JPH05334163A - Exclusive control system for shared memory between plural systems - Google Patents

Abstract

PURPOSE:To attain more efficient exclusive control by previously determining a system for managing the updating of a shared memory, and simultaneously with the start of updating processing, transferring exclusive control information to an updating management system. CONSTITUTION:Respective systems HOST1, HOST2 are respectively constituted of user application programs UAP1, UAP2 for accessing a shared data base(DB) C-1, DB accessing parts A-1, A-2 for attaining DB accessing, and DB distriibuting function parts B-1, B-2 for attaining transaction processing applied between plural systems. The HOST 1 is connected to the HOST 2 through a high-speed = bus such as a channel. In this case, a data updating system in the shared memory is previously determined, and simultaneously with the starting of a UAP for updating the shared memory, exclusive control information is transferred to the data updating system to execute the exclusive control of the memory.

Description

Detailed Description of the Invention

[0001]

[Industrial field of use] With the increase in scale and diversification of systems, the number of distributed systems for the purpose of function distribution, load distribution, and risk distribution is increasing. The present invention relates to a technique for sharing access to a database, which is a common resource, between systems in such a system.

[0002]

2. Description of the Related Art A conventional method of sharing a memory such as a database between systems is to determine a data update system in advance as described in Japanese Patent Laid-Open No. 3-222048 and update the data when a data update process occurs. This is a method of performing remote access processing via the system to realize exclusive control of the access target memory, and at the time of data reference processing, the memory is directly accessed from the process request source system without exclusion control.

[0003]

In the above prior art, the exclusive control of the memory device in the data update system is the data update processing timing (remote access processing timing). Therefore, the data exclusion and consistency guarantee of the data search process and the data update process to be executed are guaranteed by the UAP.
There is a problem that it needs to be done in.

There is also a problem that retries are not taken into consideration.

An object of the present invention is to provide a shared memory exclusive control system between a plurality of systems which solves the above problems.

[0006]

In order to achieve the above object, in a computer system in which a memory such as a database is shared by a plurality of systems, a system for managing the update of the shared memory is determined in advance, and a shared memory update management system is provided. When a system other than the above updates the shared memory, the system other than the update management source starts the process of updating the shared memory and transfers exclusive control information to the shared memory update management system at the same time Resource exclusion.

Also, the occupation of the update management source system is managed by deriving a pseudo transaction from the processing request source system.

When exclusive control is not normally performed in the update management source system, a failure notification is sent to the processing request source system and the processing is rescheduled.

If the exclusive control is not normally performed in the update management source system, the exclusive management is retried in the update management source system.

[0010]

Operation: The data update system for the shared memory is determined in advance, and at the same time when the UAP for updating the shared memory is started,
Exclusive control is performed on the memory by transferring exclusive control information to the data updating system. Since the validity of the exclusive control is guaranteed at the time of UAP synchronization processing, the consistency of data during UAP processing execution is guaranteed,
Since it is not necessary to perform data exclusive control by the UAP, the user can realize more efficient exclusive control without being aware of the data exclusive control.

Further, even when an update request for the shared memory is issued from a plurality of systems, the system is provided with a means for retrying the exclusive control process in the system or user designation, so that the load on the system operation manager is greatly reduced. It is also possible to apply the present invention to a system in which automatic operation of the system is performed, since it is also possible to retry exclusive control processing.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

The present invention can be mixed with distributed transactions, but for the sake of simplifying the explanation, a description will be given using a shared DB system in which a database (hereinafter abbreviated as DB) is shared by a plurality of systems. ..

FIG. 1 is a block diagram of a shared DB system for sharing a database according to an embodiment of the present invention among a plurality of systems.

HOST1 is a shared DB (C-1) update management system, and HOST2 is a shared DB (C-1) update management system.
It is a system to access to. HOST1, HO
Each system in ST2 has a user application program (UAP) for accessing the shared DB (C-1).
1, UAP2) and D that realizes database access
B access unit (A-1, A-2) and DB distribution function unit (B-1, B-2) for realizing transaction processing across a plurality of systems, and H
A shared DB (C-1) exists in OST1. HOS
DB (C-2) exists in T2. HOST1 and H
The OST 2 is connected by a high-speed bus 1 such as a channel, and is connected to a DB access unit (A-1, A-2) and a shared DB (C
-1), the DB access unit (A-2) and the DB (C-2) are also connected by the high speed bus 1 such as a channel.

