JP4434838B2 - Distributed data equivalence method and system, and program - Google Patents

Description

  The present invention operates a plurality of computers connected via a network as a master computer and a replica computer, respectively, and when the target data is updated to the original by the master computer, the target data of each replica computer is the same as that of the master computer. The present invention relates to a distributed data equivalent technique equivalent to content.

  In an industrial system (power, gas, water, public transportation, etc.) that is said to be the lifeline that forms the backbone of modern cities, the database used to monitor and control the system becomes huge and can be accessed at high speed. It has been demanded. And if the state of an industrial system changes, a database will be changed and the data content of the computer of several places is calculated | required to be kept equal. By keeping them equal, data is not acquired via a network at a plurality of locations, so that high-speed access is possible.

  In addition, when operating the most important computer as a server, a dual system (or multiple system) that backs up the server function so that the function of the industrial system is not lost even if this server stops is used. It is usual to configure. By adopting a multi-system configuration, even if the server is stopped, the server function is immediately transferred to another computer so that the function of the industrial system is not impaired. Even if the server function is transferred, it is necessary that the data can be accessed and accessed as before, and the other remaining computers are similarly kept in the same state. Thus, in an industrial system, a distributed data equivalent system that keeps data contents equal among a plurality of computers is required.

  In Patent Document 1, as such a distributed data equivalent system, a monitoring control method for a power system is proposed in which changes are transferred in order to keep the data of a plurality of computers in the system equal. In this method, when there is a change in the contents of the monitoring control database of the power system, the change contents are exchanged between the computers. According to this method, when the contents of the database are updated in each computer, each computer sends out the changed contents to other computers. If there is a data change at the same location at the same time, the data contents are corrected and adjusted later. In this method, the database can be updated from a plurality of computers.

  In Patent Document 2, when there are a plurality of computers, one specific computer is called a base host, and when the base host can update data, a base host switching type database management system that switches the base host Is described. In this method, when the base host is switched, information indicating that the base host has been changed is sent out from a certain computer by replication communication, and the base host content of each computer is rewritten. In addition, when the base host is switched, the data of the old and new base hosts are matched with the timing of switching to keep the database data equal to the latest data by the replication means.

  Further, Patent Document 3 describes a distributed data management technique that maintains the data contents of a plurality of computers consistently by replication. In this method, updating to the database is performed in a series of operations called transactions. The contents of the transaction are executed in the database, and the update contents executed in the database are replicated to other computers by the communication means.

Japanese Patent No. 2787108 JP 2002-297593 A JP 2002-297428 A

  In a distributed data equivalent system that keeps data in a plurality of computers equal, it is required that transmission of change contents for keeping data contents be performed at high speed. Since a multi-system is configured and continues as a system even if the server stops, it is necessary to move the server function (master) to another and continue to operate. There is a need for a data management system that has a function of switching masters and that can be used continuously with the data kept the same even after switching.

  The technique disclosed in Patent Document 1 transmits data changes from each computer, but the same location may be updated by a plurality of computers at the same time. In such a case, one computer is used. It was necessary to make an adjustment that the update contents of the remaining computers were sequentially reflected after invalidating the update contents other than. In order to perform such adjustment, it is necessary to store the update order arbitration and change history. In the end, with this method, the update throughput of the entire system was not increased, and each computer required a large amount of main memory compared to the amount of data to be changed. .

  In the method of Patent Document 2, when the base host (master) is switched, for example, there are changes such as “the base host has been changed to the host B” and “the base has been changed to the replica”. Since the management information of each computer is changed after being replicated, it was not possible to synchronize with the mode change of other computers so that another computer that confirmed that the base host no longer exists becomes the base host . Therefore, depending on the time delay of the communication time, there may be two or more base hosts simultaneously in a certain time zone. In addition, in order to match the data after switching by replication, there is a possibility that a computer whose data does not match may occur when a timeout occurs due to communication delay or the like. When multiple computers perform operations to become a base host that can update data at the same time, in order to mediate the authority to rewrite system management information, none of the management information is rewritten, and two or more base hosts Could have worked.

  Further, in the method of Patent Document 3, even if a plurality of data updates are executed in parallel, the data update is performed on the database in the form of a transaction, so that it is easily affected by the throughput of the database. Since the changes are replicated as they are, if detailed transactions continue, the communication overhead becomes larger than the actual changes, and the update throughput is limited to a level lower than the communication throughput. There was a possibility.

  Such problems are not limited to the field for monitoring and controlling the industrial system, but also exist in the distributed data equivalent technology in various fields in which the data update result of the master computer is reflected in each replica computer. Yes.

  The present invention has been proposed to solve the above-described problems of the prior art, and its purpose is to perform unnecessary data processing and storage of data equivalence between a master computer and each replica computer. It is to provide a distributed data equivalent method, system, and program that can be realized at high speed by efficiently using computer resources and communication resources.

  In order to achieve the above object, the present invention manages the mode information of the master computer and its own computer in each computer, transmits and receives the data update result based on the mode information, and retrieve requests from the replica computer By sending retrieve data from the master computer in response to the data, the data equivalence between the master computer and each replica computer can be efficiently utilized at high speed by efficiently using computer resources and communication resources without wasteful data processing and storage. It can be realized.

  The distributed data equivalent method of the present invention stores the target data in a plurality of computers connected via a network, and operates a specific one computer in the master mode and the other computers in the replica mode. This is a method for causing each replica computer to have the same content as that of the master computer when the target computer is operated as a replica computer and the target data is updated to the original by a data update operation on the master computer. In this distributed data equivalent method, each computer includes a database in which data files for storing target data are arranged in a shared area accessible from a plurality of applications, and performs a mode processing step, and also includes a master computer and a replica computer. The data update, update data merge, replication transmission, replication reception, merge data expansion, reception data reflection, retrieval request transmission, retrieval data transmission, and retrieval data reception are performed.

  Here, the mode processing step manages mode information including mode information for identifying which computer is the master computer among a plurality of computers and mode information for identifying the mode of the own computer. In this step, mode information is transmitted and received, and when mode switching is required, mode transition and a predetermined operation after mode transition are executed.

  Each step of data update, update data merge, and replication transmission is performed by a computer that operates as a master computer in the master mode. In the data update step, the result of the data update operation is temporarily stored in the update cache, and the content of the update cache is set as one update content commit data in response to the commit request, and the data file is added with unique data specifying information It is the step which stores in. The update data merging step is a step of generating merge data by merging one or a plurality of update content commit data determined in accordance with a preset merge merging condition. The replication transmission step is a step of managing transmission destination information indicating each replica computer to be equivalent based on the mode information and transmitting merge data as replication data to each replica computer based on the transmission destination information. .

  On the other hand, the steps of replication reception, merge data expansion, and reception data reflection are performed by a computer that operates as a replica computer in the replica mode. The replication reception step is a step of managing transmission source information indicating a master computer that is a transmission source of replication data based on the mode information, and receiving replication data from the master computer based on the transmission source information. The merge data expansion step is a step of expanding the merge data, which is replication data, into one or more original update content commit data. The received data reflection step determines whether or not to reflect the expanded update content commit data on each data update content commit data based on the retrieve status information indicating the retrieve status of the data. In this step, the determined update content commit data is reflected in the data file together with the data specifying information.

  Furthermore, the retrieval request transmission is a retrieval request that requests the retrieval data required for retrieval to copy all or part of the data file in a batch in order to eliminate the difference between the data file of the own computer and the master computer by the replica computer. Is transmitted to the master computer. The retrieve data transmission step is a step in which the master computer transmits the retrieve data to the replica computer in response to reception of the retrieve request from the replica computer. The retrieve data receiving step is a step of executing retrieve and updating retrieve state information when the replica computer receives retrieve data from the master computer.

  Note that the distributed data equivalent system and program of the present invention grasp the characteristics of the distributed data equivalent method as described above from another viewpoint.

  According to the present invention having such a feature, each computer performs a mode change in mode management called master switching, and the mode change is made in accordance with the mode transition rule triggered by the reception of the mode change of another computer, Data equivalence can be performed according to the state of transition. That is, it is possible to change the mode of a plurality of computers according to each other's situation, and to actively perform data equivalence according to the situation of mode management. Therefore, without causing a problem that a plurality of master computers exist, a single master computer can be determined and data can be continuously updated at high speed without adjustment with other computers. .

  In addition, when a data update operation is performed in each application, the data update result is stored in an update cache for each application, sent out by a commit operation, merged with commits made by one or more applications, and a certain amount of data Replication transmission can be performed in units. Therefore, data can be updated in parallel and at high speed by a plurality of applications. In addition, since a plurality of committed results are merged, the transmission overhead can be greatly reduced as compared with the case of individually transmitting. Therefore, even if the communication function has a lower throughput than the data update or commit, the data update result can be transmitted efficiently by making the best use of the capability of the communication function. It can be realized at high speed by efficiently using computer resources and communication resources without wasteful data processing and storage.

  As described above, according to the present invention, data equivalence between a master computer and each replica computer can be executed at high speed by efficiently using computer resources and communication resources without performing unnecessary data processing and storage. Distributed data equivalence methods and systems, and programs can be provided.

  Hereinafter, embodiments of a distributed data equivalent system to which the present invention is applied will be described with reference to the drawings. However, the embodiment described here does not limit the present invention at all, and merely illustrates one aspect of the present invention.

  The present invention is typically realized by controlling a computer with software. The software in this case realizes the functions and operational effects of each computer of the present invention by physically utilizing the hardware of the computer, and is suitable for conventional technology to which the conventional technology can be applied. Applies. Furthermore, specific types and configurations of hardware and software that realize the present invention, a range processed by software, and the like can be freely changed. For example, a program that implements the present invention is one aspect of the present invention.

[1. Glossary]
First, definitions of important terms used in the present specification will be described.

(1) Master computer:
It is a computer that can update original data, and can refer to data that has been updated to the original, instead of simply copying the target data to be equivalent. Usually, it is a specific unit in a system connected by a network.

(2) Replica computer:
The original data cannot be updated with respect to the target data to be equivalent, and the computer can only copy and refer to the original data updated by the master computer. That is, the target data handled by the replica computer is updated only by copying the original data of the master computer.

(3) Master switching:
One replica computer is set as a master computer. In general, when a master computer already exists, this means that one replica computer is set as a master. A computer that has been a master until now is sometimes called an old master, and a computer that has been newly set as a master is sometimes called a new master. In addition, when a master computer is lost due to a failure or the like and does not exist but only a replica computer exists, setting one replica computer as a master is also called master switching. Further, when there are a plurality of master computers due to setting or activation timing, setting one of a plurality of master computers or one of the other replica computers as a master is also referred to as master switching.

