JPH0679284B2 - De-dock prevention method for distributed database system - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a deadlock prevention method for a distributed database system, and sometimes to reduce the transfer amount of control information for deadlock detection between distributed systems. The present invention relates to a suitable deadlock prevention system.

[Conventional technology]

In a distributed database system, a plurality of data processing devices each have a distributed database, and each data processing device can arbitrarily access these databases. In such a distributed database system, when the data resources that the data processing device tries to access are being used by other data processing devices and all the data processing devices are in a waiting state, the system I am in a deadlock. Therefore, conventionally, the order of access to the resources of each distributed database is set in advance, and the order of access is strictly adhered to in each data processing device to prevent the system from falling into a deadlock.

The prior art of the deadlock detection and prevention method in this kind of distributed database system is, for example, GA Champain published by Planning Center Co., Ltd. Translated by Fujita "Distributed computer system-impact on management, design and analysis" (1981). , Pp. 338-341.

[Problems to be solved by the invention]

In the prior art, deadlock can be strictly prevented by setting the order of access to the distributed database in advance, but resource allocation occurs because a wait flag is created for deadlock detection. For example, the resource allocation information is transferred to all distributed systems (nodes) every time, so that the amount of information transfer between nodes increases, and the overhead for that becomes a problem.

An object of the present invention is to provide a distributed database system,
It is to reduce the transfer amount of control information for deadlock detection between distributed systems.

[Means for solving problems]

The above-mentioned purpose is a means for each distributed system to determine whether a request from a program that processes a transaction (hereinafter referred to as a process) violates a predetermined resource access order, and if it violates, the access Provide a means to determine whether the result may deadlock, allow access to the distributed database to violate a predetermined resource access order, and allocate the violated request resource to other processes It is achieved by not accepting the request only when it is requested.

[Work]

In the present invention, paying attention to the fact that the actual distributed system presets the order of access to the database in order to prevent frequent occurrence of deadlocks, and the database requested by a certain process is being allocated to another processor. The possibility of deadlock is prevented by not accepting the access request only when the access request is present and the access request violates a preset access order. Therefore, if the access sequence numbers are assigned to the databases according to the access sequence, the control information for preventing deadlock between distributed systems will be limited to the access sequence numbers indicating to which database the process is accessing. can do. In fact, by comparing the access sequence number of the database currently requested by the process with the maximum access sequence number of the databases accessed in the past,
It is possible to judge whether the current access request violates the access order.

〔Example〕

 An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an embodiment of a distributed database system to which the method of the present invention is applied. In FIG. 1, each of the data processing devices 1, 10 and 11 has a database 2 and the data processing devices are connected to each other by a communication line 3. Each data processing device 1, 10, 11 has a communication program 4, a control system 5 and a process 6, 7 or 8
Have software resources. Further, the control system 5 includes a request transfer program 51, a request acceptance program 52, a program management table 53, a resource management table 54, and an access execution routine 55.

The communication program 4 provides a function of transmitting a database access request to another data processing device and receiving a response thereof.

Processes 6, 7 and 8 are program parts for executing business, and access to the database is controlled by the control system 5.
Run through. The process status includes running status, resource waiting status, and cancel status, and the status is represented on the process management table 53.

The database 2 is a set of user data accessed through the control system 5. The resource allocation status of the database 2 has two states, i.e., empty and specific process allocation, and is represented on the resource management table 54.

The request transfer program 51 of the control system 5 provides the function of assembling the access request message of another distributed database system. Request acceptance program
52 is a core part of the present invention, which determines whether a request from each process violates a resource access sequence number, allocates a required resource to the process, and allocates the resource to another process. A function is provided to determine whether or not allocation is in progress and to cancel the process when the resource of the request that violates the access sequence number is being allocated to another process. The request acceptance program 52 also manages the process management table 53 and the resource management table 54. The access execution routine 55 executes the actual access to the database 2.

Next, the management of the access sequence and its number will be described.

The access order of the database 2 is set by the system designer, and the result is used as a resource management table 54 as shown in FIG. 3 to create a table of resource IDs and access order numbers. In this example, database A and database B are ordered to be accessed before database C and database D, and database C is ordered to be accessed before database D. The resource allocation status indicates that the processes 6, 7, and 8 access the database B, the database A, and the database D, respectively, and each database is allocated to the process. Also,
The database C is empty.

The states of the processes 6, 7 and 8 are managed by a process management table 53 as shown in FIG. In this example, the process 6 is continuing the processing after accessing the databases with access sequence numbers up to 1, and the process 8 is continuing the processing after accessing the databases with access sequence numbers up to 6. In addition, the process 7 waits for the database D because the database D is allocating to the process 8 after trying to access the database D after accessing the databases having the process sequence numbers up to 1.

