WO2016157358A1

WO2016157358A1 - Data processing system

Info

Publication number: WO2016157358A1
Application number: PCT/JP2015/059905
Authority: WO
Inventors: 裕三石田; 浩雄吉田; 正樹曽我; 俊輔村上; 正州根岸
Original assignee: 株式会社野村総合研究所
Priority date: 2015-03-30
Filing date: 2015-03-30
Publication date: 2016-10-06
Also published as: RU2017134687A; RU2686028C2; SG11201707971YA; PH12017501685A1; JP6475820B2; RU2017134687A3; JPWO2016157358A1

Abstract

[Problem] To ensure rapidity of processing when maintaining the same data on a plurality of DB servers. [Solution] In this data processing system 10, which is provided with a plurality of AP servers and a plurality of DB servers, each DB server is provided with a shared table that stores shared work data, wherein deletion and updating of records is prohibited. Each AP server is connected to all the DB servers, executes shared work processing, transmits work data which is a processing result to each DB server, and requests addition to a corresponding table. An ID including an identification code for the executing AP server is assigned to each piece of work data, and a responsible AP definition table, which defines an AP server that is responsible for specified work processing for which exclusive control is sought, is stored in each DB server. Each AP server, in the event of processing requiring exclusive control, consults the responsible AP definition table, identifies the responsible AP server, and transfers the processing thereto. The responsible AP server stores this work processing in a queue and then executes said processing in order in accordance with FIFO.

Description

Data processing system

The present invention relates to a data processing system, and more particularly to a technique for managing business data generated by a plurality of AP servers by a plurality of DB servers.

In the case of a large-scale disaster such as an earthquake, tsunami, or eruption, or when a database is physically damaged due to a nuclear disaster or terrorist or other man-made disaster, a corporate database is a backup database in which business data is installed remotely. Duplicate registration has been done for a long time.
For example, Patent Document 1 discloses a technique for registering data in a disk device of a primary control device, and then transmitting the same data to a sub-control device installed at a remote place and requesting the registration.
JP-A-11-85408 Introduction to Web System / 1st Web Site Configuration and J2EE Server Internet URL: http://www.atmarkit.co.jp/fjava/rensai2/websys01/websys01.html Search Date: March 10, 2015 Two-phase commit Internet URL: http://e-words.jp/w/2%E3%83%95%E3%82%A7%E3%83%BC%E3%82%BA%E3%82%B3% E3% 83% 9F% E3% 83% 83% E3% 83% 88.html Search date: March 10, 2015

In this way, by holding data twice by the primary and secondary control devices, even if an earthquake or the like occurs in one area and the disk device is physically damaged, the other disk device that survived By utilizing the data inside, it becomes possible to ensure the continuity of business processing.
In addition, since a failure or trouble of the server computer itself is assumed, it can be said that it is desirable to prepare for an emergency by duplicating the database server not only in a remote place but also in the same site.

By the way, when executing update processing for a plurality of databases, since a contradiction occurs between data if commits for all the databases are not established, an arbitration method called XA interface two-phase commit is usually adopted. Mutual consistency is ensured.

In this two-phase commit, first, the master server that controls the entire process inquires of each target DB server whether the commit can be executed. The DB server that is ready for update returns a “ready” response, confirms that all DB servers have completed the preparation, the master server notifies the commit start, and the database is rewritten all at once. If an error occurs in one of the databases during rewriting, the DB server where the error occurred reports the failure, and the master server notifies all servers of the withdrawal and performs a “rollback” process to restore the data. (Non-Patent Document 2).

By adopting such a two-phase commit, data consistency between DB servers is ensured, and it is possible to surely avoid the generation of illegal data.
However, since it is necessary to repeatedly exchange data between the master server and the remote DB server, the speed of processing must be sacrificed. For this reason, even if it is effective for business processing that does not cause a problem even if waiting time of several seconds occurs as in ATM trading, for example, as in online trading, a delay of several seconds is not allowed There was a problem that it could not be applied to business processing.

The present invention has been devised in view of such a current situation, and realizes a technology capable of ensuring the speed of processing while preserving the same data in a plurality of DB servers. It is an object.