(1) When the processing request source is HOST1 A processing flow of a DB access request of HOST1 from UAP1 in HOST1 will be described.

FIG. 3 is a system configuration diagram. FIG. 6 shows D
It is a figure which shows the system definition information management table in a B access part. The system definition information management table 6 includes information 61 indicating access target data, information 62 indicating an update management node, and information 63 indicating whether or not it is shared. The data A and B are stored in the shared DB (C-1), the data C is stored in the DB (C-2), and the data D is stored in the DB (C-3).

FIG. 5 is a flowchart showing the processing procedure. When UAP1 is started (5-1), the DB access unit (A-1) is called and the UAP scheduling process is executed (5-2). The UAP scheduling process is a process of loading a schema and a UAP environment definition accessed by the UAP from a system directory.

After that, the DB access unit refers to the system definition information management table 6 in the DB access unit with the identifier of the access target data as a key, and the access target data of the UAP is the data in the shared DB (C-1). It is discriminated whether or not it is (5-3).

If the data is not in the shared DB (C-1), the DB exclusive use declaration is made (5-6) and the DB access process is executed (5-), as in the normal local access.
7 and later). If it is the data in the shared DB (C-1), the own system judges whether or not it is the shared DB update management system (5-5).

If the own system is a shared DB update management system, it is judged whether or not the data is updated in the UAP process (5-14). If the data is not updated, the DB is updated.
The exclusive use declaration is not made, and thereafter the same processing as the normal local access is executed (5-7 and after). When updating is performed, the DB exclusive use declaration is made (5-6) and the access processing of the shared DB is executed (5-7 and thereafter) as in the normal local access. In this way, the exclusive control of the shared DB is realized by executing the exclusive use declaration in the DB access unit only when updating the data in the UAP process.

If the own system is not the shared DB update management system, it is judged whether or not the data is updated by the UAP process (5-4). If the update is not performed, the DB exclusive use declaration is made. No, DML (data manip
lation process) is performed (5-7 and subsequent steps are performed). If updating is performed, remote access processing described later is performed. The remote access processing will be described in the case of (2) the processing request source is HOST2.

As can be seen from the above description, for example, in the case of a data update request from the UAP1 in the HOST1 to the data A in the shared DB (C-1), which is the update management node, a normal local access is performed. Similarly, D
B exclusive use declaration is made (5-6), and shared DB access processing is executed (5-7 and thereafter).

The DML processing is similar to the normal local access and shared DB access. When the scheduling process of UAP1 and the DB exclusive use declaration are completed in the DB access unit, the control is once returned to UAP1. Upon receipt of the control, the UAP1 issues DML (5-7),
Control is passed again to the DB access unit.

In the DB access section, the DB index section is read (5-8) and the DB data section is read (5-9).
And DML processing is performed (5-10). This DML processing is processing corresponding to the issued DML, and searches for data /
Update etc. The result of this processing is returned to UAP1 (5
-11), UAP1 ends (5-12). UAP1
The data returned to is the DB access result (status
Data) or search data.

(2) When the processing request source is HOST2 Processing of an access request from UAP2 in HOST2 to data B stored in the shared DB of HOST1 will be described.

FIG. 2 is a system configuration diagram. FIG. 8 shows D
It is a figure which shows the system definition information management table in a B access part. The system definition information management table 8 is composed of information 81 indicating an identifier of access target data, information 82 indicating an update management node, and information 83 indicating whether or not the data is shared. Data A, B, C are shared DB (C-1)
Further, the data C is stored in the DB (C-2).

FIG. 7 shows the PDU format. The communication data (PDU) 7 is the communication prefix data 71.
And a DML processing data section 72. The contents of the DML processing data section 72 are different between the pseudo distributed transaction and the normal distributed transaction.
In the case of a normal distributed transaction, it is DB access information (index information, data, etc.), and in the case of a pseudo distributed transaction, it is dummy information (data for DB exclusive control information).

FIG. 5 is a flow chart showing the processing procedure. When the UAP2 is started (4-1), the DB access unit (A-2) is called, and the UAP scheduling process is executed (4-2).