The process management table 53 and the resource management table 54 are distributed and managed in the respective data processing devices 1, 10 and 11.
In the case of the process management table 53 of FIG. 2, the states of the process 6 and the process 8 are managed by the data processing device 11 and the data processing device 1, respectively, and the state of the process 7 is in the database D waiting state. Managed by 11. The resource management table 54 of FIG. 3 is managed by the data processing device in which each database exists. That is, the database A is the data processing device 1
The databases B and C are managed by the data processor 10, and the database D is managed by the data processor 11.

In the state shown in FIGS. 2 and 3, the process 8 is the database C.
When accessing, the request message in the format shown in FIG. 4 is transferred from the data processing device 1 to the data processing device 10. In FIG. 4, process ID = process 8,
Maximum access sequence number = 6, resource ID = database C. The data processing device 10 determines that this request violates the access order. However, since the database C is empty, it is assigned to the process 8 and
The database C is accessed. The access result is transferred from the data processing device 10 to the data processing device 1 in the format shown in FIG. In this case, the maximum access sequence number is set to 6.

In the state of FIGS. 2 and 3, the process 8 is the database B.
Similarly, in the case of accessing, the request message of FIG. 4 is transferred from the data processing device 1 to the data processing device 10. In this case, process ID = process 8, maximum access sequence number = 6, resource ID = database B. In the data processing device 10, the access sequence number of the database B is 1, which is equal to or less than the maximum access sequence number, and therefore the request is determined to be in violation of the access sequence. Moreover, since the database B is being allocated to another process 6, the data processing apparatus 10 requests the data processing apparatus 1 to cancel the process 8 because the deadlock may occur.

Next, the request transfer program 51 and the request accepting program 52, which are programs for executing the transfer processing and the accepting processing of the access request to the other data processing apparatus in this way, will be described.

The request transfer program 51 performs the processing shown in FIG. First, when a process issues a request addressed to another system, the maximum access sequence number of the process is obtained from the process management table 53 (process 101). The process ID and the maximum access sequence number are set in the control information to create the request message shown in FIG. 4 (process 102). The communication program 4 is requested to transfer the request message created in this way (process 103).

The request receiving program 52 performs the processing shown in FIG. First, when a request message is received from the communication program 4, the process ID and the maximum access sequence number are set in the process management table 53 (process 201). It is checked whether the requested resource is being allocated to another process (process 202). If it is being allocated, then it is checked whether the access sequence number of the resource is larger than the maximum access sequence number of the process (process 203). If the request is in the order of access, the access sequence number of the resource is higher than the maximum access sequence number of the process. In this case,
The process enters a resource waiting state (process 204). If the access order is not met, the access order number of the resource becomes smaller than or equal to the maximum access order number of the process. In this case, deadlock may occur and the process is canceled ( processing
205). If the requested resource is not being allocated,
It is checked whether the access sequence number of the resource is larger than the maximum access sequence number of the process (process 206), and if it is large, the access sequence number of the resource is set to the maximum access sequence number of the process management table 53 (process). 20
7). Then, assigning the resource to the process,
Access is executed (process 208).

According to the present embodiment, the control information transferred between the systems can be limited to the process ID and the maximum access sequence number, and the information exchange between the systems at the time of deadlock determination is completely deleted. This has the effect of reducing the amount of control information transferred.

〔The invention's effect〕

According to the present invention, in a distributed database system,
The control information for deadlock prevention can be limited to the process ID and maximum access sequence number, and the possibility of deadlock can be determined without inquiring to other systems. Is effective in reducing the control information transfer amount.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of the entire distributed database system according to the present invention, FIG. 2 is a detailed diagram of a process management table, FIG. 3 is a detailed diagram of a resource management table, and FIG. 4 is a request between systems. Detailed view of message, Figure 5 is a detailed view of access result message between systems,
FIG. 6 is a processing flow chart of the request transfer program, and FIG. 7 is a processing flow chart of the request receiving program. 1,10,11 …… Data processing device, 2 …… Database, 3 …… Communication line, 4 …… Communication program, 5 …… Control system, 6,7,8 …… Process, 51 …… Request transfer program, 52 …… Request acceptance program, 53 …… Process management table, 54 …… Resource management table, 55 …… Access execution routine.

Claims (1)

[Claims] 1. A distributed database system in which a plurality of data processing devices each distribute a database, and any of the data processing devices can arbitrarily access the distributed database, each of the distributed databases. Corresponding to the above, an access sequence number indicating the access sequence in the entire database is provided, and the maximum access sequence number indicating the maximum access sequence among the already accessed distributed databases is sent to another data processing device. A first data processing device that transfers an access request, and, when the access request is received, assigns the access request if the distributed database is not assigned to another request, and makes an access to the own distributed database. Sequence number is the maximum access order If it is not greater than the number and the distributed database is already assigned to another request, the access request is canceled, the access sequence number is larger than the maximum access sequence number, and the distributed database is already assigned to another request. A deadlock prevention method for a distributed database system, comprising: a second data processing device that puts the access request in a waiting state when allocated.