In order to achieve the above object, a data processing system according to claim 1 is a data processing system including a plurality of AP servers and a plurality of DB servers, and the DB servers are common to each other. Each table is provided with a table for storing business data, and each table has a restriction that only records can be referenced and added, and deletion and updating are prohibited. Connected to the DB server, the function to execute common business processing for requests sent from client terminals, and the business data generated as a result of processing are sent to each DB server above, It has a function to request additional registration, and each business data mentioned above is assigned a unique ID including an identification code to identify the AP server that generated it. In addition, each DB server above stores a responsible AP definition table that defines for each business process the AP server responsible for the specific business process for which exclusive control is required. If the business process request sent from the client terminal does not require exclusive control, it executes it by itself, whereas if it requires exclusive control, refer to the AP definition table in charge, and After identifying the AP server that should be in charge of business processing, delegate the relevant processing to the relevant AP server, and the responsible AP server that received this temporarily stores the delegated business processing in the queue, Each business process is executed in order according to the rules of the FIFO.

The data processing system according to claim 2 is the system according to claim 1, and in addition to the main AP application server, the main AP application server performs a process when the main AP application server does not function. A feature is that the sub AP server to be assigned is defined for each business process.

The data processing system according to claim 3 is the system according to claim 1 or 2, and each of the business data is stamped with a time stamp at the time of business processing execution. When it is necessary to update the current business data, new business data with an ascending ID different from the business data, the corrected data, and the time stamp at the time of correction is generated, and the business data for updating is also generated. It is characterized in that existing business data is substantially updated by additionally registering in a corresponding table provided in each DB server.

The data processing system according to claim 4 is the system according to claims 1 to 3, and further, when the AP server needs to delete the existing business data, the ID of the business data is to be canceled. Generate business data for data cancellation with data items to be stored, and add the business data for data cancellation to the cancellation-dedicated table provided in each DB server to effectively store the existing business data. It is characterized by being deleted.

According to the data processing system described in claim 1, business data generated by the AP server is simultaneously transmitted to a plurality of DB servers and is registered in duplicate in each table, so that the integrity of business data can be improved. it can.
In addition, when registering business data, it is prohibited to delete or update data that requires complicated processing in the event of a failure. Data can be recovered simply by copying the missing data from another DB server. Therefore, it is not necessary to adopt a strict arbitration method such as a two-phase commit, and it is possible to determine that registration is complete when a receipt notification is simply returned from each DB server.

Also, a plurality of AP servers are prepared, and each AP server can freely process a business process that does not require exclusive control. Therefore, the processing efficiency of the entire system can be improved by load balancing.
On the other hand, for some business processes that require exclusive control, the processes are aggregated in the AP server preset in the assigned AP definition table and processed according to the FIFO rules. It is possible to effectively avoid the generation of contradictory data.
Because the business data generated in each AP server is assigned a unique ID including the identification code of each AP server, even if business data is transmitted from each AP server to the DB server, Each ID does not overlap, ensuring absolute uniqueness.

In the case of the data processing system described in claim 2, since the sub AP server is defined in addition to the main AP server in the AP definition table, the main AP server is down for some reason. However, it is possible to quickly delegate the business process to the sub AP server.

In the case of this invention, the DB server table is provided with a restriction that prohibits deletion and update of records. However, by using the mechanism according to claims 3 and 4, substantial update of business data and Deletion can be realized.

FIG. 1 is an overall configuration diagram of a data processing system 10 according to the present invention.
First, in the first data center 12 located in Tokyo, a first AP server 14, a first DB server 16, and a second DB server 18 are installed.
A second AP server 22, a third DB server 24, and a fourth DB server 26 are installed in the second data center 20 located in Osaka.

Each of the first AP server 14 and the second AP server 22 has the same application program, and executes the same business process for the request transmitted from the client terminal 28 via the communication network 27 such as the Internet. It has the function to do.

For convenience of illustration, only the first AP server 14 and the second AP server 22 are shown, but in reality, a plurality of AP servers having the same function are installed in each data center, and a load (not shown) is shown. Load distribution is achieved through a balancer.
In this way, if multiple AP servers are provided in the same data center, even if any AP server is planned or unplanned, it is the same without switching to an AP server in another data center. The remaining AP server in the data center can be replaced, and downtime can be minimized.
The number of DB servers is not limited, and three or more DB servers can be installed in each data center.

