JPS6182266A - Global transaction execution control system - Google Patents

Abstract

PURPOSE:To improve the efficiency by treating a task identifier of a transmitted application program as a task identifier of a sub-transaction execution task at other site to eliminate the need of return processing (BIND processing) of the task identifier. CONSTITUTION:When a global transaction is executed at the side A, a transaction identifier TRID is added to a sub-transaction start message BEGIN1 outputted from the application program APP1 and the result is transmitted to a site B. A sub transaction execution tank TEM1 assigned by an assignment request processing section 13 at the site B forms a task identifier by adding a site identifier M2 of the site B to an outputted transaction identifier TRID. Then the communication is attained by using the task identifier.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a global transaction execution control method, and particularly in a distributed data system, when a subtransaction executed at another site is executed by any task among a plurality of tasks. When a task is assigned to a task and executed, the task can be processed without being notified of the task's identifier.

For example, in the business of banks, host computers tMM1 and M2 are installed in Tokyo and Osaka, respectively, as shown in Figure 2.
Data for customers in Kanto is stored in local database DB1, data for customers in Kansai is stored in local database DB2, and users tJ++ and T
For J, 2, U2T, and U22-, the database is 1.
(The so-called distributed database method is used, which is recognized as belonging to the prisoner and does not require any awareness, such as a database in Tokyo and a database in Osaka.

In this case, for example, if the data that user U++ wants to access spans both local databases DBI and DB2, the application program APPII executes access to local database DB1, but the data that user U++ wants to access spans both local databases DBI and DB2. It is necessary to execute this by the subtransaction execution task TEM2L and transmit it to the host computer M1.

[Problem that the invention seeks to solve]

By the way, in conventional data processing, it is necessary to execute subtransactions at sites that are far apart from each other.
When executing a global transaction, (1)
Either the subtransaction execution task of the other party's site was fixedly determined in advance, or (2) the other party's site had been notified of the task identifier of the subtransaction execution task arbitrarily assigned.

By the way, in the case of (1) above, as shown in Figure 3, the global transaction is, for example, an application program AP.
When the PII is issued and executed from APP12-, the subtransaction execution task TB of another site
M21 and TEM22- are each application program APPII,
There was a fixed one-to-one correspondence with APP12-. Of course, based on this, the task identifier is also fixed, and the destination of the subtransaction start message BEGIN is fixed by the APPII and APP12-. This type of fixed system is suitable for routine specific tasks, such as TSS (Time Sharing System).
When a user's job is generated dynamically as in the case of the use in m), there is a drawback that the correspondence cannot be determined in advance, so the method cannot be used.

In the case of (2) above, as shown in FIG.
When issued and executed from PPII, the TEM management task CM of the other site arbitrarily selects a TE from among the subtransaction execution tasks TEM21-.
M2i is assigned.

The task identifier of TEM2i of the subtransaction execution task assigned in this way is sent to the application program APP by task identifier notification processing (BIND processing).
This will be notified to the II side, and from then on it will be possible to communicate directly with TEM2 i. In this case, it is assumed that the task identifier of the TEM management task TCM is known on the application program APPII side. As described above, (2) requires a task identifier notification process, which is inefficient, and requires an execution task management task TCM, making control complicated. If the communication path is dynamically changed, there is no waiting time for BIND processing, but control becomes complicated. Moreover, as shown in FIG. 4(b), this sub-transaction start message is sent to the communication module VC, which is in charge of message communication with other sites.
The task is transmitted to the TEM management task CM via M, an arbitrary TEM is allocated in this TCM, and the task identifier of that TEM is ACK-ACKed from the allocated TEM.
It is included in BEGIH and notified to the other party's application program APP via VCM. However, if all the subtransaction execution tasks TEM are in use, a queue is created for subtransactions sb1 and sb2- that have not been executed, and the subtransaction execution tasks TEM
The problem is that the processing in the execution task management task TCM is overlapped with the transaction error recovery processing and other processing other than this TEM allocation management, and the operation of this TCM becomes a bottleneck. was there.