In the DB access unit (A-2), the system definition information management table 8 in the DB access unit shown in FIG.
By referring to, it is identified that the access target data B of UAP2 is data in the shared DB (C-1). Further, it is identified from the system definition information that the shared DB update management system is HOST1. Also, it recognizes that the data is updated in the UAP process.

Since 4-3 in FIG. 4 is the same as 5-13 (portion within the dotted line) in FIG. 5, detailed description will be omitted.

The DB access unit (A-2) creates data for DB exclusive control information, and the DB distribution function unit (B-2).
Pass control to. At this time, in the DB access unit (A-2), as a sub-transaction of the UAP2 processing transaction, the DB access unit (A in the HOST1
For -1), a pseudo distributed transaction that causes exclusive control processing of the shared DB is derived.

A normal distributed transaction is generated by the generation of a processing request from UAP or the like, and is one processing unit for executing the processing. On the other hand, the pseudo-distributed transaction is a pseudo-transaction in the system (database management system) as if a processing request had been issued, although the processing request from the UAP or the like was not actually issued. It is a distributed transaction (processing unit) that is generated dynamically.

In the DB distribution function unit (B-2) to which control is passed from the DB access unit (A-2), a distributed transaction is generated as in the normal distributed transaction processing, and as shown in FIG. The DB exclusive control information data passed from the access unit (A-2) is stored in the DML processing data unit 72, and the communication prefix data 7 is stored therein.
1 is added, converted into communication data (PDU) 7 and transferred to the shared DB update management system HOST 1 (4
-4).

In the DB distribution function unit (B-1) of HOST1, the communication data (PDU) transmitted from HOST2.
Simultaneously with receiving 7 (4-5), a pseudo transaction is generated. This pseudo transaction process is
The access unit (A-1) is for performing exclusive control of the shared DB, and is for performing shared DB exclusive control in synchronization with the DB access unit (A-2) of the HOST2 that is the processing request source. Exclusive control of the DB access unit (A-1) to the shared DB is realized as a pseudo DB access request by pseudo transaction processing from the DB access unit (A-1) (4-6).

In the DB access unit (A-1), shared DB exclusive use declaration processing is performed by the shared DB exclusive control data passed from the DB distribution function unit (B-1), and exclusive control is started (4-7). ). The shared DB exclusive use declaration processing result is returned to the DB distribution function unit (B-1) as the pseudo DB access processing result.

When the DB distribution function unit (B-1) receives the pseudo DB access result from the DB access unit (A-1), it stores the pseudo transaction processing result in the DML processing data unit 72 and communicates with it. Prefix data 71 is added and converted into communication data (PDU) 7 and H
Transfer to OST2 (4-8).

In HOST2 which has received (4-9) the communication data (PDU) 7 including the pseudo DB access result data from HOST1, in HOST2, in the DB distribution function unit (B-2), the received data is converted to the DB distribution function unit (B). -1, B-2)
DML processing data portion 72 excluding the communication prefix data 71 between
It is returned to the B access unit (A-2).

The DB access unit (A-
In 2), it is determined from the data received from the DB distribution function unit (B-2) whether or not the shared DB exclusion process in HOST1 is normally performed (4-10). If the exclusive process is normal, the process for the shared DB (C-1) is continued. When performing update processing (issue update DML), shared DB
Since the shared database update management system HOST1 directly accesses (C-1), remote access via the communication path is executed for HOST1 (4-).
11 to 4-15). Since this method is known, it will not be described in detail in this embodiment. And
UAP2 ends (4-16).

In this embodiment, the DB access unit (A-
In 2), it is determined whether or not the shared DB exclusion process has been normally performed. However, after step 4-7, a process for determining whether or not the exclusion control is normal in the DB access unit (A-1) is provided. At the time of abnormality, the DB access unit (A-1) may perform the exclusive control until the exclusive control is normally performed.

On the other hand, reference processing (reference DM
When issuing L, the local access is performed by referring to the shared DB (C-1) through the direct path of the shared DB (C-1) and the DB access unit (A-2). Since this method is the same as the data reference processing described in item (1), it will not be described again in this item.
On the other hand, when the exclusive process is abnormal, the process scheduling is executed again in the DB access unit (A-2) (4-10).