The first DB server 16, the second DB server 18, the third DB server 24, and the fourth DB server 26 each have a common table, and the same data is stored in each table. .
Here, as shown in FIG. 2, each DB server has a point addition table 50, an addition cancellation table 51, a point application table 52, an application cancellation table 53, and a responsible AP definition table 54 in common. .

In the point addition table 50, data items such as an addition ID, an account, the number of points, and a time stamp are set.
Further, in the addition cancellation table 51, data items such as cancellation targets and time stamps are set.
In the point application table 52, data items such as an application ID, an account, the number of points, and a time stamp are set.
In the application cancellation table 53, data items such as a cancellation target and a time stamp are set.
In the assigned AP definition table 54, data items such as an assigned AP definition ID, an account, a primary assigned AP, and a secondary assigned AP are set.
In the addition ID and application ID, a unique identification code using an artificial key is stored (details will be described later).

The first AP server 14 is connected to the first DB server 16 and the second DB server 18 via the LAN, and is installed in the second data center 20 via the communication line 29. The DB server 24 and the fourth DB server 26 are also connected.
The second AP server 22 is connected to the third DB server 24 and the fourth DB server 26 via the LAN, and is installed in the first data center 12 via the communication line 29. The first DB server 16 and the second DB server 18 are also connected.

In normal times, the client terminal 28 installed in eastern Japan is connected to the first AP server 14 installed in the first data center 12, and the client terminal 28 installed in western Japan is connected to the second data center 20. It is connected to a second AP server 22 installed inside.

On the other hand, for example, when a large earthquake occurred in the Tokyo metropolitan area and the first data center 12 was devastated, it was installed in eastern Japan by changing the settings of the DNS server (not shown). Since the connection destination of the client terminal 28 is switched to the second AP server 22 installed in the second data center 20, continuity of service for users in East Japan is ensured.
Similarly, when the second data center 20 is damaged, the connection destination of the client terminal 28 installed in West Japan is switched to the first AP server 14 set in the first data center 12. .

As described above, each data center normally provides services to neighboring client terminals 28, but also provides services to client terminals 28 nationwide in the event of a disaster. DB servers usually need to keep data of the same content mutually.
In addition, since multiple DB servers with the same data are installed in the same data center, the AP server can request any DB server to extract data when referring to the data. Dispersion can be achieved.

The number of data centers is not limited to two. As many data centers as possible are provided in various locations, and multiple AP servers with the same functions are installed in each data center. It is desirable to connect to all DB servers via a communication network.

FIG. 3 shows the internal configuration of the first AP server 14 and the second AP server 22. The business processing unit 30, the artificial key management unit 32, the data control unit 34, and a plurality of DB communications are shown. Part 38.

The business processing unit 30 executes data generation processing, arithmetic processing, registration processing in the DB server, etc. in response to a request from the client terminal 28, and the AP server CPU operates according to a dedicated application program. Is realized.

The artificial key management unit 32 has a function of issuing an artificial key (surrogate key) in response to a request from the business processing unit 30 (details will be described later).
The artificial key management unit 32 is realized by the CPU of the AP server operating according to dedicated middleware.

When the data control unit 34 receives a data registration instruction from the business processing unit 30, the data control unit 34 has a function of replicating data as many as the number of DB servers and passing them to the DB communication unit 38. This is realized by the CPU of the AP server operating according to the above.
In addition, the data control unit 34 has a function of receiving an instruction from the artificial key management unit 32 and instructing the DB communication unit 38 in charge of each DB server installed in the same data center to update the artificial key management data. (Details will be described later).

Each DB liaison unit 38 performs functions such as issuing an SQL statement to a DB server assigned to itself in advance and instructing additional registration and reference of data, and the CPU of the AP server according to dedicated middleware. It is realized by operating.

Here, based on the flowchart of FIG. 4, the mechanism of artificial key assignment in the system 10 will be described.
First, the business processing unit 30 that has generated data as a result of the business processing requests the artificial key management unit 32 to designate a table and issue an artificial key (S10). In the case of generating data to be stored in the above point addition table 50, it is requested to issue an artificial key for “addition ID”.