[Means for solving problems]

In order to improve the above-mentioned problems, the present invention provides a global transaction execution control method that uses multiple host computers, a local database connected to the host computers, and a global transaction execution control system that spans multiple sites. In a distributed database system in which a transaction is broken down into subtransactions for each site and executed, each site has an identifier generating means for identifying an application program task that has occurred at that site, and a communication module has an identifier for identifying an application program identifier. A recognition means is provided, and the identifier generated by the identifier generation means is added to the subtransaction execution task executed at another site and sent, and this identifier is used in the communication module of the other site as a task of the subtransaction execution task. By handling the subtransaction execution task as an identifier, when dynamically allocating and executing a subtransaction execution task of another site, processing can be performed without waiting for notification of the identifier of the task.

[Effect]

With the above configuration, the global transaction generating site uniquely determines the task identifier of the subtransaction execution task of another site that executes the application program and subtransaction of the local site, and this is assigned to the subtransaction execution task of the other site. Since processing can now be performed based on the task identifier, task identifier notification processing (BIND processing), which was previously necessary, has been made possible.
This makes it possible to eliminate the need for processing and improve processing efficiency.

〔Example〕

An embodiment of the present invention will be described based on FIG.

FIG. 1 (81 is a task identifier used in the present invention,
FIG. 2B is a schematic diagram for explaining the shield operation of the communication module in the present invention, and FIG. 1C is a configuration diagram of an embodiment of the present invention.

In the figure, 10 is a communication module (VCM), which corresponds to the communication module VCM in FIG. 4 (bl), and 11 is a TEM management task (TCM), which corresponds to the communication module VCM in FIG.
corresponding to the TEM management task (TCM) of
12 is a receiving section, 20 is an identifier generating section, and is shown in FIG.
The application program AP generates a task identifier as shown in l.
21 is a transmitter that is added to P1, APP2-, etc.

The communication module (VCM) 10 receives the sub-I/ranzaktion start message BEGIN transmitted from the other site.
receive subtransaction execution task TEMI,
This is passed to T E"M2"-, or when all subtransaction execution tasks are in use and there is no free space, a queue is created for the transmitted unexecuted subtransaction sb. VCMIO
The allocation request processing unit 13 allocates the transmitted BE'GIN to the subtransaction execution task TEM that is waiting for BEGIN, and the BEGIN for which allocation has not yet been executed.
a request storage unit 14 that creates and holds a queue by holding the data, and a communication control unit 15 that controls communication in this VCMIO.
Equipped with etc.

The identifier generator 20 generates a transaction identifier (TRT
D), for example, the time (time stamp) at which the transaction started execution. Typically, when executing a global transaction GT, a transaction identifier (TR'TD) that uniquely identifies the transaction across sites is used.
is added to each subtransaction ST, and this TRID is added to each subtransaction ST and communicated to each site.

Application program APP of own site and TE of other site
[Site identifier +
Use T'R'IDj. In this way, the task identifier will also be unique at each site, and at the subtransaction S'T execution site, if the site identifier of another site related to the subtransaction is known, the task identifier can be known and communication can be performed. Therefore, notification of the task identifier becomes unnecessary.

Here, in FIG. 1 (81 shows an example of a task identifier, the local time indicating the time when the transaction started execution, and the site identifier (81) where the transaction occurred)
Usually the host computer identification number M provided at that site
). For example, figure F tel
When the application program APPI shown in is started at time To and passed to site B, if the identification number of the host computer at site B is M2, the time To output from the identifier generation unit 20 is communicated from APPI. If it is included in BEGIN and executed by TEMI, then TEM
The task identifier of I is "M2ToM1". In this way, when the TEM receives any BEGIN message, the TRID added to the BEGIN message is
and the site identifiers of his/her site 1~ are combined to form a task identifier, and communication will be performed using this task identifier from now on.

The operation of the present invention will be explained.

Now, in Figure 1 (not shown in C1), the global
When the application program APP1 executes a transaction GT, this transaction is decomposed into a subtransaction of its own site A and a subtransaction of another site B. When the global transaction GT is executed, the local time generated from the identifier generation unit 20 at that time is added as a transaction identifier TRTD to the subtransaction start message BEGINI output from this application program APPI,
It is sent to Sai) B via the transmitter 21.