(4) Data file:
It is a file that stores the data body of the target data to be equivalent. Specifically, it is arranged on a storage device such as a computer hard disk or main memory. The data file is arranged in a shared area accessible from a plurality of applications, and is updated / referenced by each application. In general, a data file is composed of a data body and a data state table storing update serial numbers. In this case, the data body is composed of one or a plurality of data each having a data structure such as a table, a list, and a queue. In the following embodiments, as an example, the case where the data body is composed of a plurality of tabular data (tables) is described. The data file of the present invention is a concept including both data on the hard disk and mapping data described later.

(5) Data status table:
The data body is divided into several units, and each update serial number is stored for each unit. Specifically, when the data body is composed of a plurality of tables, the data state table stores the serial number of each table. If the table is divided into columns, the data state table stores the serial numbers of the table and the column.

(6) Serial number:
This number is incremented every time data is updated. Generally, a serial number is assigned to each unit constituting data. For example, when serial numbers are assigned in units of tables and columns, if there is an update, the serial number of the updated table and the serial number of the corresponding column of the table are updated.

(7) Update cache:
This is a temporary storage area provided for each application. That is, it is an area for storing data update contents made by each application.

(8) Data update:
“Data update” by an application is a concept including all of data rewriting, addition, and deletion. When such an operation is performed, the data file is normally stored once in a cache and is not immediately changed. When one application updates data, the data update result is not shown to other applications when the data update result is not committed.

(9) Commit:
The data update result stored in the cache is reflected in a data file shared by a plurality of applications. Reflection is to change the data value before updating the specific data to the data value obtained as a data update result when data update for specific data in the data file is performed. It is. When one application updates data, the data update result is shown to other applications only when the data update result is committed. The committed data for each time is called commit data, but may be abbreviated as commit.

(10) Mode:
The state of the calculator. There are at least a master mode and a replica mode. In the following embodiments, in addition to “in master operation” as the master mode and “in replica operation” in the replica mode, modes of “in master transition” and “in replica transition” are used. In addition, from the viewpoint of simplifying the description, mode information indicating the mode of the own computer or another computer is generally abbreviated as “mode”.

(11) Mode management:
It is at least maintaining the mode of the own computer, and if necessary, also includes maintaining a specific computer in the master computer. Furthermore, the mode of all computers may be maintained. In basic mode management, mode information from another computer is received and the mode of the own computer is changed, or mode information after the change is transmitted after the mode is changed. In addition, when mode information is received or the mode of the local computer is changed, replication and retrieval are also performed in cooperation with each other as necessary.

(12) Mode management table:
The mode management table stores at least mode information of the own computer. Actually, in order to use it for the functions of mode management, retrieval, and replication, the mode management table includes the name of the master computer (a unique identifier such as name or IP address) recognized by the local computer, and mode transmission / reception. The mode of other computers may be stored.

(13) Mode change:
It is to change the mode of the calculator. When the mode is changed, if necessary, the changed mode information is transmitted as part of the mode management.

(14) Retrieve:
Copying all or part of a data file from one computer at a time. For example, in the case of a table data file, an operation of copying an entire table or an operation of copying a part of a table column is a retrieve. In normal retrieval, the side that wants to receive a copy issues a retrieval request, and the side that receives the request transmits a necessary part (for example, a table or a column). Transmission is one to one.

(15) Retrieve status table:
It is the table | surface which stored the information which shows the state which the said data is in the retrieval for every unit which carries out the transmission / reception of the data main body in a data file. For example, there are states such as “Retrieving”, “Retrieve completed”, and the like. In the case of table data, the units for retrieving and transmitting may be column units, table units, or the like. In this case, the retrieval state table is a table in which the state for each column, the state for each table, and the like are described.

(16) Replication:
When the data file is updated in the master computer, the update content is transmitted to one or a plurality of replica computers. Usually, the same transmission content is sent simultaneously to a plurality of replica computers. The update content to be transmitted is called replication data. Informing the master computer that the replica computer is the target of receiving replication data in the replica computer is called replication connection. In the replica computer, the replication data can be received from the master computer by performing this replication connection.

(17) Replication management table:
7 is a table storing information indicating computers on the transmission / reception partner side on the transmission / reception side of replication. In other words, the replication destination computer name is stored on the replication transmission side, the replication transmission source computer name is stored on the replication reception side, and if both are the transmission side and the reception side, both are stored in a table containing both. is there.

(18) Merge:
It is to handle multiple commit data together. The collected data is called merge data. Conversely, dividing the merge data into a plurality of original commits is called merge expansion.

(19) Merge number:
This is a number assigned to each merge data. In the following embodiment, the retrieval of the merge number is checked in the merge development process, which triggers the retrieval request.

(20) Mapping data:
This is one method for realizing a data file shared by a plurality of applications. When the contents of the database are placed on a fixed hard disk and the database is started, the data on the hard disk is reflected in the process space on the main memory. If the mapping data reflected in the process space on the main memory is updated, the data on the hard disk is also updated immediately (depending on the OS implementation). If the mapping data is referred to, the data on the hard disk is read out.

(21) Simultaneous broadcast:
The simultaneous transmission to a plurality of computers. It is a one-sided communication that is transmitted regardless of the reception status on the receiving side. There may be a case where a plurality of computers are a specific group or an unspecified number. Multicast communication multi-broadcast communication in an IP network is an example of simultaneous broadcast communication.

(22) Application:
Although the application runs on each computer, an external application outside the system that generates data processing and mode processing of the distributed data equivalent system is particularly shown.

[2. System configuration]
FIG. 1 is a block configuration diagram showing each computer constituting a distributed data equivalent system (hereinafter abbreviated as this system) according to an embodiment to which the present invention is applied. As shown in FIG. 1, each computer has a database 11, a data processing unit 12, a merge data processing unit 13, a replication processing unit 14, a mode processing unit 15, a retrieve as constituent elements for realizing distributed data equivalent processing. A processing unit 16 and the like are provided. The data operation application 17 and the mode setting application 18 are external applications outside the system that operate on each computer and generate data processing and mode processing of the system. In addition, the function and structure of each component 11-16 of this system are as follows.

  The database 11 is a part for storing target data such as distributed data equivalent system monitoring control data, and includes a data file 111 and its mapping data 112 arranged on a hard disk. Among these, the data file 111 is arranged in a shared area accessible from a plurality of applications on the hard disk. The mapping data 112 is data that reflects the data file 111 in the process space on the main memory. The construction of the database 11 from the data file 111 and the mapping data 112 in this way is one of the methods for realizing the present invention on a computer.

  The data processing unit 12 is a part that processes data in the database 11, and includes a data reference processing unit 121, a data update processing unit 122, a received data reflection unit 123, and an update data cache 124. Among these, the data reference processing unit 121 is a part that operates in both the master mode and the replica mode, and refers to the data in the database 11 in response to a data reference request from the data operation application 17.

  The data update processing unit 122 is a part that operates in the master mode and updates the data in the database 11. The data update processing unit 122 updates the result of the data update operation in response to the data update request from the data operation application 17. The data is temporarily stored in 124, and the content of the update data cache 124 is set as one update content commit data in response to the commit request, and a unique serial number is added and stored in the database 11.

  The received data reflecting unit 123 is a part that operates in the replica mode and reflects the data update contents of the database 11 on the master computer side to the database 11 of the own computer. The received data reflecting unit 123 acquires the update content commit data obtained on the master computer side through the merge data processing unit 13 of the own computer, and for each update content commit data, based on the retrieve state table described later. It is determined whether or not to be reflected in the database 11 of the computer, and the update content commit data determined to be reflected is reflected in the database 11 together with the serial number.

  The merge data processing unit 13 is a part that merges update content commit data and merges and expands merged data, and includes an update data merge unit 131 and a merge data expansion unit 132. Among them, the update data merge unit 131 is a part that operates in the master mode, and generates merge data by merging one or a plurality of update content commit data determined in accordance with a preset merge merge condition. The merge data expansion unit 132 is a part that operates in the replica mode. The merge data obtained on the master computer side is acquired through the replication processing unit 14 of the own computer, and the original one or more update contents are acquired. Expand to commit data.

  The replication processing unit 14 is a part that transmits and receives merge data of update content commit data as replication data, and includes a replication management unit 141, a replication communication unit 142, and a replication management table 143. Among these, the replication management unit 141 is a part that operates in both the master mode and the replica mode, and manages the replication management table 143.

  The replication management table 143 stores replica computer information (transmission destination information) indicating each replica computer to be equivalent in the master mode, and a master indicating the master computer that is the replication data transmission source in the replica mode. Computer information (source information) is stored. The replication communication unit 142 is a part that operates in both the master mode and the replica mode. In the master mode, the merge communication data is transmitted as replication data to each replica computer based on the replica computer information. Sometimes, replication connection is made to the master computer based on the master computer information and the replication data is received.

  The mode processing unit 15 is a part that performs management including transmission / reception of mode information, mode management of the own computer, and mode change, and includes a mode management unit 151, a mode transmission unit 152, a mode reception unit 153, and a mode management table 154. ing. Here, the mode information is information including master computer information specifying which computer is the master computer and own computer mode information specifying the mode of the own computer, and is stored in the mode management table 154.

  The mode management unit 151 is a part that operates in both the master mode and the replica mode, manages the mode management table 154, and requires a mode change in response to a mode change request from the mode setting application 18 The mode transition and the operation after the predetermined mode transition are executed. The mode transmission unit 152 and the mode reception unit 153 operate in both the master mode and the replica mode, and transmit / receive mode information to / from other computers.

  The retrieval processing unit 16 is a part that transmits / receives data necessary for retrieval to / from other computers as retrieval data. The retrieval processing unit 161, retrieval data transmission unit 162, retrieval request transmission unit 163, and retrieval data reception unit 164 The retrieval state table 165 is provided.

  Among these, the retrieve request accepting unit 161 and the retrieve data transmitting unit 162 are parts that operate in the master mode. The retrieve request accepting unit 161 accepts the retrieve request from the replica computer, and the retrieve data transmitting unit 162 is the request source. Send retrieve data to the replica computer.

  The retrieve request transmission unit 163 and the retrieve data reception unit 164 operate in the replica mode. The retrieve request transmission unit 163 transmits a retrieve request to the master computer, and the retrieve data reception unit 164 retrieves from the master computer. Receive data. The retrieve state table 165 is used in the replica mode, and stores retrieve state information indicating the state of the retrieve for each data unit transmitted and received as retrieve data. The retrieve state table 165 is updated by the retrieve request transmission unit 163 according to the retrieve state.