Receiving this, the artificial key management unit 32 refers to the artificial key management table 56 stored in the DB server installed in the same data center (S12), and works on the latest artificial key related to the point addition table 50. The processing unit 30 is notified (S14).

As shown in FIG. 5, the artificial key management table 56 includes data items of “table ID”, “AP server ID”, and “next key value”, in units of “table ID” + “AP server”. The latest artificial key (next key value) is managed.

Here, the “artificial key” is a numerical value having a predetermined length (for example, 32 bits or 64 bits).
In the last digit (end) of the numerical value, an identification code (any numerical value from 0 to 9) for specifying the AP server that generated the data is set.

Specifically, the relationship between the ID of each AP server and a numerical value of 0 to 9 is defined in the lower-order digit management table 57 of the artificial key.
In the AP server management table 58 and the data center management table 59, the relationship between each AP server and the data center is defined.
Similar to the artificial key management table 56, the lower one digit management table 57, the AP server management table 58, and the data center management table 59 of these artificial keys are also stored in the DB server in the same data center.

The artificial key management unit 32 that has finished notifying the latest value of the artificial key to the business processing unit 30 updates the value of each record in the artificial key management table 56 (S16).
Specifically, for the “next key value” in the point addition table 50 related to the AP server, “the last value (the latest value above) is composed of the part excluding the identification code at the end. Is set to a value obtained by adding 1 to a numerical value.

However, the update method of the latest value of the artificial key is not limited to such “ascending order”, but “in the immediately preceding value (the latest value described above) excluding the identification code at the end. It may be “descending order” in which “a value obtained by subtracting 1 from a configured numerical value” is set. The identification code of the AP server is not limited to the numerical value of “0 to 9” as described above, and other character types may be used. Further, an identification code may be inserted at the head of the artificial key.
Further, only the numerical value excluding the AP server identification code is stored in the “next key value” of the artificial key management table 56, and the artificial key management unit 32 sets the corresponding identification code at the time of issuing the artificial key. It can also be added to the end of the numerical value and issued to the business processing unit 30.

Upon receiving the latest artificial key from the artificial key management unit 32, the business processing unit 30 generates data in which the artificial key is set as an ID in a primary key or an external key (S18).
This data is transmitted to all DB servers via the data control unit 34 and the DB communication unit 38, and is additionally registered in a corresponding table (point addition table 50) provided for each.

As described above, the uniqueness of the artificial key is ensured in units of “AP server x table”, and each AP server is associated with a specific data center in the data center management table 59. Even if an artificial key is issued, the uniqueness of the last digit is ensured, so there is no risk of duplicated artificial keys being issued, and the uniqueness of each data is reliably ensured.

In addition, since the artificial key issuance requests issued from the application server's business processing unit 30 are once queued and processed according to the FIFO (first-in first-out) rules, Therefore, there is no risk that an artificial key with a small value is issued by mistake, and the order is guaranteed.

Data generated by the business processing unit 30 of each AP server is stored in a table in each DB server as described above, but deletion (delete) prohibition and update (update) prohibition in each table These restrictions are imposed in advance.
In short, in this system 10, only addition (insert) and reference (select) to each table is allowed.

For this reason, when it becomes necessary to invalidate existing data, an equivalent state is realized by adding a new record to another table instead of deleting the record.
For example, when the point addition data shown in FIG. 2 is invalidated, the addition cancellation data is added to the addition cancellation table 51.
Since the “cancellation target” of the addition cancellation data is filled with the addition ID of the point addition data to be invalidated, the business processing unit 30 adds the point addition data registered in the point addition table 50 to the addition. Those whose IDs are registered in the addition cancellation table 51 are not counted.

Similarly, application revocation data is added to the application revocation table 53 when the point application data is invalidated.
Since the application ID of the point application data to be invalidated is filled in the “cancellation target” of this application cancellation data, the business processing unit 30 applies the application among the point application data registered in the point application table 52. Those whose IDs are registered in the application cancellation table 53 are not counted.

In addition, when the registration data needs to be corrected in this system 10, it is possible to respond by registering a new record with the corrected data instead of updating the existing record.
For example, when correcting the number of points in the point addition data, a new addition ID in ascending order different from the addition ID of the point addition data is paid out, and the point addition data in which the number of points is changed in the same account Add to 50.
The data before correction and the data after correction are stamped with millisecond precision time stamps, and each is given an addition ID that is different in ascending order. Can definitely be identified.

