JPS61131043A - Inter-task queue controlling system - Google Patents

Abstract

PURPOSE:To cope with a fault by only recording and erasing a relevant QUEUE entry in and from a check point file, by setting implementing order information indicating the implementing order of QUEUE entries in a relevant task in each QUEUE entry. CONSTITUTION:When a system fault occurs, a CPU 41 first reads the content of a check point file 46 in an external storage device 44 into a QUEUE entry pool 60. Then the CPU 41 checks all QUEUE entries in the pool 60 and couples tasks by setting next entry pointers in the order of serial numbers at every task ID. Thereafter, the CPU 41 sets the location of a QUEUE entry 62 restored in the pool 60 in the first entry pointer of a control entry 61 in a queue control table 50 and the location of another QUEUE entry 63 in the final entry pointer. Furthermore, the CPU 41 sets the serial number 3 of the entry 63 in the serial number counter of the control entry 51.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to an inter-task queue control method in online task processing.

[Technical Background of the Invention] Generally, when constructing an online system, in order to improve response performance and establish a module configuration, processing is divided into processing tasks for each function (in the example of , tasks A, B, and C). Processing requests (intertask queue, i.e., intertask queue, hereinafter referred to as intertask Q
(referred to as LJE) is transferred and processed.

As shown in FIG.
It is created in a free QUE entry in the UE entry pool 14. This is called generation of an intertask queue entry (QUE entry). FIG. 5 shows QUE entries 5 to 17 for task A and task B.
A state in which QtJE entry 8 is generated is shown. QUE entries 5 to 18 are inter-task queue information (hereinafter referred to as QUE information) passed to the corresponding task.
has. QUE information is retrieved in the order of input based on a first-in, first-out method.

Therefore, in order to realize the first-in, first-out method, as shown in Figure 5, the QUE for each task is
A queue control table 20 for managing entries is secured in the main storage device 13. In the queue control table 20, control entries specific to each task are prepared. Fifth
The figure shows a control entry 21 for task A and a control entry 21 for task B.
A control entry 22 for is shown. control entry 2
1.22, a first entry pointer indicating the position of the first entry among all QLJE entries belonging to the relevant task, and a last entry pointer indicating the position of the RR entry are set. The first entry pointer determines the QIJE entry to be retrieved next in the relevant task, and the final entry pointer determines the QUE entry to be linked to the newly generated QLIE entry.

In addition, next entry pointers indicating the position of the QUE entry to be processed next to the same entry are set in QUE entries 15 to 18 (however, since QUE entries 17 and 18 are the final QLIE entries for tasks A and B, (0 is set as the entry pointer) QuE entries belonging to the same task are linked in the order of input.

Therefore, the contents of the queue control table 20 and the QUE entry pool 14 secured in the main storage device 13 allow the extraction and processing of each QUE information and the creation and deletion of QtJE entries to be executed correctly. However, if there is a system failure, the queue control table 20
And the contents of the QUE entry pool 14 may be destroyed or lost.

Therefore, conventionally, when a QLJE entry is created or deleted with update of pointer information, the entire queue control table 20 and the QtJE entry pool 14 are used as a checkpoint, and an external storage device 22 such as a magnetic disk device is used.
The data is written to the checkpoint file 24 prepared in the. According to this method, even if a system failure occurs, by using the contents of the checkpoint file 24, the queue control table 20 and Q
The UE entry boolean 14 can be restored to the state before the failure occurred.

[Problems with the background art] However, as described above, the queue 11i11 I21
In the method of writing one table and the entire QUE entry bool as a checkpoint, especially in the case of a large-scale system in which the QUE entry bool becomes large, there is a problem that writing the checkpoint takes a long time and the response performance deteriorates.

[Purpose of the Invention] This invention was made in view of the above circumstances, and its purpose is to reduce write data loss during checkpoint writing performed in preparation for the occurrence of a system failure, thereby shortening checkpoint writing time. The object of the present invention is to provide an inter-task queue control method that can further improve response performance.

[Summary of the Invention] In this invention, the QUE entries (intertask queue entries) held in the QUE (queue) entry boolean prepared in the main memory contain not only intertask QUE (queue) information and next entry pointers, but also It also has a task ID indicating the task to which the entry belongs. Further, the main storage device is provided with a queue cut surface table in which a first queue entry pointer and a last queue entry pointer are set for each task in the QUE entry pool.