  FIG. 2 is a block diagram showing a distributed data equivalent system configured by connecting a plurality of computers shown in FIG. 1 through a communication network. Among a plurality of computers, one specific computer is in a master mode in which the original data of the database 11 can be updated, and the other computers are in a replica mode in which the original data cannot be updated and the master data update contents are duplicated. By operating, it operates as a master computer and a replica computer.

  FIG. 3 is a block configuration diagram showing a system configuration when the plurality of computers shown in FIG. 2 are operated as a master computer and a replica computer. In FIG. 3, only the components used when operating as the master computer and the replica computer among the components shown in FIG. 1 are shown, and the relationship between those components in the master computer and the replica computer is shown. Show.

  Below, the detail of each component 11-16 of the computer shown in these FIGS. 1-3 is demonstrated sequentially.

[2-1. Database]
The data file 111 of the database 11 stores two types of data, a tabular data file and a data status table.

  The tabular data file is a tabular data file composed of rows and columns as shown in FIG. 4A, and one or a plurality of tabular data files are stored in the data file 111. In the following description, the tabular data file is abbreviated as “table” from the viewpoint of simplification. The table is given a unique name for uniquely identifying the table in the data file between computers, and each table is identified by this name. The name may be anything that can uniquely identify the table, and may be a simple character string or symbol string composed of numbers and symbols.

  As shown in FIG. 4B, the data state table is a table in which a serial number value is assigned to each unit of data in a table, a column, or a row. In FIG. As an example, a serial number value is assigned to each column number. This serial number is used as follows to determine whether the data of each unit is the same between computers. That is, when one data value that can be specified by the column number and row number of a table is changed in the data file 111, the serial number value of the data state table corresponding to the table and column number is added to By using different serial number values, it is determined that the data contents indicated by the column numbers in the table are the same if the serial number values are the same between computers, and if they are not the same, it is determined that they are not the same.

  In FIG. 4B, the serial number value is given for each column number. However, the serial number value may be given for each table, each row number, or each table, row number, and column number for management. In this embodiment, as an example, the data body of the data file 111 is a table in a table format. However, as a modification, the format of the data body of the data file 111 is not limited to tabular data, and the addition of data contents Any format can be used as long as it can be deleted, updated, and referenced.

  Further, the mapping data 112 of the database 11 directly maps the tables and data status tables in the data file 111 arranged on the hard disk to virtual memory pages on the main memory from the system and the plurality of data manipulation applications 17. It is a thing. When a plurality of data manipulation applications 17 operate, mapping data 112 is created for each data manipulation application 17.

  In the process of reflecting the data update contents to the database 11, the reflection to the data file 111 is indirectly performed by reflecting the mapping data 112 temporarily without performing the direct reflection process to the data file 111. Reflection to the mapping data 112 mapped to the virtual memory page of the system and the plurality of data manipulation applications 17 is performed. As a modification, the mapping data 112 is not created, and writing to and reading from the table on the data file 111 and the file constituting the data state table are performed each time the update or reference process is performed on the database 11. Good.

[2-2. Data processing section]
As shown in FIG. 5, the update data cache 124 is managed by a table name, a row number, a column number, a serial number value, a changed data value, etc. of the changed data, and is a data operation which operates on the master computer. Created and managed individually for the application 17. When a plurality of data manipulation applications 17 operate in the master computer, an update data cache 124 is created for each data manipulation application 17, and data update processing unit 122 adds, changes, deletes, and references data. Data is referenced by the unit 121.

  When receiving a data reference request from the data operation application 17, the data reference processing unit 121 performs the following data reference processing. That is, when the own computer is operating as a master computer, it is checked whether or not the data of the corresponding table name, row number, and column number exists in the update data cache 124. The data value in the cache 124 is returned. When such update data cache data does not exist when operating as a master computer and when operating as a replica computer, the data values of the row number and column number of the table are returned.

  When the data update processing unit 122 receives a data change request from the data operation application 17 on the own computer operating as the master computer, the data update processing unit 122 stores the change contents (data update operation When a commit request from the data manipulation application 17 is received, the following commit process is performed.

  That is, in the commit process, the data update processing unit 122 first reflects the post-change data value stored in the update data cache 124 to the table, and sets the corresponding table name and column number serial number value of the data status table to “ 1 ”is added, and the added serial number value is stored in the serial number value of the update data cache 124. Thereafter, the update data merge unit 131 is notified together with the contents of the update data cache 124, and the contents of the update data cache 124 are emptied.

  The received data reflection unit 123 performs the following data reflection process when receiving a notification of changes to the database by the master computer from the merge data expansion unit 132 of the own computer operating as a replica computer. That is, in the data reflection process, the reception data reflection unit 123 refers to the retrieve state table 165, and if the change contents can be reflected, reflects the change in the table and updates the serial number value in the data state table. If the change contents cannot be reflected, it is temporarily saved, reflected in the table at the timing when the change can be reflected, and the serial number value in the data status table is updated.

[2-3. Merge data processing unit]
When the update data merging unit 131 receives a notification of the contents of the update data cache 124 from the data update processing unit 122 of the own computer operating as the master computer, Merge processing with the contents of the update data cache being temporarily saved notified to. Here, if there is update data cache content that has been notified before and temporarily stored, merge with the update data cache content, and if there is no such update data cache content that is temporarily stored Temporarily stores the update data cache contents notified this time.

  Such a merge process is performed according to a preset merge merge condition. That is, when the merge merge condition is satisfied, merge data is created by merging the contents of a plurality of update data caches that are temporarily stored, and this merge data is divided into replication data as it is or as needed, A replication data transmission request is notified to the replication communication unit 142, and the contents of the update data cache being temporarily saved are emptied.

  Here, the merge merge condition is a combination of one or more condition items selected from the following various condition items. As various condition items, the data size of the update data cache contents being temporarily saved, the previous merge process (processing until the transmission request notification to the replication communication unit 142 is notified and the update data cache contents being temporarily saved are emptied) ) The number of notifications received from the data update processing unit 122 after the completion, the elapsed time from the completion of the previous merge process, and the like can be considered.

  The merge data expansion unit 132 expands the replication data received from the replication communication unit 142 of the own computer operating as a replica computer into update content commit data that can be reflected in the table, and notifies the received data reflection unit 123 I do.

[2-4. Replication processing section]
As shown in FIG. 6, in the replication management table 143, when the local computer is operating as a master computer, information on one or more replica computers that are data equivalence targets is stored. When the computer is operating as a replica computer, information on the master computer is stored.

  In response to the request from the mode management unit 151, the replication management unit 141 initializes the replication management table 143 and adds master computer information.

  When the replication communication unit 142 operates as a master computer and receives a replication connection from a replica computer, the replication communication unit 142 adds replica computer information indicating the replica computer to the replication management table 143. When a transmission request notification is received from the update data merge unit 131 of the own computer operating as the master computer, the replication data contents are transferred to the replication communication unit 142 of one or more replica computers according to the replication management table 143. Send.

  When the replication communication unit 142 operates as a replica computer, the replication communication unit 142 performs replication connection to the master computer according to the master computer information stored in the replication management table 143, receives replication data from the master computer, and merge data expansion unit. Notification to 132 is performed.

[2-5. Mode processing section]
In the present embodiment, the mode management table 154 includes “own computer mode information” and “master computer information” as shown in FIG. “Own computer mode information” stores mode information of the own computer. The “master computer information” stores information on the master computer when the master computer exists in the network.

  As a modification, only the own computer mode information or the mode information of all computers may be stored. In addition, using simultaneous broadcast communication for mode information transmission, each computer automatically acquires mode information of other computers by sending the mode information to multiple computers simultaneously at the same time. It is also possible to update the mode management table 154 in real time.

  When the mode management unit 151 receives a mode setting change request to the master mode from the mode setting application 18 or receives mode information from another computer through the mode receiving unit 153, the mode management unit 151 changes the mode and mode of the own computer. Change of management table 154, request for transmission of mode information of own computer to mode transmission unit 152, notification of addition of replica computer or notification of change of master computer to replication management unit 141, notification of request for start of retrieval to retrieve request transmission unit 163, etc. I do.

  The mode transmission unit 152 transmits mode information data to the mode reception unit 153 of one or a plurality of computers on the network by using broadcast communication such as broadcast or multicast according to a transmission request from the mode management unit 151. Do.

  The mode reception unit 153 receives the mode information data transmitted from the mode transmission unit 152 of another computer and notifies the mode management unit 151 of the mode information data.

  In the present invention, the mode type handled in the mode management is that at least “master” and “replica” can be identified, but in this embodiment, in addition to the master mode and replica mode in the present invention, these modes are used. The intermediate state is also handled as an independent mode.

  In the following, the master mode and the replica mode are referred to as “master operation” and “replica operation”, respectively, and the transition states to the replica mode and the master mode are respectively “master transition” and “replica transition”. Call. That is, in the present embodiment, four modes of “master operation”, “replica operation”, “master transition”, and “replica transition” are handled as mode types.

  Among these modes, only the “master operation” which is the master mode is a mode in which original data can be updated with respect to the data in the database 11, not simply copying. The replica mode “during replica operation” is a mode in which the original data of the database 11 cannot be updated, and the data that has been originally updated by the master computer can be copied and referenced.

  Since “Master Transition” is a mode in the middle of transition to “Master Operation”, when transitioning to “Master Operation”, it always goes through “Master Transition” and then “Master Operation”. Transition. For example, if a master switch is set to operate as a master computer from an external application in a “replica operating” computer, transition to “master transition” and whether there is another master computer or not After confirming, transition to “Master Operation”. Similarly, “Replica Transition” is a mode in the middle of transition to “Replica Operation”. Therefore, when transitioning to “Replica Operation”, always go through “Replica Transition” and then “Replica Operation”. Transition to “medium”.

  Note that when the computer is in either the “master transition” or “replica transition” mode, the computer is present in the same manner as the replica computer in relation to the application.

  FIG. 8 is a diagram showing transition triggers (triggers and triggering events) and transition destinations in each mode as transition conditions when transitioning from four types of modes to another mode. The content described on the arrow is “trigger”. In FIG. 8, in addition to the four types of modes, the “no database” mode is introduced as a state where there is no database (the state before or after the initialization of the database) for convenience. "Mode" is a mode that does not appear in the mode management table 151. Hereinafter, transition flows in a total of five modes including the “no database” mode will be described with reference to FIG.

  First, transition triggers between five types of modes can be broadly classified into three types: “database start / stop”, “mode setting of external application”, and “reception of mode information from other computers”.

  The first trigger “database start / stop” is a trigger having the “no database” mode as a transition destination or a transition source. The mode immediately before starting the database is “no database”. In this case, when the database 11 is activated, first, the state transits to “in replica transition”. Transition to the “no database” mode is possible from any of the other four modes, and the trigger for that is the shutdown of the database 11.