As described above, the same data is stored in each DB server, and each AP server has a function to execute the same business process. Service can be provided, and load distribution of the entire system 10 can be performed.

However, depending on the contents of the business process, there are things that require exclusive control between AP servers.
First, there is no need for exclusive control for the process of adding points to users who have purchased goods or services. For example, if multiple employees within a company simultaneously generate points for a company account, There is no particular problem even if processing is performed in parallel on the AP server.
Similarly, it can be said that exclusive control is unnecessary for the point balance inquiry processing for the same account.
On the other hand, when a product is purchased using points accumulated in the account, the number of accumulated points is limited, so if there is a request for applying points concurrently, exclusive control is executed. There is a need.

Hereinafter, according to the schematic diagram of FIG. 6 and the flowcharts of FIGS. 7 to 10, a general processing procedure when the system 10 is applied to point management work will be described, and then the exclusive control method will be described.
First, when a user purchases a product at a store and presents a point card, the point card account (card number) from the client terminal 28A in the store and the number of points generated by the product purchase (the number of points added) It is transmitted to the second AP server 22.

Upon receiving this point addition request, the business processing unit 30 of the second AP server 22 (S10 in FIG. 7) receives the addition ID from the artificial key management unit 32, and then generates point addition data (S12). .
Next, the business processing unit 30 requests the first DB server 14 to the fourth DB server 26 to additionally register the point addition data via each DB communication unit 38 (S14).
Upon receiving this, the first DB server 14 to the fourth DB server 26 add the point addition data to the respective point addition tables 50 all at once.

Next, the business processing unit 30 calculates the point balance of the user (S16).
Specifically, all point addition data associated with the account is acquired from any DB server (S16-01 in FIG. 8), and an addition ID is extracted from each point addition data (S16-02).
Next, this addition ID is transmitted to an arbitrary DB server, and corresponding addition cancellation data is acquired (S16-03).
Then, except for the point addition data in which the addition cancellation data is registered, the number of points of the remaining valid point addition data is totaled (S16-04), thereby determining the number of addition points of the user.

Next, all point application data associated with the account is acquired from any DB server (S16-05), and an application ID is extracted from each point application data (S16-06).
Next, this application ID is transmitted to an arbitrary DB server, and corresponding application cancellation data is acquired (S16-07).
Then, by excluding the point application data in which the application cancellation data is registered, the number of points of the remaining valid point application data is totaled (S16-08), thereby determining the application point number of the user.

Next, by subtracting all valid application points from all valid addition points (S16-09), the current point balance is determined.
This point balance is transmitted to the client terminal (S18 in FIG. 7) and printed on the receipt.

On the other hand, when a user wishes to apply points at the store, the point card account (card number) from the client terminal 28B and the number of points applied according to the price of the purchased product are sent to the second AP server 22. Sent.

Upon receiving this point application request, the business processing unit 30 of the second AP server 22 (S30 in FIG. 9) is in charge of the point application process for the account by referring to the assigned AP definition table in any DB server. The AP server to be executed is specified (S32). And if it happens that self is designated as the main AP server, the point application process described later is executed as it is, but another AP server (for example, the first AP server 14) is designated as the responsible AP server. If so, the request (account, number of applied points, etc.) is transferred to the corresponding AP server, and the point application processing is delegated (S34).

The first AP server 14 that has received this processing delegation temporarily stores this request in the FIFO queue (S36), and then executes the point application processing in order (S38).
This point application process specifically follows the following procedure.
First, the first AP server 14 calculates the point balance of the account by the same procedure as described above (S38-01 in FIG. 10).

Next, point application data is generated within the range of this point balance (S38-02).
For example, if the current point balance exceeds the number of application points this time, the business processing unit 30 receives the application ID issued from the artificial key management unit 32, and then the point application data (application ID, account, number of points) , Timestamp).
Then, all the DB servers are requested to register additional point application data (S38-03).
On the other hand, if the point balance is less than the number of applied points at this point, the business processing unit 30 generates point application data for the remaining number of points and requests all DB servers to perform additional registration. As a result, the point balance of the account becomes zero.

