JPH03235130A - Exclusive priority control system - Google Patents

Abstract

PURPOSE:To improve the response of a transaction of a high degree of urgency by executing preferentially a transaction of a high degree of urgency against a transaction of a low degree of urgency which is being executed, in the case plural transactions use exclusively shared resources. CONSTITUTION:The table constitution of a transaction processing queue in a date base system in which plural transactions use exclusively shared resources has a transaction processing queue 10, a lock requiring pointer area 11, a transaction group table 12, a pointer 13, and a transaction 14. In such a state, with respect to resources for which the transaction of the high degree of urgency executes an exclusive control requirement in the case another transaction of the low degree of urgency has the exclusive control right, the transaction of the low degree of urgency is ended abnormally, and the exclusive control right is delivered to the transaction of the high degree of urgency. In such a manner, the response of the transaction of the high degree of urgency can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a database system having shared resources that are exclusively used by a plurality of transactions.

It also relates to exclusive priority control methods for transactions and processes in data processing systems that have shared resources that are used exclusively by multiple processes.

[Conventional technology]

Conventionally, as an exclusive priority control method for shared resources, for example, as described in Japanese Patent Application Laid-Open No. 17764/1982,
There is a method for improving the responsiveness of transactions with high urgency by chaining resource allocation requests for transactions in queues according to priority level and allocating resources in order of priority.

[Problem to be solved by the invention]

The above-mentioned conventional technology does not take into consideration the fact that when a high-urgency request is registered in the queue while a low-urgency transaction is already being executed, the high-urgency request is executed preferentially. However, there were limits to the improvement of responsiveness for highly urgent transactions.

An object of the present invention is to provide a transaction exclusive priority control method in which a transaction with a high degree of urgency is executed with priority over a transaction with a low degree of urgency that is currently being executed in a database system in which a plurality of transactions use shared resources exclusively. It is about providing.

A further object of the present invention is to provide a process exclusive priority control method that gives priority to processes with high urgency over processes with low urgency in a data processing system that has shared resources that are used exclusively by multiple processes. .

[Means to solve the problem]

In order to achieve the above object, the present invention provides that, in a database system having a shared resource that is exclusively used by a plurality of transactions, a transaction with a high degree of urgency can control the resource for which exclusive control is requested by another transaction with a low degree of urgency. is characterized in that when a transaction has exclusive control rights, the low-urgency transaction is abnormally terminated and exclusive control rights are transferred to a high-urgency transaction.

Furthermore, in a data processing system having a shared resource that is exclusively used by a plurality of processes, the present invention provides that a process with a low level of urgency has exclusive control over the resource for which a process with a high level of urgency has requested exclusive control. If there is a problem, the process with a low degree of urgency is suspended and exclusive control is passed to a process with a high degree of urgency.

[For production]

In database systems and data processing systems,
Lock requests by transactions and processes for shared resources are processed in order of priority, but in particular, the highest priority lock request is processed with the highest priority even if the lock is granted to another transaction or process with a lower priority. The lock grant is changed for each transaction or process.

This makes it possible to further improve the responsiveness of transactions and processes with high urgency.

〔Example〕

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

FIG. 1 shows the table structure of a transaction processing queue in a database system in which multiple transactions exclusively use shared resources. here,
There are N levels from 1 to N as the transaction processing priority for exclusive control requests (lock requests) to shared resources, and the higher the number, the higher the priority, and in particular, transactions with priority N have higher priority than other priorities. The transaction group table 12 assigns search chains to the six transamos that are to be processed with priority even if the transaction processing is currently being executed, and the transaction group table 12 assigns the above search chains to the transaction processing queue 10 according to the priority. Connected. Transaction 14 is.

Each group is connected to the transaction group table 12 of the group to which it belongs on a first-come, first-served basis.

Further, a lock request pointer area 11 of the transaction processing queue 10 exists for each of N priorities, and a pointer 13 is used to indicate the next lock request of each priority.