  The second trigger “mode setting of the external application” is a mode setting by the mode setting application 18 outside the system. When a mode setting operation for setting the own computer as a master computer is performed, if the mode of the own computer is “during replica operation” or “during replica transition”, transition is made to “during master transition”.

  The third trigger “reception of mode information from another computer” is to receive mode information indicating the mode of the computer from another computer. When mode information is received from another computer, the mode of the own computer is changed to a transition destination mode corresponding to the mode information. FIG. 9 is a table showing the transition conditions at the time of such mode reception, and the transition of the own computer when the mode information indicating the mode shown in the upper right column in the current mode shown in the left column is received. The previous mode is shown in each column in the center where the current mode and the reception mode intersect.

  As shown in FIG. 9, when the own computer is in the “master operation” mode and receives the “master transition” mode from another computer, the own computer is finally set to the “replica operation” mode. Therefore, the mode management unit 151 first changes its own computer to the “in transit replica” mode, which is an intermediate mode, and sets the “own computer mode” column of the mode management table 154 as a previous step. Change to "Replica transitioning".

  In addition, when the local computer is in "Replica in operation" mode, if it receives "Master transition in progress" mode from another computer, it will eventually become "Replica in operation" mode, so mode management The unit 151 changes the own computer to the “in replica transition” mode as a previous step for that purpose, and changes the “own computer mode” column in the mode management table 154 to “in replica transition”.

  In this way, when the current mode is the “in replica operation” mode in the situation that should finally be in the “in replica operation” mode, as a modification, without changing to the “in replica transition” mode, Although it is conceivable to maintain the “in replica operation” mode as it is, in this embodiment, from the viewpoint of equalizing the processing when receiving the “master transition” mode from another computer regardless of the current mode, the current mode Even if “Replica is in operation”, it is temporarily changed to “Replica transition”.

  In addition, when the local computer is in "Master Transition" mode, if the "Replica Transition" mode is received from the previous master computer, it is confirmed that the previous master computer is no longer the master computer. Therefore, the mode management unit 151 changes the own computer to the “master operation” mode, and changes the “own computer mode” column of the mode management table 154 to “master operation”.

  On the other hand, when the local computer is in “Replica Transition” mode, if it receives “Master Operation” mode from the new master computer, it is confirmed that a new master computer has been established. Then, the mode management unit 151 changes the own computer to the “in replica operation” mode, and changes the “own computer mode” column of the mode management table 154 to “in replica operation”.

  After performing each mode transition as described above, the mode management unit 151 executes each operation determined according to the new mode after the transition. FIG. 10 shows an operation performed after such mode transition.

  As shown in FIG. 10, when the mode transition is made to “in master operation”, the mode management unit 151 simply transmits only the new mode “in master operation” by the mode transmission unit 152.

  When the mode transition is made to “in replica operation”, the mode management unit 151 transmits a new mode “in replica operation” by the mode transmission unit 152 and the master computer by the replication management unit 141 and the replication communication unit 142. And a retrieval request transmission unit 163 transmits a retrieval request.

  In addition, when the mode transition is made to “in master transition”, the mode management unit 151 transmits a new mode “in master transition” by the mode transmission unit 152 and finally transits to the mode in “master operation”. As a preparation for the operation at that time, replication transmission is started by the replication management unit 141 and the replication communication unit 142, and the retrieve request receiving unit 161 is activated. Here, at the start of replication transmission, the previous replication management table 143 is deleted and a replication connection is awaited.

  When the mode transition is made to “in replica transition”, the mode management unit 151 simply transmits only the new mode “in replica transition” by the mode transmission unit 152.

  As described above, the mode processing unit 15 performs mode transition based on predetermined transition conditions as shown in FIGS. 8 and 9, and after the mode transition, the replication processing unit 14 and the retrieval processing unit 16 In cooperation with each other, an operation after mode transition having a predetermined content as shown in FIG. 10 is executed.

[2-6. Retrieve processing unit]
The retrieve state table 165 stores the retrieve state of each table as shown in FIG.

  The retrieve request accepting unit 161 operates only when the own computer is a master computer. The retrieve request transmitting unit 163 of another computer that is a replica computer sends a “data state table” transmission request or “data about a table column”. ”And“ data state table ”are transmitted to the retrieve data transmission unit 162.

  The retrieve data transmission unit 162 operates only when the own computer is a master computer. From the table and the data state table in the database 11 of the own computer, the retrieve data transmission unit 162 operates according to the request content of the transmission request received by the retrieve request reception unit 161. The data is acquired, and the acquired data is transmitted as “data corresponding to the content of the transmission request” to the retrieve data receiving unit 164 of the request source replica computer.

  The retrieve request transmission unit 163 operates only when the local computer is a replica computer, and when receiving a notification from the mode management unit 151, the merge data expansion unit 132, and the reception data reflection unit 123 of the local computer, A transmission request for “status table” or a transmission request for “data relating to table columns” and “data status table” is sent to the retrieval data transmission unit 162 of the master computer, and the retrieval status table 165 is updated.

  When the transmitted transmission request is a “data state table transmission request”, the retrieve request transmission unit 163 transmits a “data state table” transmitted as “data corresponding to the content of the transmission request” from the requesting master computer. The data is received through the retrieve data receiving unit 164 and compared with the entire data state table of the database 11 of the own computer. When “table, column” with different serial number values is obtained as a result of this data state table comparison, a transmission request for “data on table column” and “data state table” for that “table, column” To the retrieve data transmission unit 162 of the master computer, and the retrieve state table 165 is updated.

  The retrieve request transmission unit 163 also sends a request request master if the transmission request made is a transmission request for “data relating to table columns” and “data state table” for “table, column” as described above. A completion notification including such data is received from the retrieve data transmission unit 162 of the computer through the retrieve data reception unit 164, and the retrieval status table 165 is updated.

  When the content received from the retrieve data transmission unit 162 of the requested master computer is the “data state table”, the retrieve data reception unit 164 notifies the retrieval request transmission unit 163 of the content. On the other hand, if the received contents are “data related to table columns” and “data status table” for “table, column”, they are reflected in the table and data status table in the database 11 of the own computer. The completion request is notified to the retrieval request transmission unit 163.

[3. Initial startup / mode change processing]
FIG. 12 is a flowchart showing an outline of the operation from the initial activation by each computer of the distributed data equivalent system shown in FIGS. 1 to 3 to the system stop.

  As shown in FIG. 12, each computer performs “initial startup processing” in cooperation with the mode processing unit 15, the replication processing unit 14, and the retrieval processing unit 16 (S 1201). The “replica operation” mode is set, and the system operates as a replica computer (S1202). After that, when the master setting of the own computer is performed by the mode setting application 18 (YES in S1203), the mode processing unit 15, the replication processing unit 14, and the retrieval processing unit 16 cooperate with each other, By performing “mode change processing” (S1204), the master operation mode is set, and the system operates as a master computer (S1205).

  Further, when the master setting of another computer is performed (YES in S1206), the mode processing unit 15, the replication processing unit 14, and the retrieval processing unit 16 cooperate to “mode change processing at the time of other computer master setting”. (S1207), the replica operation mode is set, and the system operates as a replica computer (S1208). Each computer repeats the mode change process (S1203 to S1208) according to the master setting of the own computer or another computer until the database 11 of the own computer stops (NO in S1209), and the time when the database 11 of the own computer stops The operation ends in (YES in S1209).

  In the following, details of the initial startup process (S1201), the mode change process at the time of setting the own computer master (S1204), and the mode change process at the time of setting the other computer master (S1207) among the operations shown in FIG. .

[3-1. Initial startup process]
FIG. 13 is a flowchart showing an outline of the processing procedure of the initial activation process (S1201) shown in FIG. This processing is performed in cooperation with the replication management unit 141 and the retrieve request transmission unit 163 with the mode management unit 151 as a processing subject.

  At the time of initial startup, since the computer is started as a replica computer, the first own computer mode in the mode management table 154 is set to replica transition. The first mode at the time of initial activation may be any of the four modes. However, from the viewpoint of simplifying the processing flow, the case where the first mode is “during replica transition” will be described in the present embodiment.

  First, the mode management unit 151 checks whether or not a master computer exists in the communication network (S1301). Specifically, as a process for performing this confirmation, the mode management unit 151 transmits an “inquiry” as to whether or not the master computer exists through the mode transmission unit 152 to the network by multicast or broadcast, for a certain period of time. Wait for a response.

  In response to this “inquiry”, if there is a computer operating as a master computer in the communication network, the “inquiry” is received by the mode receiving unit 153 on the master computer, and mode management on the master computer is performed. Notification is made to the unit 151. In this case, the mode management unit 151 on the master computer transmits information indicating that the own computer is a master computer through the mode transmission unit 152. When the inquiry source computer receives the master computer information by the mode receiving unit 153 and receives the notification by the mode management unit 151, the presence of the master computer is confirmed.

  If there is no master computer (NO in S1302) as a result of such a master computer confirmation process (S1301), the mode management unit 151 automatically executes until the own computer or another computer on the communication network becomes the master computer. The own computer mode information in the computer mode and mode management table 154 is maintained as “in replica transition”, and operates as a replica computer (S1303).

  On the other hand, if there is a master computer (YES in S1302), the mode management unit 151 sets the master computer information in the mode management table 154 (S1304), and the mode management unit 151 sends the retrieval request transmission unit 163. The retrieval request transmission unit 163 initializes the retrieval status table 165 (S1305). In the initialization of the retrieval status table 165, the retrieval request transmission unit 163 sets the status of the retrieval status table 165 to “not retrieved” for all the tables in the data file 111.

  After completing the initialization (S1305) of the retrieve state table 165, the replication manager 141 notifies the replication manager 141 of the master computer information, and the replication manager 141 performs processing for starting replication reception. (S1306).

  In this case, the replication management unit 141 stores the master computer information notified from the mode management unit 151 in the replication management table 143 as a process for receiving replication, and the replication communication unit 142 sends the master computer information to the master computer. Perform replication connection processing. The replication connection process is to notify the master computer that the local computer is a target for receiving replication. As a result, the replication communication unit 142 can receive replication from the master computer.

  As a result of the replication reception start processing (S1306), after the replication reception becomes possible, the retrieval request transmission unit 163 notifies the retrieval request transmission unit 163, and the retrieval request transmission unit 163 retrieves the entire database. Is performed (S1307).

  FIG. 14A is a flowchart showing the processing procedure of the retrieval processing (S1307) of the entire database. This processing is performed with the retrieval request transmission unit 163 as a processing subject.