The first AP server 14 that has completed the delegated point application processing notifies the second AP server 22 of the processing result (number of applied points and point balance) (S40 in FIG. 9).

Receiving this, the business processing unit 30 of the second AP server transmits the processing result to the client terminal 28B (S42).
That is, when points can be applied for the total purchase amount, the client terminal 28B is notified of this and the point balance.
On the other hand, when the points are applied only to a part of the purchase amount, the client terminal 28B is notified of the fact and the shortage amount. At this time, cash payment is made for the shortage at the store.

As described above, when the point application processing that requires exclusive control is reached, the processing authority is transferred from the second AP server 22 to the first AP server 14, and the first AP server 14 determines the FIFO. The point application process is executed according to the rule. For this reason, even if a point application request related to the same account is transmitted from the client terminal 28C to the first AP server 14 at substantially the same time, adjustments between competing requests can be made according to FIFO (first-in first-out) rules. In addition, it is possible to effectively avoid the occurrence of illegal processing (eg, the application of points is executed even though the point balance is actually zero).

In contrast, in the case of point addition processing or point balance calculation processing for the same account, since it is used for parallel processing in a plurality of AP servers, efficiency can be improved by distributed processing.

In addition, the assigned AP definition table 54 stores the secondary assigned AP in addition to the primary assigned AP, so if the AP server designated as the primary assigned AP is down, the secondary assigned AP is immediately available. Authority can be delegated to the AP server.
If it is uneasy only to prepare the sub AP server, a lower AP server such as the sub AP can be set in advance.
The assigned AP definition table 54 is stored in all DB servers, so even if any DB server is temporarily down, the AP server identifies the main assigned AP server from other DB servers. It becomes possible.

In order to effectively execute this exclusive control, in the application program for the business processing section, processing that requires exclusive control (point application processing) and unnecessary processing (point addition processing, point balance confirmation processing, etc.) are defined in advance. It is necessary to keep it.
In addition, in the assigned AP definition table 54, specific information (URL or the like) of the primary responsible AP and the secondary responsible AP is defined for each account (or for each account group).

However, according to the present invention, as described above, the type of business processing that requires exclusive control (point application processing) is defined in the application program, and the correspondence between the processing target (account) and the AP server is defined in the assigned AP definition table 54. It is not limited to the method defined on the side.

For example, you must define exclusive control for multiple types of business processes in the application program as shown below, and set the primary AP server and secondary AP server for each type of business process in the assigned AP definition table. You can also.
[Application program settings]
Exclusive control required-> Process A, Process B, Process C [Settings in charge AP definition table]
Process A-> Main AP: Server a1, Sub AP: Server a2
B processing-> Main AP: Server b1, Sub AP: Server b2
C processing-> Main AP: Server c1, Sub AP: Server c2

In the above, the point management work is illustrated, but the same exclusive control can be applied to other work processes.
For example, when applied to an inventory management system, as shown in FIG. 11, an arrival table 60, an arrival cancellation table 61, a shipping table 62, a shipping cancellation table 63, and a responsible AP definition table 64 are stored in each DB server. Keep it inside.

In the arrival table 60, data items such as an arrival ID, a product code, the number, and a time stamp are set.
Further, in the arrival cancellation table 61, data items such as a cancellation target and a time stamp are set.
In the shipping table 62, data items such as a shipping ID, a product code, the number of items, and a time stamp are set.
In the shipping cancellation table 63, data items such as a cancellation target and a time stamp are set.
In the assigned AP definition table 64, data items such as assigned AP definition ID, product code, main assigned AP, and sub-assigned AP are set.
In the above-mentioned arrival ID and shipping ID, a unique identification code using an artificial key is stored.

Here, when a product arrival request is received from the client terminal 28, it is only necessary to simply add arrival data to the arrival table 60. Since exclusive control is not required, any AP server is authorized to process.
In addition, when an inquiry request for the number of stocks is received from the client terminal 28, it can be calculated by “the number of uncancelled arrivals−the number of uncancelled shipments”, so exclusive control between AP servers is unnecessary.

On the other hand, when a product order (shipment) request is transmitted from the client terminal 28, orders exceeding the stock quantity cannot be received, so exclusive control between AP servers is required.
Specifically, when a product order request is generated, the processing authority is delegated to the AP server in charge of shipping processing of the product (product code).