Furthermore, in this invention, a checkpoint file is prepared. This checkpoint file contains the above QU
When generating a QIJE entry in the E entry boolean, the corresponding QUE entry is recorded. Also, QtJE
When a QUE entry disappears in the entry boolean, the corresponding QUE recorded in the checkpoint file will be deleted.
The UE entry is cleared.

According to this invention, when a system failure occurs,
Each QUE entry recorded in the above checkpoint file is read into the QUE entry pool during the recovery process, and the next entry pointer in each QUE entry is set for each common task ID and concatenated. In addition to restoring the contents of
By setting the positions of the first and last QUE entries for each task in the restored QUE entry boolean in the queue control table, the contents of the table are restored.

In addition, in this invention, the contents of the QUE entry boolean are
For systems that require restoration in the order in which QUE entries are created, each QUE entry is set with creation order information indicating the order in which the same entry is created within the corresponding task. In this case, when each QUE entry recorded in the checkpoint file is read into the QUE entry boolean after a system failure, the next entry pointer of each QUE entry is set in the generation order indicated by the above generation order information for each common task ID. By concatenating them, the contents of the QUE entry Boolean are restored to the order in which the QUE entries were created before the system failure. Furthermore, the contents of the queue control table are restored by setting the positions of the first and last QUE entries for each task in the QUE entry boolean thus restored.

[Embodiment of the Invention] An embodiment of the invention will be described below with reference to FIGS. 1 to 3.

In FIG. 1, a computer main body 43 having a CPU 41 and a main storage device 42 and an external storage device 44 such as a magnetic disk device are connected via a channel (not shown), an external bus 45, and the like. The main storage device 42 has a queue control table 50 that manages intertask QUE entries for each task, and a QUE entry pool 60 that holds intertask QUE entries for the entire system.

In the samurai matrix control table 50, the QUE of tasks A and B is
Control entries specific to each task are provided, such as control entries 51 and 52 for indicating the entry. Control entries 51 and 52 are the first Q to the corresponding task (A, B).
It has a first entry pointer indicating the position of the UE entry, an R end entry pointer indicating the position of the final QUE entry for the corresponding task, and a serial number counter indicating the number of QUE entries generated for the corresponding task.

The QtJE entry boolean 60 has an area reserved for QtJE entries, and QtJE entries can be added to the empty area as needed.
A UE entry is generated. FIG. 1 shows QUE entries 61 to 63 for task A and QUE entries 61 to 63 for task A.
A state in which an entry 64 is generated is shown. Q
The UE entries 61 to 64 contain a next entry pointer indicating the position of the QUE entry to be processed next to the relevant entry, and QUE information (inter-task QUE) passed to the relevant task.
In addition to the well-known information such as UE information), a task ID indicating the destination task of the QUE entry and an intra-task serial number (hereinafter simply (referred to as a serial number).

In this example, task A's QUE entry is QUE
Entry 61, QUE entry 62, and finally QUE entry 63 are generated in this order. Therefore, the sequence numbers of QUE entries 61, 62, 63 are 1, 2, .
It is set to 3. Further, the next entry pointer of the QUE entry 61 points to the position of the QUE entry 62 □,
The next entry pointer of QIJE entry 62 is QIJE
It is set to point to the position of entry 63. Since this QIJE entry 63 is the final QUE entry for task A at that point, the next entry pointer is set to 0. On the other hand, QUE entry 64 is the first QUE entry to task B, and is also the final QUE entry at that point. Therefore, Q
The next entry pointer of the UE entry 64 is set to 0 and the sequence number is set to 1.

The external storage device 44 has a checkpoint file 46. Checkpoint file 46 contains QU
An entry area corresponding to each QUE entry area in the E entry pool 60 is reserved.

Each time a QUE entry in the QUE entry pool 60 is generated, the QUE entry is written to the corresponding entry area in the checkpoint file 46. Further, the QUE entry written in the checkpoint file 46 is erased (for example, cleared to 0) in response to the disappearance of the corresponding QUE entry in the QUE entry boolean 60. Therefore, the QUE entry pool 60
In this example, QUE entries 61-64 are generated in the checkpoint file 46, as shown in FIG. 1, QUE entries 61-64 are written as QtJE entries 71-14. Since a QUE entry is written when the entry is generated as described above, the next entry pointers of the QLIE entries 71 to 74 are all set to O.