  First, a data state table is acquired from the master computer (S1401), each table is compared (S1402), and is compared with the data state table of its own computer (S1403). (YES in S1404) sets the state of the corresponding table in the retrieve state table 165 to “Retrieved” (S1405), and for other tables having columns with different version number values (NO in S1404) ”Retrieve request processing” is performed (S1406).

  FIG. 14B is a flowchart showing the processing procedure of this “table / column retrieval request processing” (S 1406). First, the table for which the request has been received is set to “retrieving” (S14061), and a transmission request is made to the master computer together with information about columns having different version numbers for each table (S14062).

  The retrieval request receiving unit 161 of the master computer that has received this transmission request notifies the retrieval data transmission unit 162, and the data corresponding to the content of the transmission request is transmitted as retrieval data. "Retrieve data according to the content" is received by the retrieve data receiving unit 164 of the computer requesting the retrieval (S14063). As a result, as the “retrieve execution process”, the retrieve data receiving unit 164 collectively reflects the retrieve data on the table and the data state table, and the retrieve request transmission unit 163 is notified of the retrieval completion (S14064).

  In the retrieval processing of the entire database shown in FIG. 14A, after a series of processing steps (S1402 to S1407) are completed for all tables, a retrieval completion notification is sent from the retrieval data receiving unit 164 by the retrieval request transmission unit 163. Receive (S1410). As a result, the retrieval request transmission unit 163 sets the state of the corresponding table in the retrieval state table 165 to “retrieved” (S1411), and when all the tables are “retrieved” (“YES” in S1408). The retrieval process for the entire database is completed (S1409).

  In the “initial start-up process” shown in FIG. 13, after the “retrieve process for the entire database” (S1307) shown in FIG. 14A is completed, the mode management unit 151 uses the own computer mode and the mode management table 154. The computer mode information is changed to “in replica operation” (S1308), and thereafter, the own computer maintains the “in replica operation” mode until a master change occurs and operates as a replica computer (S1309).

  Note that “replication connection” in “replication reception start processing” (S1306) in FIG. 13 is not a connection in the communication layer (IP layer) but a replication management layer (replication processing unit 14) in this system. Connection. In addition, IP layer TCP or UDP (multicast, broadcast) is used for replication communication. When TCP, which is connection-type IP communication, is used, connection processing at the TCP level is performed in synchronization with replication connection processing.

[3-2. Mode change process when setting own computer master]
FIG. 15 is a flowchart showing an outline of the processing procedure of the mode change process (S1204) at the time of setting the own computer master shown in FIG. This processing is performed in cooperation with the replication management unit 141 and the retrieve request reception unit 161 with the mode management unit 151 as a processing subject.

  Note that the specific processing procedure of the “mode change process when setting the own computer master” differs depending on whether the own computer is a master computer or a replica computer. If the computer is a replica computer, the current mode is Depends on whether replica operation or replica transition is in progress. Below, the processing procedure according to these situations will be described.

  When there is no master computer in the communication network at the time of initial startup, the own computer mode information in the mode management table 154 is set to “in replica transition” in each computer, and the master computer information is empty. . On the other hand, if a star computer exists at the time of initial startup, the own computer mode information in the mode management table 154 is set to “in replica operation” in other computers, and the master computer information is currently operating as a master computer. Stores information about the computers that are running.

  Further, after the mode change process is performed by one or more master settings after the initial startup, in each computer, the own computer mode information in the mode management table 154 includes “master operation”, “replica operation”, It is set to either “master transition in progress” or “replica transition in progress”, and the master computer information stores information on a computer currently operating as a master computer. Here, when the own computer is operating as a master computer, the own computer mode information in the mode management table 154 is set to “master operation in progress”, and information on the own computer is stored in the master computer information. .

  In such various situations, when the master setting of the own computer is performed by the mode setting application 18 (S1501), processing is performed by the mode management unit 151. The mode management unit 151 refers to its own computer mode information in the mode management table 154 and confirms whether it is a master computer (S1502). When the own computer is a master computer (YES in S1502), the master setting process is completed as it is.

  If the own computer is not the master computer (NO in S1502), the own computer mode information in the mode management table 154 is changed to “during master transition” (S1503).

  After changing the own computer mode information to “in master transition”, the mode management unit 151 transmits the information in master transition by the mode transmission unit 152 (S1504). In transmitting the master transition information, the master transition information is transmitted to the network by multicast or broadcast as its own computer mode information.

  Next, the mode management unit 151 checks whether or not there is a master computer (current master computer) currently operating in the “master operation” mode in the mode management table 154 of the own computer (S1505). If there is a master computer (YES in S1505), it waits for a certain period of time until the replica transition information is transmitted from the current master computer, and the mode reception unit 153 receives the replica transition information from the current master computer. Receive (S1506). When the mode management unit 151 receives the replica transition information from the current master computer, or when the current master computer does not exist (NO in S1505), the mode management unit 151 adds the information of the own computer to the master computer information in the mode management table 154. Is stored (S1507).

  After that, the mode manager 151 notifies the replication manager 141 of its own computer master information, and the replication manager 141 performs replication transmission start processing (S1508). Specifically, as the replication transmission start processing, the replication management unit 141 initializes the replication management table 143. When the replication communication unit 142 is in replication connection with the master computer, the replication management unit 141 performs disconnection processing. Enables reception of replication connection requests from computers. By performing such a replication transmission start process, the master computer can receive a replication connection request from the replica computer, establish a connection, and transmit replication data generated thereafter.

  Next, in response to a notification from the mode management unit 151, the retrieval request transmission unit 163 is stopped (S1509). If there is a process that is currently requested, the retrieve request transmission unit 163 stops the process.

  Subsequently, in response to a notification from the mode management unit 151, the retrieval request reception unit 161 is started (S1510). The retrieve request accepting unit 161 can respond to the retrieve request from the replica computer by performing the start process.

  Thereafter, the mode management unit 151 changes its own computer mode information in the mode management table 154 to “master operation” (S1511), and the mode transmission unit 152 transmits the master operation information (S1512). Start operation as a computer.

[3-3. Mode change processing when setting other computer master]
16 and 17 are flowcharts showing an overview of the processing procedure of the mode change processing (S1207) at the time of other computer master setting shown in FIG. 12, and FIG. 16 shows the first half until the transition to the “in replica transition” mode. FIG. 17 shows the latter half until the transition to the “in replica operation” mode. This processing is performed in cooperation with the replication management unit 141 and the retrieve request transmission unit 163 with the mode management unit 151 as a processing subject.

  Note that the specific processing procedure of the “mode change process when setting the own computer master” differs depending on whether the own computer is a master computer or a replica computer. If the computer is a replica computer, the current mode is Depends on whether replica operation or replica transition is in progress. Below, the processing procedure according to these situations will be described.

  When there is no master computer in the communication network at the time of initial startup, the own computer mode information in the mode management table 154 is set to “in replica transition” in each computer, and the master computer information is empty. . On the other hand, if a star computer exists at the time of initial startup, the own computer mode information in the mode management table 154 is set to “in replica operation” in other computers, and the master computer information is currently operating as a master computer. Stores information about the computers that are running.

  Further, after the mode change process is performed by one or more master settings after the initial startup, in each computer, the own computer mode information in the mode management table 154 includes “master operation”, “replica operation”, It is set to either “master transition in progress” or “replica transition in progress”, and the master computer information stores information on a computer currently operating as a master computer. Here, when the own computer is operating as a master computer, the own computer mode information in the mode management table 154 is set to “master operation in progress”, and information on the own computer is stored in the master computer information. .

  When the master setting is performed in one of the other computers, the above-described “mode change process at the time of own computer master setting” is performed in the computer, and the information during master transition is transmitted (S1504).

  In the own computer, when the mode receiving unit 153 receives such master transition information from another computer (S1601), the mode management unit 151 refers to the own computer mode information in the mode management table 154, and It is checked whether the mode is “master operation” or “replica operation” (S1602).

  As a result of the confirmation, when the mode of the computer is neither “master operation” nor “replica operation”, that is, “master transition in progress” or “replica transition in progress” (NO in S1602), The process is completed as it is. For example, in the process when the master computer does not exist at the time of initial startup, the own computer is “in replica transition”, but at this time, the mode change process is performed even if master transition information from another computer is received. Never do.

  If the mode of the computer is “master operation” or “replica operation” (YES in S1602) as a result of the confirmation, the mode management unit 151 notifies the replication management unit 141 and the replication management unit 141, a replication stop process is performed (S1603). Specifically, as the replication stop process, the replication management unit 141 initializes the replication management table 143, and performs a disconnection process when the replication communication unit 142 is in replication connection with the master computer. If a replication connection is received from, disconnect processing is performed. By performing such a replication stop process, the replication transmission / reception process does not occur until the replication process is started.

  Next, the mode management unit 151 determines whether or not the mode of the local computer is “in master operation” (S1604). When the mode of the local computer is “master operation in progress” (YES in S1604), the retrieval request accepting unit 161 is stopped (S1605) in response to a notification from the mode management unit 151. At this time, the retrieve request accepting unit 161 stops the currently accepted request process when the retrieve data transmitting unit 162 is executing the process.

  If the mode of the local computer is not “master operation” (NO in S1604), the retrieval request transmission unit 163 is stopped (S1606) by a notification from the mode management unit 151. If there is a process that is currently requested, the retrieve request transmission unit 163 stops the process.

  Next, the mode management unit 151 changes the own computer mode information in the mode management table 154 to “in replica transition” (S1607). Since the mode information before the change is also necessary for the subsequent processing, it is stored at the time before the change.

  The mode management unit 151 changes the mode of its own computer to “in replica transition”, and then transmits information during replica transition by the mode transmission unit 152 (S1608). In the replica transition information transmission process, replica transition information is transmitted to the network by multicast or broadcast. Here, when the mode before the change of the own computer is “master operation”, the new master computer is waiting to receive the replica transition information.

  After performing the replica transition information transmission processing (S1608), the mode management unit 151 waits for reception of master operating information from the new master computer.

  Thereafter, when the master operating information is transmitted from the new master computer which is the transmission source of the master transition information by the “mode change processing at the time of setting the own computer master” of the computer (S1512), it is shown in FIG. As described above, the own computer receives master operating information from the new master computer by the mode receiving unit 153 (S1701). In this case, the mode management unit 151 changes the master computer information of the mode management table 154 of the own computer to the information of the new master computer (S1702), initializes the retrieve state table 165 (S1703), and replicates. A reception start process (S1704) and a retrieve process for the entire database (S1705) are performed. After the retrieve process for the entire database is completed, the local computer mode information is changed to “in replica operation” (S1706), and the system operates as a replica computer. (S1707).

  Here, a series of processes (S1702 to S1707) from the setting of the master computer information to the operation as a replica computer are the same as the series of processes (S1304 to S1309) in the initial startup process of FIG.