At site B, the reception unit 12 receives this and sends it to VCMIO. The communication control unit 15 stores a part of this BEGIN in the request storage unit 14, and the allocation request processing unit 13 stores the BEGI
This BEGTN is assigned to the subtransaction execution task TEM waiting for N. In this way, for example, TEMI
When executes this BEGINI, TEMI adds the site identifier M2 of site B (usually uses the host computer identifier as described above) to the transaction identifier TRID of BEGINI output from the application program APPI, and executes the task. Create an identifier. In this case, when the TRID is TOM+, the task identifier is M 2 T
o M 1, and communication is thereafter performed using this task identifier. Since this task identifier is uniquely determined as described above, the task identifier for it will be recognized at the time BEGINl is transmitted.

By the way, as mentioned above, even though the subtransaction start messages BEGTNI and BEGIN2- are held in the 70M request storage unit 14, if all the subtransaction execution tasks TEM are being executed, it is necessary to wait until the execution of the subtransaction execution tasks TEM is completed. Become. In the request storage section 14,
Create an unexecuted subtransaction queue sb1.5b2- corresponding to these unexecuted requests (see Figure 1 (bl
Reference) Wait for the completion of execution of the subtransaction execution task TEM.

When the TEM finishes executing one subtransaction, it requests the VCM to receive the next unexecuted subtransaction. If there is a BEGIN message that has not been executed at that time, the allocation request processing unit 13 of the VCM passes it to the TEM.

Otherwise, the TEM will be forced to wait until a new BEGIN arrives.

By the way, even if the subtransaction transmitted to site B is not executed at another site A because there is no free space in the TEM, it can be recognized because its task identifier is uniquely determined as described above. Processing can be carried out even without the return of the item. Therefore, there is no need to wait for the reception of ACK-BEGIN as in the past, resulting in improved efficiency. Furthermore, notification of task identifiers (BIND processing) is no longer necessary at the subtransaction execution site, which also improves efficiency.

In the conventional case, the TEM management task CMII is BEG
Although IN acceptance and queuing, TEM allocation processing, and a path function for messages such as BEGIN to TEM were required, the present invention eliminates the need for these functions at all, and TCM is capable of processing transaction error recovery in TEM and Since the functions such as generation of TEM can be limited, control is simplified, and the flow of messages via TCM is eliminated, and all messages are sent directly to TEM, which also improves efficiency. Become.

〔Effect of the invention〕

According to the present invention, when executing a global transaction, the processing could not proceed without knowing the identifier of the task of another site, but the process of returning the task identifier in such a case, that is, the BIND process, etc. Since this is no longer necessary, data processing efficiency can be improved.

[Brief explanation of the drawing]

Fig. 1 f8+ is an example of a task identifier used in the present invention, Fig. 1 (bl) is a schematic operational explanatory diagram of a communication module in the present invention, Fig.
The figure is an explanatory diagram of a distributed data storage system, Figure 3 is an explanatory diagram of a conventional fixed task identifier, and Figures 4 (a), (
bl is an explanatory diagram of the state of dynamic task assignment in a conventional distributed database system. In the figure, 10 is a communication module, 11 is a 78M management task, 12 is a reception section, 13 is an allocation request processing section, 14 is a request storage section, 15 is a communication control section, 20 is an identifier generation section, and 21 is a transmission section. .

Claims (1)

[Claims] In a distributed database system that includes multiple host computers, a local database connected to these host computers, and a global transaction that spans multiple sites, it is divided into subtransactions for each site and executed. The communication module is provided with an identifier generating means for identifying the application program task generated at the application program, and an identifier recognition means for identifying the application program identifier in the communication module, and the identifier generated by the identifier generating means is provided in the subtransaction execution task executed at another site. By adding and sending this identifier as the task identifier of the sub-transaction execution task in the communication module of the other site, when dynamically allocating and executing the sub-transaction execution task of the other site, the identifier of that task can be used. A global transaction execution control method characterized by being able to perform processing without waiting for notification.