Next, the operation of an embodiment of the present invention will be explained in which task A is at the head of task Q.
An example will be explained in which QUE information in the UE entry 61 is extracted and processed, and then a QUE entry for task B is generated. When the processing of the QUE information of the QUE entry 61 is completed, the CPtJ 41 transfers the start entry pointer of the control entry 51 for task A prepared in the queue control table 50 to the QUE entry 61 in order to delete the QUE entry 61. Update the contents of the next entry pointer within. This causes control entry 5
The first entry pointer of 1 is Q as shown in Figure 2.
It now points to the position of the QLJE entry 62 to be processed next to the UE entry 61.

Next, the CPU 41 creates a QUE entry for task B by overlaying it on the QUE entry 61 that task A was processing. That is, the CPU 41 writes QUE information to task B (inter-task QUE information) to the QLJE entry 61.
information) and then sets the next entry pointer to O. Further, the CPU 41 updates the final entry pointer of the control entry 52 for task B prepared in the queue control table 50 to point to the position of the QtJE entry generated by the above-described overlay, and updates the serial number counter. Add +1 to 2. Also, c p U
41 sets the position of the generated QUE entry to the next entry pointer of the QUE entry 64 that the last entry pointer of the control entry 52 has been pointing to.

As a result, the QUE entry generated above is the last QUE entry (and the first QUE entry is
(also the E entry) is connected to the QUE entry 64 that was there. Further, the CPU 41 updates the task ID of the generated QtJE entry to indicate task B, and sets the serial number to the value (2) of the serial number counter after the counting. In this way, the QUE entry 65 for task B is generated in the 8i area of the QLIE entry 61 for task A, as shown in FIG. That is, when the QtJE entry 61 disappears, the QUE entry 65 is generated.

c p U 41 is QtJE entry 6 to task B
5, the area corresponding to the same entry 65 in the QUE entry pool 60 (on the external storage device 44)
aji in checkpoint file 46! Write the contents of QLJE entry 65 to . That is, the CPU 41 has Q
The contents of the QUE entry 65 are stored in the area of the QUE entry 71 corresponding to the UE entry 61 as shown in FIG.
Overlaid as UE entry 75. That is, Q(
Q corresponding to JE entry 61 (delete JE entry 11 and Qt corresponding to QUE entry 65)
5 recording is performed simultaneously. In addition, the QU that should disappear
When creating a QUE entry in an area different from the E entry area, it is necessary to erase the contents of the corresponding area of the checkpoint file 46 when the QUE entry disappears.

Now, assume that a system failure occurs in the state shown in FIG. In this case, the contents of the queue control table 50 and the QUE entry pool 60 in the main storage device 42 may be destroyed or lost. For this reason,
In recovery processing after a system failure, it is necessary to restore the contents of the queue control table 50 and QUE entry boolean 60 to the state immediately before the system failure occurred. Therefore, the CPU 41 first reads the contents of the checkpoint file 46 in the external storage device 44 into the QLJE entry boolean 60. As a result, the contents of QUE entries 2 to 75 in the checkpoint file 46 are written to the QLJE entry pool 60 as QtJE entries 82 to 85, as shown in FIG.

Next, the CPU 41 examines all QUE entries in the QUE entry pool 60, sets and connects next entry pointers in ascending order of serial numbers for each task ID. for example,
The QUE entry for task A is QUE entry 82.
83, and their serial numbers are 2 and 3. Therefore, the CPU 41 sets the next entry pointer of the QUE entry 82 to the position of the QUE entry 83 and sets the next entry pointer of the QtJE entry 83 to O. As a result, QIJE entry 82.83 becomes
Restore as QLJE entry 62.63 in FIG.

Subsequently, the CP tJ 41 reads the queue control table 50.
The first entry pointer of the control entry 51 in
QtJE entry 6 restored to E entry pool 60
2 (first QUE entry to task A), and also set the position of QUE entry 63 (first QUE entry to task A) to the last entry pointer.
Set the position of the first QLIE entry to task A. Then, the CPU 41 sets the serial number (3) of the QLIE entry 63, which is the final QUE entry, in the serial number counter of the control entry 51. As a result, the contents of the control entry 51 are restored. By performing the above processing for all valid tasks in the QUE entry bool 60, the contents of the queue control table 50 and the QUE entry bool 6° are restored.