[4. Data change / reflection process]
FIG. 18 is a flowchart showing an outline of the distributed data equivalent processing procedure from the data change process of the master computer to the data reflection process of the replica computer in the distributed data equivalent system shown in FIGS.

  As shown in FIG. 18, in the master computer operating in the “in master operation” mode (S1205), the data change unit 12 stores the data change contents stored in the update data cache 124 by the data processing unit 12. Processing (S1801) is performed, and commit processing (S1802) is performed in which the changed content is reflected in the database 11 as update content commit data. Next, the update data merging unit 131 performs merge processing (S1803) for merging the update data cache content that is update content commit data in accordance with the merge merge condition, and then, the replication processing unit 14 converts the merged data into the merged data. A replication transmission process (S1804) is performed to transmit the replication data to the replica computer.

  Thereafter, in the replica computer operating in the “in replica operation” mode (S1208), the replication processing unit 14 performs a replication reception process (S1811) for receiving the replication data transmitted from the master side. Next, the merge data processing unit 13 performs merge data expansion processing (S1812) for expanding the replication data into update content commit data that can be reflected in the database 11, and subsequently, the update content expanded by the data processing unit 12 By performing the data reflection process (S1813) for reflecting the commit data in the database 11 of the own computer, it is equivalent to the same content as the database 11 of the master computer.

  Details of each process shown in FIG. 18 will be sequentially described below.

[4-1. Data change processing]
FIG. 19 is a flowchart showing an outline of a data change processing procedure by the data processing unit 12 of the master computer shown in FIG.

  As shown in FIG. 19, when a data change request is made by the data manipulation application 17 (S1901), the data update processing unit 122 stores the change contents corresponding to the data change request in the update data cache 123. Data change processing is performed. Here, as an example, the data change request is “table name of table” of the table to be changed, “row number in the table”, “column number in the table”, “data value after change” of the data to be changed. Shall be given by

  First, the data update processing unit 122 refers to the own computer mode information in the mode management table 154 to check whether the own computer is a master computer operating in the “master operation” mode (S1902). . If the own computer is not the master computer (NO in S1902), the data change process is terminated as it is.

  When the own computer is a master computer (YES in S1902), the data update processing unit 122 uses the “table name of the table”, “row number in the table”, and “column number in the table”. Then, it is confirmed whether or not the cache data corresponding to the data change request already exists in the update data cache 123 (S1903). If the corresponding cache data exists (YES in S1903), the “changed data value” of the corresponding cache data is updated (S1905).

  If there is no cache data corresponding to the data change request in the update data cache 123 (NO in S1903), the cache data is newly added to the update data cache 123, and the added cache data is used as information. , “Table name of table”, “row number in table”, “column number in table”, “data value after change” are stored (S1904).

[4-2. Change commit process]
FIG. 20 is a flowchart showing an outline of a change commit procedure performed by the data processor 12 of the master computer shown in FIG.

  As shown in FIG. 20, when a change commit request is made by the data manipulation application 17 (S2001), whether or not cache data is stored in the update data cache 123 by the data update processing unit 122. (S2002). When cache data is stored in the update data cache 123 (YES in S2002), the following series of processes (S2004 to S2006) are performed for each cache data (S2003).

  That is, for each cache data (S2003), the “changed data value” is reflected in the data value of the row number and column number of the corresponding table name in the data file 111 (S2004), and the corresponding data state table 112 The serial number value of the table name and column number to be added is added (S2005), and the serial number value after the addition is stored in the “serial number value” of the cache data (S2006).

  After a series of processing (S2004 to S2006) has been completed for all cache data (S2007), the update data merge unit 131 is notified of the contents of the update data cache 123 at that time, and a merge request is made. Request processing "(S2008) is performed, and subsequently, the cache data in the update data cache 123 is made empty (S2009).

[4-3. Commit data merge process]
FIG. 21 is a flowchart showing an outline of the commit data merge processing procedure by the merge data processing unit 13 of the master computer shown in FIG.

  Here, the merge merge condition for the timing of merging the contents of a plurality of cache data that has received a merge request and making a transmission request to the replication communication unit 142 is a combination of the following condition items: It shall be. In other words, the merge merge condition is “the total data size of a plurality of cache data currently merged”, “the time since the transmission request to the previous replication communication unit 142”, “the transmission request to the previous replication communication unit 142” It is assumed that any one of a plurality of items “the number of times the merge request has been received since the execution of“ is exceeded the set threshold ”is set.

  As shown in FIG. 21, in the merge process, the update data merge unit 131 “waits for reception” with a timeout for the merge request of the update data cache contents from the data update processing unit 122 (S2101). When “waiting for reception” is canceled due to the occurrence of an event (S2102), it is determined whether the event that occurred is a timeout or a merge request (S2103).

  If a timeout has occurred (NO in S2103), it is determined whether there is merged data currently being saved in the temporary save area (S2104). Note that “data being merged” refers to merged data obtained by merging the update data cache contents, especially data that is being processed before being determined as data for a replication transmission request. This means data that may be merged with other update data cache contents based on conditions.

  If there is no merged data currently stored in the temporary storage area (NO in S2104), the number of request receptions and timeout time are reset (S2106), and the merge request reception waiting is again performed (S2102). If merge data currently being saved exists in the temporary save area (YES in S2104), the merged data being temporarily saved is determined as replication transmission request data and replicated to the replication communication unit 142. A transmission request is made (S2105). Subsequently, the merge data for which transmission has been requested, the number of request receptions, and the timeout time are reset (S2106), and a merge request reception wait is again performed (S2101).

  On the other hand, if a merge request is received (YES in S2103), it is determined whether there is currently merging data being stored in the temporary storage area (S2107). If there is merged data that is currently saved in the temporary save area (YES in S2107), the merged data that is currently saved is merged with the contents of the update data cache that has received this request (S2108). . If there is no merged data currently stored (NO in S2107), the requested update data cache content is stored in the temporary storage area as merged data (S2109).

  After that, based on the merge merge condition, “whether the size of the merged data currently being stored exceeds the threshold” (S2110), “the number of requests received to configure the merged data currently being stored is the threshold Is exceeded ”(S2111). If any threshold is exceeded (YES in S2110 or YES in S2111), a replication transmission request (S2105) is made to the replication communication unit 142, and the request is made. The merged data currently being stored, the number of requests received, and the timeout period are reset (S2106), and the merge request reception wait is again performed (S2101).

  In the above determination based on the merge merge condition, if neither threshold is exceeded (NO in S2110 and NO in S2111), a timeout time is set (S2112), and a merge request reception wait is again performed (S2102). .

  In the merge processing as described above, when the replication transmission request (S2105) is performed, the confirmed original merge data may be used as it is. However, as another method, the original merge data is divided or Data obtained by compression may be used. That is, the data obtained by dividing the content of the original merge data into a plurality of data, or the data obtained by compressing the original merge data using a data compression means such as Huffman coding, or the division thereof, The replication transmission request may be made using any one of the data obtained by executing both of the compressions.

[4-4. Replication send / receive processing]
FIG. 22 is a flowchart showing an overview of a replication transmission processing procedure and a replication reception processing procedure by the replication processing unit 14 of the master computer and the replica computer shown in FIG.

  As shown in FIG. 22, the replication management unit 141 of the master computer is waiting for a replication transmission request (S2201). When receiving a replication transmission request from the update data merging unit 131 of its own computer, the replication management table 143 is updated. Referring to this, it is determined whether or not there is a replica computer that is a transmission destination of replication data (S2202).

  If there is no destination replica computer (NO in S2202), the contents of the requested data are discarded and the transmission request is waited again (S2201). If there is a destination replica computer (“YES” in S2202), the replication communication unit 142 refers to the information of the destination replica computer in the replication management table 143 and the replication communication of the destination replica computer. Data is transmitted to the unit 142 (S2203).

  On the other hand, the replication management unit 141 of the replica computer waits for reception of replication data from the master computer (S2211), and receives replication data from the master computer (S2212). The merge data expansion request is sent to the merge data expansion unit 132 (S2213).

[4-5. Merge data expansion / data reflection processing]
FIGS. 23 and 24 are flowcharts respectively showing an outline of the merge data expansion processing procedure and the data reflection processing procedure by the merge data processing unit 13 and the data processing unit 12 of the replica computer shown in FIG.

  As shown in FIG. 23, when the merge data expansion unit 132 receives a merge expansion request from the replication communication unit 142 (S2301), the merge data content that is the replication data received by the replication communication unit 142 is displayed as individual update content. The update contents to be expressed are expanded into commit data (S2302). Next, a data reflection request is made to the received data reflection unit 123 for the expanded update content commit data (S2303).

  As shown in FIG. 24, when receiving the data reflection request from the merge data expansion unit 132, the reception data reflection unit 123 performs a process on each update content commit data (S2401), and sets the update content commit data during the processing. The retrieval state of the corresponding table is confirmed in the retrieval state table 165 (S2402), and it is determined whether or not it is “retrieved” (S2403).

  If the retrieved state of the table corresponding to the update content commit data being processed is not “retrieved” (NO in S2403), the update content commit data is stored in the temporary storage area (S2404), and then the next update content commit is performed. Process the data.

  Further, when the retrieved state of the corresponding table is “retrieved” (YES in S2403), the received data reflecting unit 123 confirms “update content commit data for the corresponding table” in the temporary storage area. (S2406), it is determined whether or not there is commit data for update contents being saved for the corresponding table (S2407).

  If the update content commit data being saved exists for the corresponding table (YES in S2407), the update content commit data being saved is referred to the data status table in the mapping data 112, and the data status table The serial number value of the table and column to be stored is compared with the serial number value of the update content data being saved (S 2411), and it is determined whether the update content commit data being saved can be reflected in the mapping data 112. This is performed (S2412).

  If the saved update content commit data can be reflected in the mapping data 112 (YES in S2412), the data value of the saved update content commit data is reflected in the corresponding table in the mapping data 112 (S2413), and the data The serial number values in the state table are added (S2414).

  Also, if the update content commit data being saved cannot be reflected in the mapping data 112 (NO in S2412), the "update content commit data being saved" and the "update content commit data being processed" that are individual processing targets are included. (S2415), “retrieve request processing of table and column” is performed for the corresponding table and column (S2416), and the next update content commit data is processed.

  On the other hand, when the update content commit data being saved is reflected (S2413) and the subsequent addition of the serial number value is completed (S2414), or the update content data being saved for the corresponding table does not exist in the temporary save area ( In step S2407 (NO), the following series of processing (S2421 to S2426) is performed on “in-process update content commit data” that is an individual processing target.