FIG. 2 shows the entry structure of the transaction group table 12. The transaction group table 12 includes a guaranteed processing value 16 of this group that is set in advance as an external parameter, a processing performance value 18 that indicates the number of transactions of this group that have been processed since the last time it was cleared to zero, and a processing result value 18 that indicates the number of transactions that have been completed in this group since the last time it was cleared to zero. Contains entries for 20 transactions. Also, 22,2
4 are entries into which the forward pointer and backward pointer of the search chain are entered, respectively. Note that the processing guarantee value 16 and the processing performance value 18 are prepared to prevent transaction sinking, and this is detailed in Japanese Patent Application Laid-Open No. 19957/1983.

FIG. 3 is a flowchart for selecting groups to which a primary lock is to be granted in order of priority in the table configuration shown in FIG. 1. first.

In step 31, a search for the transaction processing queue 10 is started from the N groups with the highest priority. Processing 32
It is determined whether the first lock request pointer 13 is null or not. If it is not null, a lock is granted to the first lock request in step 33, and the pointer 13 is shifted to the next lock request in step 34, and the search ends. Also, processing 32
If it is determined that the first lock request pointer 13 is null, in step 35 the process moves on to searching for a group with a lower priority.

In this way, when it is found in process 36 that the search has been completed up to the group with priority 1, the lock granting process is ended. As a result, locks will be granted from the group with the highest priority.

FIG. 4 is a flowchart when a lock is granted to a group with priority n (n is smaller than N) and a transaction with priority N is registered while a transaction with priority n is being executed. It shows the basic idea of

First, in step 41, it is confirmed that the current priority level n is not N. If a lock has already been granted to a group with priority N, or if the first lock request pointer 13 of priority N is null in step 42, there is no need to change the lock, and the search ends immediately. do. If it is determined in process 41 that no lock has been granted to a group with priority N, and in process 42 that a new lock request with priority N has occurred.

In process 43, the lock on the current priority level n is released, and in process 44, the lock is applied to the priority level N. and,
In process 45, the pointer 13 is shifted to the next lock request and the search is ended.

In this manner, by following the flowchart of FIG. 4, even if a lock is granted to a priority level lower than N, it is possible to release the lock and grant a lock to a group with priority level N.

FIG. 5 is a flowchart when selecting a transaction to be processed next. First, in process 51, a search is started in the transaction processing queue 10 starting from the N group with the highest priority. In process 52, it is determined whether all groups of the priority level have been searched, and if they have not been searched, process 5
In step 3, the forward pointer points to one of the transaction group tables, and in step 54 it is confirmed that there is a transaction waiting to be processed in the group.In step 55, the processing actual value of the group is compared with the guaranteed processing value. , if the processing guarantee value has not yet been reached,
In process 56, select the first transaction waiting in the transaction group on a first-come, first-served basis and end the search. 0 If there is no transaction to be selected in the group, the processing performance value of the transaction group has reached the guaranteed processing value. If so, the next transaction group pointed to by the forward pointer is searched for in step 53. On the other hand, if it is determined in process 52 that all transaction groups of that priority have been searched, the process moves on to search for transaction groups of lower priority in process 57. In this way, process 5
8 indicates that the search for the transaction group with priority 1 has been completed, the search ends without selecting any transaction.

The transaction selection process in Figure 5 described above is as follows:
This is done when the processing of one of the transactions has been completed and online system resources have become available, allowing the processing of a new transaction to begin.

Next, as a specific example, an example will be described in which a transaction with priority N is executed while a transaction with a long processing time is being executed in FIG. 11.

In FIG. 11, it is assumed that transaction 111, transaction 113, and transaction 115 share resource B, and exclusive control requests (lock requests) for resource B have priority 1, priority 3, and priority N, respectively. In particular, since transaction 115 has priority N, even if other transactions are being executed, the lock request of the transaction being executed is released according to the flowchart of FIG. 4, so that transaction 115 can be executed with priority. In FIG. 11, the processing is represented by the time axis t.