The scheduling process in the DB access unit (A-2) is to discard the first pseudo sub-transaction process and perform the same pseudo sub-transaction process (distributed transaction process) again. still,
As this pseudo sub-transaction processing discard processing,
There are a plurality of methods, such as a method using only HOST2 and a method using HOST1 instruction, but in the present embodiment, the latter method is described.

In the distributed system, as a means for guaranteeing the consistency of transaction processing across a plurality of systems, 2
Adopts phase commitment. The two-phase commitment is well known and will not be described.

Therefore, the DB access part (A
The guarantee of the validity of the exclusive control in -1) can be realized by deriving the pseudo distributed transaction as described above. That is, a pseudo transaction is started before the transaction for processing the data in the shared DB (C-1) is started, and the shared DB (C-
At the end of the transaction for which the data in 1) is to be processed, prepare for commit including the pseudo transaction. As a result, it is possible to judge whether all the target transactions have been processed normally, and it is possible to confirm the validity of the exclusive control in the DB access unit (A-1) of HOST1. By rolling back the entire transaction when the commit preparation fails (failure occurs), the consistency of the data reference / update processing in HOST1 and HOST2 can be maintained.

The UAP mode to be activated is MPP (mes
In the case of the sage processing program), even if the synchronization process fails, the exclusive control can be newly started by deriving the pseudo distributed transaction at the start of the next new transaction.

[0046]

According to the present invention, a UAP for determining a shared memory data update system in advance and updating the shared memory.
At the same time as activating, the exclusive control is performed on the memory by transferring the exclusive control information to the data updating system. UA guarantees the validity of the exclusive control
Since it is performed during the P synchronous processing, the consistency of the data during the UAP processing is guaranteed, and there is no need to perform the data exclusive control by the UAP. Exclusive control can be realized.

Further, even when an update request for the shared memory is issued from a plurality of systems, the system is provided with a means for retrying the exclusive control process in the system or user designation, so that the load on the system operation manager is greatly reduced. It is also possible to apply the present invention to a system in which automatic operation of the system is performed because the exclusive control process is also retried.

[0048]

[Brief description of drawings]

FIG. 1 is a system configuration diagram of an embodiment of the present invention.

FIG. 2 is a system configuration diagram of an embodiment of the present invention.

FIG. 3 is a system configuration diagram of an embodiment of the present invention.

FIG. 4 is a flowchart showing a processing procedure.

FIG. 5 is a flowchart showing a processing procedure.

FIG. 6 is a diagram showing an example of a system definition information management table.

FIG. 7: Transfer data format (PDU) between DB distribution function units
It is a figure which shows the format of.

FIG. 8 is a diagram showing another example of the system definition information management table.

[Explanation of symbols]

HOST1 ... Shared database update management system, HO
ST2 ... System for accessing shared database, UAP1, UAP2 ... User application program, A-1, A-2 ... Database access unit, B
-1, B-2 ... DB distributed function unit, C-1 ... Shared database, C-2 ... Database.

 ─────────────────────────────────────────────────── ─── Continued Front Page (72) Inventor Tadashi Kitamura 6-81 Onoue-cho, Naka-ku, Yokohama-shi Kanagawa Hitachi Software Engineering Co., Ltd. In-house

Claims (4)

[Claims] 1. In a computer system in which a memory such as a database is shared by a plurality of systems, a system for managing the update of the shared memory is determined in advance, and a system other than the shared memory update management system updates the shared memory. , The system other than the update management source starts the process of updating the shared memory, and at the same time, transfers exclusive control information to the shared memory update management system to execute resource exclusion to the shared memory. A shared memory exclusive control method between multiple systems, which is characterized in that 2. A shared memory exclusive control system between a plurality of systems according to claim 1, wherein the occupation of said update management source system is managed by deriving a pseudo transaction from said processing request source system. 3. The update management source system, when exclusive control is not normally performed, issues a failure notification to the processing request source system and reschedules the processing. Shared memory exclusive control method between multiple systems. 4. The sharing between a plurality of systems according to claim 1, wherein when the exclusive control is not normally performed in the update management source system, the exclusive management is retried in the update management source system. Memory exclusive control method.