  That is, for the update content commit data being processed, the serial number value is compared with the data state table in the mapping data 112 (S2421), whether or not it can be reflected is determined (S2422), and can be reflected ( In S2422 (YES), the data value is reflected in the mapping data 112 (S2423), the serial number value in the data state table is added (S2424), and the next update content data is processed.

  In addition, if the update content commit data being processed cannot be reflected (NO in S2422), the commit information for the update content being processed is discarded, and if there is a saved update content commit data that has been reflected, the update content commit data is stored. The middle update content commit data is discarded (S 2425), “table / column retrieve request processing” is performed (S 2426), and the next update content commit data is processed.

  A series of processes (S2421 to S2426) for such “in-process update content commit data” is the same as the above-described series of processes (S2411 to S2416) for “stored update content commit data”.

  In the data reflection process as described above, when individual update content commit data is in units of tables or columns of tables, in order to perform data reflection processing at high speed, Individual data reflection processes may be executed in parallel.

  The retrieval status table 165 is confirmed, and the data stored in the temporary storage area because the retrieval is in progress is received data when the retrieval is completed in the retrieval request transmission unit 163. You may make it notify to the reflection part 123. FIG. In this case, the reception data reflecting unit 123 checks whether or not the data of the table that has been retrieved is stored in the temporary storage area. By performing S2411 to S2414 or S2416), it is possible to reflect the data being saved immediately after the retrieval is completed.

  Alternatively, as shown in FIG. 21, when a replication transmission request (S2105) is made by the update data merge unit 131 of the master computer, a merge number assigned to each merge data is added to the merged merge data. The merge number may be checked with respect to the received merge data by any of the processing units 12 to 14 of the replica computer. In this configuration, when a communication failure occurs, the received data reflection unit 123 of the replica computer notifies the retrieval request transmission unit 163, and performs a data retrieval request process according to the merge number check result, thereby merging. It is also possible to recover from a failure without having to compare the data status table and the update content data for each update content data that has been expanded.

[5. effect]
According to the present embodiment as described above, the following effects can be obtained.

(1) In the master computer, when the database is updated by performing the data change process and the commit process with respect to the data update operation results by the plurality of data operation applications, the merged data of the plurality of updated contents of the changed contents is merged. After the merged data is split and compressed, it is sent for replication. As a result, it is possible to reduce the load on the communication network as compared with the case of performing replication transmission for each commit process, and as a result, it is possible to improve the throughput when reflecting the data update contents to a plurality of replica computers. it can.

(2) In replica computers, when replicating data that has been split / compressed after receiving replication is expanded into multiple commit data for changes, and the commit data for changes is reflected in the database, it is reflected using the data status table. Check if the change is possible, and if necessary, retrieve the table and column data from the master computer. Thereby, even when a failure occurs temporarily in the communication network and the replication data is lost, the replication data in the database of the master computer can be reliably maintained.

(3) When retrieving is performed at initial startup and master switching, a data state table is acquired from the master computer and compared with the data state table of the local computer, and only the table and column data that need to be retrieved are mastered. Acquire data from the computer. As a result, when there is little difference in data contents with the master computer, the retrieval processing time at the initial startup and master switching can be greatly shortened as compared with the case of acquiring the entire contents of the database.

(4) When master switching is performed, the master computer is in the master operation state during the master transition, and the replica computer is in the replica computer state during the replica transition due to the mode management. In the new master computer, when the replica transition information from the current master computer is received, transition to master operation is made. The replica computer shifts to replica operation when it receives master operation information from the new master computer. Thereby, it is possible to always keep only one master computer existing in the network. In the unlikely event that there are multiple master computers due to a network failure, the replica computer always reflects only the replication data from the master computer recognized by its own computer, so it is inconsistent with the database. Will not occur.

(5) In addition, even when a situation occurs in which there are multiple master computers, one of the computers can transmit the status of master transition and master operation by mode management. Since all transitions to the replica computer, retrieval processing of the entire database, and retrieval of the database of the master computer can be performed, recovery processing in the event of such a failure can be easily performed.

[6. Other Embodiments]
It should be noted that the present invention is not limited to the above-described embodiments, and various other variations can be implemented within the scope of the present invention. For example, the system configuration and the functional configuration of the computer shown in the drawings are merely examples, and a specific system configuration and functional configuration can be selected as appropriate. Further, the data configuration shown in the drawings is merely an example, and a specific data configuration can be selected as appropriate. Similarly, the flowcharts shown in the drawings are merely examples, and specific operation procedures and details of each process can be selected as appropriate.

  For example, in the above-described embodiment, the case where the comparison of the data state table (data specifying information) at the time of the retrieval request is performed on the replica computer side has been described, but by transmitting the data state table from the replica computer to the master computer, The master computer side may compare the data state table and specify the retrieve data to be transmitted according to the comparison result.

The block block diagram which shows each computer which comprises the distributed data equivalent system which concerns on one Embodiment to which this invention is applied. The block block diagram which shows the distributed data equivalent system comprised by connecting the some computer shown in FIG. 1 with a communication network. FIG. 3 is a block configuration diagram showing a system configuration when a plurality of computers shown in FIG. 2 are operated as a master computer and a replica computer. The figure which shows an example of the data structure of the tabular data file (a) and data status table (b) which are stored in the data file shown in FIG. The figure which shows an example of the data structure of the update data cache shown in FIG. The figure which shows an example of the data structure of the replication management table | surface shown in FIG. The figure which shows an example of the data structure of the mode management table shown in FIG. FIG. 4 is a diagram illustrating a mode transition trigger and a transition destination as transition conditions when transitioning from four types of modes managed by the mode management unit illustrated in FIG. 3 to another mode. The figure which shows the transition destination mode at the time of receiving mode information in the present mode as a transition condition at the time of mode reception among the transition conditions shown in FIG. The figure which shows each operation | movement determined according to the new mode after a transition, after performing each mode transition on the transition conditions shown in FIG. 8, FIG. The figure which shows an example of the data structure of the retrieve state table | surface shown in FIG. The flowchart which shows the operation | movement outline | summary from the initial starting by each computer of the distributed data equivalent system shown in FIGS. 1-3 to a system stop. The flowchart which shows the outline | summary of the process sequence of the initial starting process shown in FIG. FIG. 14A is a flowchart showing a processing procedure of a retrieval process for the entire database shown in FIG. 13, and FIG. 14B is a flowchart showing a processing procedure of a table / column retrieval request process in FIG. The flowchart which shows the outline | summary of the process sequence in the mode change process at the time of the own computer master setting shown in FIG. The flowchart which shows the outline | summary of the first half part until it changes to the replica transition mode among the process steps in the mode change process at the time of other computer master setting shown in FIG. The flowchart which shows the outline | summary of the latter half part until it changes to replica operation mode among the process steps in the mode change process at the time of other computer master setting shown in FIG. The flowchart which shows the outline | summary of the distributed data equivalent process sequence from the data change process of a master computer to the data reflection process of a replica computer in the distributed data equivalent system shown in FIGS. The flowchart which shows the outline | summary of the data change process procedure by the data processing part of the master computer shown in FIG. FIG. 4 is a flowchart showing an outline of a change processing commit procedure performed by a data processing unit of the master computer shown in FIG. 3. FIG. The flowchart which shows the outline | summary of the merge processing procedure after the commit execution by the merge data processing part of the master computer shown in FIG. 4 is a flowchart showing an overview of a replication transmission processing procedure and a replication reception processing procedure by a replication processing unit of the master computer and the replica computer shown in FIG. 3. The flowchart which shows the outline | summary of the merge data expansion | deployment processing procedure by the merge data process part of the replica computer shown in FIG. 4 is a flowchart showing an outline of a data reflection processing procedure by a data processing unit of the replica computer shown in FIG. 3.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 ... Database 12 ... Data processing part 13 ... Merge data processing part 14 ... Replication processing part 15 ... Mode processing part 16 ... Retrieve processing part 17 ... Data operation application 18 ... Mode setting application 111 ... Data file 112 ... Mapping data 121 ... Data Reference processing unit 122 ... data update processing unit 123 ... received data reflection unit 123 ... update data cache 131 ... update data merge unit 132 ... merge data expansion unit 141 ... replication management unit 142 ... replication communication unit 143 ... replication management table 151 ... mode Management unit 152 ... mode transmission unit 153 ... mode reception unit 154 ... mode management table 161 ... retrieve request reception unit 162 ... retrieve data transmission unit 163 ... retrieve request transmission unit 164 ... retrieval Budeta receiving section 165 ... retrieve the state table

Claims (14)

