of database servers;
The load distribution system according to claim 1. The load balancing system includes:
A maintenance device comprising backup information for backing up the identification information stored in association with the correspondence storage means provided in the plurality of database servers, the related information, and a storage medium;
The update request means makes an update request to the database server identified by the database identification information stored in the identification information and the identification information storage means while being backed up by the backup means.
The load distribution system according to claim 2. The database access device is:
Further comprising ratio storage means for storing database server identification information for identifying each of the plurality of database servers and ratio information for updating the database server in association with each other;
The selection means selects a database server to be updated based on the ratio information associated with each database server stored in the ratio storage means.
The load distribution system according to any one of claims 1 to 3. The load distribution system includes a plurality of the database access devices,
The database access device is:
Correspondence storage means for storing database server identification information for identifying a database server and database access identification information for identifying a database access device in association with each other is further provided.
The update request means is for database access apparatus identified by the database server identification information for identifying the database server selected by the selection means and the database access identification information associated with the correspondence storage means. , Notifying the related information update request,
The load distribution system according to any one of claims 1 to 5. A plurality of database servers comprise a plurality of the correspondence storage means,
The plurality of correspondence storage means provided in the plurality of database servers are divided for each group, the correspondence storage means in the same group stores the same identification information, and the correspondence storage means between different groups Different identification information is stored,
The load distribution system according to any one of claims 1 to 5. The selection means stops the selection of the database server according to the rule for a group to be backed up when performing backup in units of the plurality of database servers.
The load distribution system according to claim 6. In a data access device connected via a network to a plurality of database servers that store unique identification information as a storage target and related information associated with the identification information in association with each other,
Selecting means for selecting a database server to be updated according to a rule determined so that a load is distributed to each database server from the plurality of database servers;
Update request means for making an update request for the related information associated with the identification information and the database server, with respect to the database server selected by the selection means;
Identification information storage means for storing the identification information and the database identification information for identifying the database server that made the update request last for the related information associated with the identification information by the update request means; ,
Requests the database server identified by the database identification information associated with the identification information and the identification information storage means to read out the related information associated with the identification information and the correspondence storage means. A read request means;
A data access device comprising: The selection means stops the selection of the database server according to the rule when performing backup for the plurality of database servers;
The data access device according to claim 8. Further comprising ratio storage means for storing database server identification information for identifying each of the plurality of database servers and ratio information for updating the database server in association with each other;
The selection means selects a database server to be updated based on the ratio information associated with each database server stored in the ratio storage means.
The data access device according to claim 8 or 9. Correspondence storage means for storing database server identification information for identifying a database server and database access identification information for identifying other database access devices in association with each other is further provided,
The update requesting means is provided to another database access apparatus identified by the database server identification information for identifying the database server selected by the selection means and the database access identification information associated with the correspondence storage means. In response to the request for updating the related information,
The data access device according to any one of claims 8 to 10. A load balancing method executed by a data access device and a plurality of database servers,
The plurality of database servers include correspondence storage means for storing unique identification information as a storage target and related information associated with the identification information in association with each other,
The database access device includes identification information storage means for storing the identification information in association with the database identification information for identifying the database server that last updated the related information associated with the identification information. Prepared,
The database access device, from the plurality of database servers, a selection step of selecting a database server to be updated according to a rule determined so that a load is distributed to each database server;
The database access device, for the database server selected in the selection step, an update request step for making an update request for the related information associated with the identification information and the correspondence storage means,
The database access device sends the related information associated with the identification information in the correspondence storage means to the database server identified by the database identification information associated with the identification information and the identification information storage means. A read request step for requesting reading from
A load balancing method.

Images