Transaction 111 is a transaction with a long processing time, starts execution at time t, and uses resources A and B.
, C is used to perform transaction processing. Also, transaction 113 starts execution from time t, but at time t, resource B is transferred to transaction 111.
Since resource B is in use, it is in a state where it is waiting for resource B to be unlocked.

On the other hand, transaction 115 starts execution at time t, but since it is a transaction with priority N for resource gB, transaction 111 using resource gB is abnormally terminated at time t7, and resource B is to continue processing. Since the transaction 111 unlocks the capital gB at time 1s and terminates abnormally, it is rescheduled after partial transaction recovery processing is performed.

Meanwhile, transaction 115 executes processing using resource B and ends at time t.

Next, an embodiment in which the present invention is applied to exclusive priority control of processes will be described.

FIG. 6 shows a table structure of a process queue in a data processing system in which multiple processes exclusively use shared resources. Here, as in the case of Figure 1, it is assumed that there are N levels of process priority from 1 to N for exclusive control requests (lock requests) to shared resources, and the higher the number, the higher the priority. shall be. In particular, it is assumed that a process with priority N is processed with priority even if processes with other priorities are being executed. The process processing queue 60 is
Each priority has a search chain, and the process group table 62 is connected to the search chain of the process processing queue 60 according to the priority. Each process 64 is connected to the process group table 62 of the group to which it belongs on a first-come, first-served basis. In addition, the lock request pointer area 61 of the process processing queue 60
exists for each of N priorities, and uses a pointer 63 to indicate the next lock request for each priority.

FIG. 7 shows the entry structure of the process group table 62. The process group table 62 is
The processing guarantee value 66 of this group, which is set in advance as external parameters, the processing performance value 68, which indicates the number of processes of this group that have completed processing since the last time it was cleared to zero, and the entry of 70, the number of processes currently waiting to be executed in this group. include.

Further, 72 and 74 are entries into which the forward pointer and backward pointer of the search chain are entered, respectively.

FIG. 8 is a flowchart for selecting the next group to be locked in descending order of priority in the table configuration shown in FIG. 6. First, in process-81, a search is started from the N group with the highest priority. In process 82, it is determined whether or not the first lock request pointer 63 is null. If it is not null, a lock is granted to the first lock request in process 83, and in process 84, the pointer 63 is shifted to the next lock request and the search is performed. finish. Also, if the first lock request pointer 63 is null as determined in process 82;

In process 85, the search moves to a group with a lower priority.

In this way, when it is found in process 86 that the search has been completed up to the group with priority 1, the process ends. As a result, locks will be granted from the group with the highest priority.

FIG. 9 is a flowchart when a lock is granted to a group with priority n (n is smaller than N) and a process with priority N is registered while the process with priority n is being executed. First, in step 91, it is confirmed that the current priority level n is not N. If a lock has already been granted to the group of priority N, or if the first lock request pointer 63 of priority N is null in step 92, there is no need to change the lock, and the process ends as is. When it is found in process 91 that no lock has been granted to the group with priority N and that a new lock request with priority N has occurred in process 92, process 9
3 to unlock the current priority n, and process 94
gives a lock to priority N. And processing 9
In step 5, the pointer 63 is shifted to the next lock request and the search is ended.

FIG. 10 is a flowchart when selecting a process to be processed next. First, in process 101, a search is started from the group with the highest priority N. In process 103, one process group table 62 of the priority level is pointed to, and when it is confirmed in process 104 that there is a process waiting to be processed in the group, in process 105, the processing performance value of the group is set as the guaranteed process value. After comparison, if the guaranteed processing value has not been reached yet, in step 106, the first process waiting in the process group is selected on a first-come, first-served basis, and the search is ended. If the process to be selected is not in the group, or if the processing performance value of the process group has reached the guaranteed processing value, in step 103, the next process group pointed to by the forward pointer is searched. on the other hand.

When it is determined in process 102 that the search for the process group with the priority level has been completed, the process moves on to search for a process group with a lower priority level in process 108.

When it is determined in step 108 that the search for the process group with priority 1 has been completed, the search ends without selecting any process.