By storing target data in multiple computers connected via a network, operating one specific computer in the master mode and each other computer in the replica mode to operate as a master computer and a replica computer, respectively. In the distributed data equivalent method in which the target data of each replica computer is equivalent to the same content as the master computer when the target data is updated to the original by the data update operation on the computer,
A database comprising a data file for storing the target data in a shared area accessible from a plurality of applications;
Manages mode information including mode information specifying which computer is the master computer of the plurality of computers and mode information specifying the mode of the own computer, and transmits / receives mode information to / from other computers In addition, when mode switching is necessary, a mode processing step is executed to execute the mode transition and the operation after the predetermined mode transition,
By the master computer,
Data that temporarily stores the result of the data update operation in the update cache, sets the content of the update cache as one update content commit data in response to a commit request, adds unique data specifying information, and stores the data in the data file An update step;
An update data merging step of generating merge data by merging one or a plurality of the update content commit data determined in accordance with a preset merge merge condition;
Based on the mode information, a transmission destination step is performed for managing transmission destination information indicating each replica computer to be equivalent, and transmitting the merge data as replication data to each replica computer based on the transmission destination information. ,
By the replica computer,
Based on the mode information, managing the transmission source information indicating the master computer that is the transmission source of replication data, and receiving a replication data from the master computer based on the transmission source information,
A merge data expansion step of expanding the merge data that is the replication data into the original one or a plurality of the update content commit data;
The update contents determined to be reflected by determining whether or not to reflect the expanded update content commit data to the data file based on the retrieve status information indicating the retrieve status of the data for each update content commit data Performing a received data reflecting step of reflecting the commit data in the data file together with the data specifying information;
By the replica computer,
Retrieval request transmission that sends a retrieval request to the master computer for retrieval data required for retrieval to copy all or part of the data file at once in order to eliminate the difference between the data file of the own computer and the master computer Do the steps,
By the master computer,
Performing a retrieve data transmission step of transmitting the retrieve data to the replica computer in response to reception of the retrieve request from the replica computer;
By the replica computer,
When the retrieve data is received from the master computer, a retrieve data reception step is performed for executing the retrieve and updating the retrieve state information.
A distributed data equivalent method characterized by the above. The data file stores tabular data composed of rows and columns as the data body of the target data, and specifies data in any unit of the tabular data or the columns and rows constituting the tabular data. Stores a data status table consisting of data identification information and serial numbers assigned to each.
The retrieve state information is a retrieve state table composed of the table format data or data specifying information for specifying the data in any one of the columns and rows constituting the table and the respective retrieve states,
The retrieved data is data in units of either the tabular data obtained as a result of comparing the data state tables or the columns and rows constituting the data.
The distributed data equivalent method according to claim 1, wherein: The mode information managed by each computer is managed as mode management information including master computer information specifying which computer is the master computer and own computer mode information specifying the mode of the own computer,
The mode processing step by each computer is as follows:
At the time of initial startup of the own computer, when receiving mode information specifying that the other computer is a master computer, the master computer information is set and maintained in the other computer, and the own computer mode information is An initial startup step to set and maintain replica mode;
After the initial startup step, when master setting by the application is performed on the own computer, the master computer information is set and maintained in the own computer, and the own computer mode information is set in the master mode. And a master mode changing step for transmitting mode information identifying that the own computer is a master computer to other computers,
After the initial activation step, when the master setting by the application is performed on the other computer, as a result of receiving mode information specifying that the other computer is the master computer, the master computer information is set in the other computer. And including a replica mode change step for setting and maintaining the own computer mode information in the replica mode.
The distributed data equivalent method according to claim 1 or 2, characterized in that: Each of the initial activation step and the replica mode change step includes:
A step of setting and maintaining the master computer information in the other computer when receiving mode information specifying that the other computer is a master computer; and
Enabling the replication reception step to begin;
Performing the retrieve request transmission step and the retrieve data reception step on the master computer, and performing batch equalization of the master computer and the data in the data file;
Including setting and maintaining the own computer mode information in the replica mode;
The master mode changing step includes
When master setting by the application is performed on the own computer, based on reception of the mode information from the master computer information or another computer, checking the presence or absence of other computers operating as the master computer;
When it is confirmed that there is no other computer operating as the master computer, setting and maintaining the master computer information in the own computer; and
Enabling the replication sending step to begin;
Enabling the retrieval data transmission step to begin;
Setting and maintaining the computer mode information in the master mode;
Including transmitting to the other computer mode information specifying that the own computer is the master computer,
The distributed data equivalent method according to claim 3. In the mode processing step by each computer, in addition to the master mode and the replica mode, a state in the middle of becoming the master mode or the replica mode is handled as a transition mode,
The master mode changing step includes
Temporarily setting and maintaining the computer mode information in the transition mode; and changing and maintaining the transition mode from the master mode to the master mode;
The replica mode changing step includes:
Temporarily setting and maintaining the computer mode information in the transition mode, and changing and maintaining the transition mode from the master mode to the master mode,
5. The distributed data equivalent method according to claim 3 or 4, wherein: In addition to the master mode and the replica mode, the mode processing step by each computer treats a state in the middle of becoming the master mode and a state in the middle of becoming the replica mode as a master transition mode and a replica transition mode, respectively. ,
The master mode changing step includes
Including setting and maintaining the self-computer mode information once in the master transition mode, and changing and maintaining from the master transition mode to the master mode,
The replica mode changing step includes:
Temporarily setting and maintaining the own computer mode information in the replica transition mode, and changing and maintaining the replica mode from the master mode to the master mode,
6. The distributed data equivalent method according to claim 5, wherein: The master mode changing step includes
When master setting by the application is performed on the own computer, the step of setting and maintaining the own computer mode information in the master transition mode,
Transmitting master transition information indicating that the local computer is in the master transition mode to another computer;
Checking the presence or absence of other computers operating as the master computer based on reception of the mode information from the master computer information or other computers,
When it is confirmed that there is no other computer operating as the master computer, setting and maintaining the master computer information in the own computer; and
Enabling the replication sending step to begin;
Enabling the retrieval data transmission step to begin;
Setting and maintaining the computer mode information in the master mode;
Including a step of transmitting master operating information indicating that the local computer is operating as the master computer in the master mode to another computer,
The replica mode changing step includes:
When the master transition information is received from another computer, the own computer mode information is set and maintained in the replica transition mode; and
A step of setting and maintaining the master computer information in the other computer when the master operation information is received from another computer;
Enabling the replication reception step to begin;
Performing the retrieve request transmission step and the retrieve data reception step on the master computer, and performing batch equalization of the master computer and the data in the data file;
Including setting and maintaining the own computer mode information in the replica mode,
The distributed data equivalent method according to claim 6. The initial activation step includes:
When mode information specifying that another computer is a master computer cannot be obtained at the initial startup of the own computer, the own computer mode information is set to the transition mode or the replica transition mode and maintained. Including steps,
The distributed data equivalent method according to claim 5, wherein Each of the initial activation step and the replica mode change step includes:
Including performing the retrieval request transmission step and the retrieval data reception step on the master computer to collectively equalize the master computer and the data in the data file,
The retrieval request transmission step or the retrieval data transmission step includes:
Data for identifying the retrieve data according to the comparison result by comparing the data specifying information stored in the data file of the replica computer that is the request source and the master computer that is the request destination for the retrieve request Including specific information comparison step,
6. The distributed data equivalent method according to claim 3, wherein The retrieval request transmission step includes a step of transmitting the data specifying information of the own computer to the master computer when transmitting the retrieval request to the master computer.
When the retrieval request is transmitted from the replica computer that is the retrieval request source and the data identification information is received, the data identification information and the local computer are compared as the data identification information comparison step. Comparing the data identification information and identifying the retrieved data according to the comparison result,
The distributed data equivalent method according to claim 9. The retrieval request transmission step includes:
Before transmitting the retrieval request to the master computer, transmitting a transmission request for the data specifying information of the master computer;
When data specifying information corresponding to the data specifying information transmission request is received from the master computer, as the data specifying information comparing step, the data specifying information is compared with the data specifying information of the own computer, and the comparison result Identifying the retrieved data in response to:
The retrieve data transmission step includes a step of transmitting the data specifying information of the own computer to the replica computer when the transmission request of the data specifying information transmitted from the replica computer is received.
The distributed data equivalent method according to claim 9. The update data merging step includes:
One or more conditions selected from a plurality of condition items including the total number of the update content commit data obtained after the previous merge, the total data size, and the elapsed time since the previous merge as the merge merge condition Determining whether a merge merge condition configured from a combination of items is satisfied;
Determining the number of update content commit data to be merged when the merge merge condition is satisfied,
12. The distributed data equivalent method according to claim 1, wherein the distributed data equivalent method is any one of claims 1 to 11. By storing target data in multiple computers connected via a network, operating one specific computer in the master mode and each other computer in the replica mode to operate as a master computer and a replica computer, respectively. In a distributed data equivalent system in which the target data of each replica computer is equivalent to the same content as the master computer when the target data is updated to the original by the data update operation on the computer,
Each of the computers is
A database in which data files for storing the target data are arranged in a shared area accessible from a plurality of applications;
Among the plurality of computers, mode information including information specifying which computer is the master computer and information specifying the mode of the own computer is managed. A mode processing unit that executes an operation after the determined mode transition and transmits / receives mode information to / from another computer,
In the master mode, the result of the data update operation is temporarily stored in the update cache, and the content of the update cache is set as one update content commit data in response to a commit request, and unique data specifying information is added to the data. A data update processing unit to be stored in the file;
An update data merging unit that generates merge data by merging one or a plurality of the update content commit data determined according to a preset merge merge condition in the master mode;
In the master mode, based on the mode information, it manages transmission destination information indicating each replica computer to be equivalent, and transmits the merge data as replication data to each replica computer based on the transmission destination information. A replication processing unit that manages transmission source information indicating the master computer that is a transmission source of replication data in the replica mode, and receives replication data from the master computer based on the transmission source information; ,
A merge data expansion unit that expands the merge data that is the replication data into the original one or a plurality of the update content commit data during the replica mode;
In the replica mode, the expanded update content commit data is determined for each update content commit data based on the retrieve status information indicating the retrieve status of the data, and whether or not to reflect the data file is reflected. Then, the received data reflecting unit that reflects the determined update content commit data in the data file together with the data specifying information,
In the replica mode, the master computer issues a retrieval request for retrieving data required for retrieval to copy all or part of the data file at once in order to eliminate the difference between the data file of the own computer and the master computer. When the retrieval data is received from the master computer, the retrieval is executed to update the retrieval state information. In the master mode, the retrieval data is received in response to the reception of the retrieval request from the replica computer. It has a retrieve processing unit that sends it to the replica computer.
A distributed data equivalent system. By storing target data in multiple computers connected via a network, operating one specific computer in the master mode and each other computer in the replica mode to operate as a master computer and a replica computer, respectively. Distributed data equivalence applied to each computer to construct a distributed data equivalent system in which the target data of each replica computer is equivalent to the same content as the master computer when the target data is updated to the original by the data update operation on the computer In the system construction program,
When the computer includes a database in which a data file for storing the target data is arranged in a shared area accessible from a plurality of applications,
Among the plurality of computers, mode information including information specifying which computer is the master computer and information specifying the mode of the own computer is managed. A mode processing function that executes the operation after the determined mode transition and transmits / receives mode information to / from other computers,
In the master mode, the result of the data update operation is temporarily stored in the update cache, and the content of the update cache is set as one update content commit data in response to a commit request, and unique data specifying information is added to the data. Data update processing function stored in the file,
An update data merge function for generating merge data by merging one or a plurality of update content commit data determined in accordance with a preset merge merge condition in the master mode;
In the master mode, based on the mode information, it manages transmission destination information indicating each replica computer to be equivalent, and transmits the merge data as replication data to each replica computer based on the transmission destination information. A replication processing function for managing transmission source information indicating the master computer that is a transmission source of replication data in the replica mode, and receiving replication data from the master computer based on the transmission source information; ,
A merge data expansion function for expanding the merge data as the replication data into the original one or a plurality of the update content commit data in the replica mode;
In the replica mode, the expanded update content commit data is determined for each update content commit data based on the retrieve status information indicating the retrieve status of the data, and whether or not to reflect the data file is reflected. Then, the update data commit function that reflects the determined update content commit data in the data file together with the data specifying information,
In the replica mode, the master computer issues a retrieval request for retrieving data required for retrieval to copy all or part of the data file at once in order to eliminate the difference between the data file of the own computer and the master computer. When the retrieval data is received from the master computer, the retrieval is executed to update the retrieval state information. In the master mode, the retrieval data is received in response to the reception of the retrieval request from the replica computer. Retrieve processing function to send to the replica computer,
A program for building a distributed data equivalent system characterized by realizing

Images