In the above embodiment, a case has been described in which a serial number counter is provided in the control entry, but the present invention is not limited to this. For example, when generating a QUE entry,
By setting the value obtained by adding 1 to the sequence number (intra-task sequence number) of the final QUE entry at that time as the sequence number of the generated QUE entry, it is possible to eliminate the need for a serial number counter in the control entry. Furthermore, in the above embodiment, a case has been described in which a serial number (intra-task serial number) is applied as information indicating the generation order of QUE entries (for each task), but it is also possible to use time information based on the system clock, etc. It is. In short, when restoring the queue control table and QUE entry pool,
Any information that can connect QIJE entries (for each task) in the order of generation is sufficient.

Furthermore, in the above embodiment, a case was explained in which the generation order of QLIE entries is also completely restored, but in a system where it is not necessary to restore the QLIE entries in the generation order, it is sufficient that the destination task can be identified, and the sequence number (not to mention the serial number counter) is sufficient. is not necessary.

[Effects of the Invention] As detailed above, according to the present invention, when a QUE entry is created or deleted, only the relevant QUE entry is recorded or deleted from the checkpoint file, and system failures can be prevented. I can handle it. Therefore, compared to the conventional method in which the entire queue control table and QtJE entry boolean are subjected to checkpoint writing, the checkpoint writing time can be shortened and the response performance can be further improved.

[Brief explanation of the drawing]

The first reason is a correlation diagram between the queue control table, QUE entry boolean, and checkpoint file before QUE entry generation in an embodiment of the present invention, and FIG. 2 is a diagram showing the state after QUE entry generation in FIG. 1. , FIG. 3 is a diagram showing the state in which the contents of the checkpoint file are read into the QUE entry after the system failure occurs in FIG. 2, and FIG. 4 is a diagram explaining the relationship between tasks and intertask queues (QUE). FIG. 5 is a correlation diagram between a queue control table, a QUE entry pool, and a checkpoint file before QUE entry generation in the prior art. 41...CPtJ, 42...Main storage device, 46...
- Checkpoint file, 50... Queue control table, 51° 52... Control entry, 60...
QUE (queue) entry boolean, 61~65.71~
75.82-85...QtJE entry (intertask queue entry).

Claims (2)

[Claims] (1) In a system where each task performs processing by passing inter-task queue information, an inter-task queue entry having the above-mentioned inter-task queue information, the task ID indicating the task to which the entry belongs, and the task next to the same entry. a queue entry pool formed on the main storage device that holds an intertask queue entry further having a next entry pointer indicating the position of the intertask queue entry to be processed; and a first queue entry for each task in the queue entry pool. A queue control table that has a pointer and a final queue entry pointer and is formed on the main memory, and a checkpoint file in which the inter-task queue entries are recorded and deleted when the inter-task queue entries are created and deleted. and reads each of the inter-task queue entries recorded in the checkpoint file into the queue entry pool after a system failure, and sets the next entry pointer of each of the inter-task queue entries for each common task ID. The contents of the queue entry pool are restored by concatenating with An intertask cue control method characterized by content restoration. (2) In a system in which each task performs processing by passing inter-task queue information, an inter-task queue entry having the above-mentioned inter-task queue information, the task ID indicating the task to which the entry belongs, and the task ID within the corresponding task. A system formed on the main memory that holds an inter-task queue entry that further includes generation order information indicating the generation order of the same entry and a next entry pointer indicating the position of the inter-task queue entry to be processed next after the same entry. A queue control table having a queue entry pool, a first queue entry pointer and a last queue entry pointer for each task in the queue entry pool, and formed on the main storage device, and creation/destruction of the inter-task queue entries. In this case, a checkpoint file is provided in which the corresponding inter-task queue entries are recorded and deleted, and each of the above-mentioned inter-task queue entries recorded in the above-mentioned checkpoint file is read into the above-mentioned queue entry pool after a system failure, and the above-mentioned inter-task queue entries are recorded and deleted. By setting and concatenating the next entry pointers of queue entries in the generation order indicated by the generation order information for each common task ID, the contents of the queue entry pool are restored, and the tasks in the restored queue entry pool are An inter-task queue control method characterized in that the contents of the queue control table are restored by setting the positions of the first and last queue entries for each queue in the queue control table.