Receiving this, the main AP server temporarily stores the order request in the FIFO queue, and executes the processing in order from the earliest.
That is, the business processing unit 30 of the main AP server calculates the stock of the product related to the product code included in the order request, and if the number included in the order request is within the stock range, Generate and request all DB servers to add to the shipping table 62. Thereafter, the processing result is notified from the consignee AP server to the consignee AP server.
When the inventory of the product is less than the number of orders, the consignee AP server notifies the consignment AP server of the result of the shortage of inventory without generating shipping data.

Even in this case, in order to effectively implement exclusive control, the application program for the business processing unit 30 defines in advance processing that requires exclusive control (order reception processing) and unnecessary processing (arrival reception processing, inventory confirmation processing). It is necessary to keep it.
In addition, there are some products such as software and electronic books sold in download format that do not need to worry about the number of stocks depending on the sales target.
In such a case, it is required to define the necessity of exclusive control for each target product (for each product code) even in the same order receiving process.

Each DB server in this system 10 receives the data to be added from the DB contact unit 38 and stores it in its own memory, and then sends a receipt completion notice to the DB contact unit 38. Thereafter, the hard disk, SSD, etc. The corresponding record is stored in a table provided in the external storage device.

The reception completion notification transmitted from the DB server is collected in the data control unit 34 via each DB communication unit 38.
The data control unit 34 outputs a registration completion notification to the business processing unit 30 when reception completion notifications are returned from all the DB communication units 38.

The business processing unit 30 recognizes the data as a target to be read when the registration notification completion notification is received from the data control unit 34.
In other words, until the registration completion notification is returned, the business processing unit 30 excludes the data from being read, and the registration timing in each DB server is shifted and incorrect data is read. It avoids the danger of end up.

However, if a receipt completion notification is not returned after a certain period of time from any DB server, the data control unit 34 recognizes that a communication failure or machine trouble has occurred and places the DB server in offline mode. Transition.
Specifically, the DB server is temporarily disconnected from the system 10, and the operation mode is based on only the remaining DB servers.
That is, when a reception completion notification has been received from all DB servers other than the separated DB server, a registration completion notification is output from the data control unit 34 to the business processing unit 30, and the registered data is to be read.

During this time, the disconnected DB server is inspected and necessary recovery measures are taken. For example, if it is determined that the receipt completion notification has not arrived due to a failure of the communication device, after replacing the communication device with a new communication device, the DB server is set to the write mode (only data writing is permitted and data reference is not allowed). (Prohibited state) and copy the difference data from another DB server installed in the same data center.

When the stored data catches up to the latest state, the DB server is set to the read / write mode (a state in which data writing and reference is permitted) and reconnected to the system 10.
Thereafter, the data control unit 34 resumes data transmission / reception via the DB communication unit 38 in charge of the DB server.

In the case of this system 10, as described above, each business data is allowed to be added only, and a rule that does not allow deletion or update is applied. Therefore, data recovery is extremely easy.
In other words, if it is assumed that deletion or update of data is allowed, in order to recover data, addition, deletion, or update of data is started from a certain point based on the update history information held by the DB server. It must be reproduced in order, and this takes a long time and a large load. In addition, it is necessary to provide a mechanism on the DB server side to securely save the update log.

On the other hand, if it is assumed that only addition of data is allowed and deletion and update are prohibited as in this system 10, data registered in other normal DB servers and registered in the failed DB server It is only necessary to compare the data and copy the missing data, and there is no need to consider the order of the data (prior relationship).

Further, in this system 10, since the reception completion notification is issued from the DB server to the AP server when the data transmitted from the DB communication unit 38 is stored in the memory of the DB server as described above, In a very short time, the task processing unit 30 can read the data.

Conventionally, memory is a volatile storage means, and data is lost when power supply is stopped, so when data is stored in a nonvolatile storage means (hard disk, etc.) A notification should be returned.
On the other hand, in the case of this system 10, even if data loss occurs in a certain DB server, it is possible to catch up simply by copying the missing data from the adjacent DB server. There is no need to wait until it is stored.