Note that the selection of the process shown in FIG. 10 is performed when the processing of one of the processes is completed and there is sufficient online system resources, etc., and it becomes possible to start processing execution of a new process.

As a specific example of the above process processing, an example in which the OS task releases the work area reserved for another task after the system abnormally ends will be described with reference to FIG.

In FIG. 12, it is assumed that process D, process E, and process F share a work area, and process D has priority N for a lock request for the work area. Therefore, the lock request of the task 121 of the process D can be executed with priority according to the flowchart of FIG. 9 even if the task 123 of the process E and the task 125 of the process F are being executed. In FIG. 12, the task processing of the process is represented by the time axis t.

When tasks 121, 123, and 125 are activated in each process, they are processed after securing a work area. Time t
When a failure occurs in task 121 in step 6, process D starts system error processing, but task 1 of process E
23 and task 125 of process F have secured a large operating area, it is not possible to secure a working area for the above-mentioned system abnormality processing. In this case, since the task of process D has priority N, the work areas secured by tasks 123 and 125 are released, and process E
and suspends process F. This allows process D
A work area for the system abnormal termination outer ring is secured, and the system abnormal processing is completed at time t.

〔Effect of the invention〕

As is clear from the above description, according to the present invention.

In a database system where multiple transactions use shared resources exclusively, when a transaction with high urgency is registered while a transaction with low urgency is being executed, the transaction with low urgency is abnormally terminated and the transaction with high urgency is executed. transaction can be executed,
This has the effect of allowing intermittent response processing to be performed during partial recovery processing for transactions that take a long time to execute.

In addition, in a data processing system in which multiple processes use shared resources exclusively, when a process with a high level of urgency is registered while a process with a low level of urgency is being executed, the process with a low level of urgency is paused and the level of urgency is increased. Since it is possible to execute a process with high performance, it is possible to secure an area from an OS task with a system error and release an area for other tasks.

[Brief explanation of drawings]

Figure 1 is a diagram showing the table structure of the transaction processing queue, Figure 2 is a diagram showing the entry structure of the transaction group table, Figure 3 is a flowchart of granting a lock to a transaction group, and Figure 4 is a diagram showing the maximum Flowchart for preferentially granting a lock to a priority transaction group, Figure 5 is a flowchart for preventing transaction sinking, Figure 6 is a diagram showing the table structure of the process processing queue, and Figure 7 is a diagram showing the process processing queue table structure.
A diagram showing the entry configuration of the group table, Figure 8 is a flowchart of granting a lock to a process group,
Figure 9 is a flowchart for preferentially granting a lock to the highest priority process group, Figure 10 is a flowchart for preventing process stagnation, and Figure 11 is a flowchart for locking the highest priority process group while the highest priority transaction is being executed. FIG. 12 is a diagram illustrating an example of executing the abnormal termination process by causing the OS task to release the work area secured by another task after the system terminates abnormally. DESCRIPTION OF SYMBOLS 10... Transaction processing queue, 11... Lock request pointer area, 12... Transaction group table, 13... Pointer, 14
... Transaction, 60 ... Process processing queue, 61 ... Lock request pointer area, 62 ... Process group table. 63...Pointer, 64...Process. Figure 6 62. -7'mouth tsufuqL-7'1 child-7IL-63
76

Claims (2)

[Claims] (1) In a database system that has a shared resource that is used exclusively by multiple transactions, another transaction with a lower urgency has exclusive control over the resource for which a transaction with a higher degree of urgency has requested exclusive control. 1. An exclusive priority control method characterized in that when a transaction with a low degree of urgency is terminated abnormally, the exclusive control right is passed to a transaction with a high degree of urgency. (2) In a data processing system that has shared resources that are used exclusively by multiple processes, when a less urgent process has exclusive control over the resource for which a more urgent process has requested exclusive control. An exclusive priority control method characterized in that the process with a low degree of urgency is suspended and the exclusive control right is passed to a process with a high degree of urgency.