In addition, since data recovery is easy in this way, when storing the same data in multiple DB servers, it is not required to go through the troublesome procedure of two-phase commit as in the past, and for each DB server Thus, it is possible to simply request registration and enable reading when each of the reception completion notifications is returned (so-called autocommitment).
As a result, the exchange of messages between the AP server and the DB server is drastically reduced, and the overall processing speed can be improved.

The above-mentioned various business data (point addition data, addition cancellation data, point application data, application cancellation data, etc.) are stamped with a millisecond precision time stamp by the business processing unit 30 as described above. Even when the value of MAX reaches MAX and IDs that are already issued are reissued (recycled), the business processing unit 30 can determine old and new data by comparing time stamps.

In the above, an example in which a plurality of AP servers and DB servers are arranged in a plurality of data centers has been described, but the present invention is not limited to this.
In other words, in an environment where multiple AP servers and multiple DB servers are installed in a single data center, and each AP server and all DB servers are connected via a network, each AP server handles business processes that do not require exclusive control. In addition to independent parallel processing, business processing that requires exclusive control can also be effectively applied to consolidating processing authority to a preset AP server.

1 is an overall configuration diagram of a data processing system according to the present invention. It is a figure which shows an example of the table stored in each DB server. It is a block diagram which shows the internal structure of a 1st AP server and a 2nd AP server. It is a flowchart which shows the issuing procedure of an artificial key. It is a figure which shows structures, such as an artificial key management table. It is a schematic diagram which shows the process sequence at the time of applying this system to point management business. It is a flowchart which shows the procedure of a point addition process. It is a flowchart which shows the procedure of a balance calculation process. It is a flowchart which shows the process sequence which concerns on the exclusive control between AP servers. It is a flowchart which shows the procedure of a point application process. The It is a figure which shows the other example of the table stored in each DB server.

10 Data processing system
12 First data center
14 First AP server
16 First DB server
18 Second DB server
20 Second data center
22 Second AP server
24 Third DB server
26 Fourth DB server
27 Communication network
28 Client terminal
29 Communication line
30 Business processing department
32 Artificial Key Management Department
34 Data control unit
38 DB Liaison Department
50 point addition table
51 Addition cancellation table
52 point application table
53 Undo table
54 Assigned AP definition table
56 Artificial key management table
57 Lower digit management table for artificial key
58 AP server management table
59 Data center management table
60 Arrival table
61 Receipt cancellation table
62 Shipping table
63 Shipping cancellation table
64 Assigned AP definition table

Claims

A data processing system comprising a plurality of AP servers and a plurality of DB servers,
Each of the above DB servers has a table storing business data common to each other, and each table is provided with a restriction that only allows reference and addition of records, and prohibits deletion and update.
Each of the above AP servers is connected to each of the above DB servers, and functions for executing common business processing for requests sent from client terminals, and business data generated as a result of the processing. It has the function to send to the DB server and request additional registration to the corresponding table,
Each business data above is assigned a unique ID including an identification code to identify the AP server that generated it.
In addition, each DB server above stores a responsible AP definition table that defines for each business process the AP server that is responsible for specific business processes that require exclusive control.
Each AP server above executes itself when a business process request sent from the client terminal does not require exclusive control. After referring to the table and identifying the AP server that should be in charge of the business process, delegate the corresponding process to the AP server in charge,
Upon receiving this, the AP server in charge stores the delegated business process once in the queue, and then executes each business process in the queue in order according to the FIFO rules.
In the above-mentioned responsible AP definition table, in addition to the primary responsible AP server, a secondary responsible AP server that should be in charge of processing when the primary responsible AP server does not function is defined for each business process. Item 4. The data processing system according to Item 1.
Each of the above business data is stamped with a time stamp at the time of business processing execution.
When it is necessary to update the existing business data, the AP server newly generates business data having an ascending ID different from the business data, data after correction, and time stamp at the time of correction,
3. The data management system according to claim 1, wherein the existing business data is substantially updated by additionally registering the business data for update in a corresponding table provided in each DB server. .
The AP server generates business data for data cancellation with a data item that stores the ID of the business data as a cancellation target when it is necessary to delete existing business data,
4. The existing business data is substantially deleted by additionally registering the business data for data cancellation in a table dedicated to cancellation provided in each DB server. The data processing system described.