JPH1185713A - Multi-computer system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To effectively use the resources of processor elements or the like, to realize a degradation operation and to improve redundancy without increasing the number of the processor elements. SOLUTION: Application programs divided into plural tasks are stored in a storage medium 1 and the tasks are overlapped on the plural processor elements in a CPU 2 and are executed. The processing result of the task is- transmitted/received between the processor elements through an interprocessor element interface 3 and it is decided by majority decision. The task giving the processing result different from the result of majority decision is stopped and the task similar to the task is executed on the other processor element as an alternate task. Thus, the task is set to be the unit of redundancy management.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a multi-computer system, and more particularly to a multi-computer system having fault tolerance.

[0002]

2. Description of the Related Art Conventionally, a multi-computer system of this type is redundant with respect to a system which is expected to cause a great deal of damage to a user of the system if the system is stopped or an erroneous processing result is output. It is used to prevent a stop or an incorrect processing result by performing management.

Generally, in order to realize fault tolerance in a multi-computer system having a plurality of processor elements, procedures such as fault detection, fault isolation, and system reconfiguration are required. As a method of failure detection, a method of deciding a result by majority decision using triple hardware or a timeout indicating that a job is not completed within a predetermined time as disclosed in Japanese Patent Laid-Open No. 05-094428 is used. There is a method to detect. As a method of fault isolation, a method in which a failed processor element is replaced with a spare normal processor element by using a changeover switch, or a failed processor element is logically converted from a processor element connection network as disclosed in Japanese Patent Application Laid-Open No. 06-250992. There is a method of separating. Also, the system reconfiguration is performed without considering the load of each processor element,
For example, there is a method of making the load of each processor element as equal as possible as disclosed in Japanese Patent No. 9360. The conventional multi-computer system combines these methods so that the processor element has fault tolerance as a unit of redundancy management, and when a transient error (temporary error) as a processing result occurs or when an application program is executed. When processing is performed while being divided into a plurality of application tasks, even if an error occurs only in some of the application tasks, the processor element in which the error or the error has occurred is not used as a failed processor element. I was

[0004]

In the above-mentioned conventional multicomputer system, a processor element is used as a unit of redundancy management, and when a transient error (temporary error) occurs in the processing result of the multicomputer system, or when a partial error occurs. Even if an error occurs only in the application task, the processor element in which the error or error occurred is not used as a faulty processor element, so degraded operation that continues processing essential to system operation while degrading functions. However, there is a problem that it is difficult to realize an operation referred to as “processor operation”, and resources such as a processor element are not effectively used.
Since the unit of the redundancy management is a processor element, the number of processor elements needs to be increased when increasing the redundancy to increase the reliability of the system, so that there is a problem that the cost is increased and the system is enlarged. .

[0005] An object of the present invention is to eliminate such conventional disadvantages, so that it is not difficult to realize a degenerate operation, resources such as processor elements can be effectively used, and furthermore, redundancy is increased to improve system reliability. When you improve your sex,
An object of the present invention is to provide a multi-computer system that does not require increasing the number of processor elements.

[0006]

According to a multi-computer system of the present invention, in a multi-computer system having a plurality of processor elements, an application program is divided into a plurality of tasks, and the plurality of tasks are processed by the plurality of processor elements. At times, the task is made a unit of redundancy management so as to have fault tolerance.

Further, in the multi-computer system of the present invention, one of the plurality of tasks is processed by a plurality of processor elements in an overlapping manner.

Further, in the multi-computer system according to the present invention, a plurality of the tasks are processed by one processor element.

Further, in the multi-computer system according to the present invention, the plurality of processor elements are connected by an interface between the processor elements, and information is transmitted and received between the plurality of processor elements via the interface between the processor elements. I have.

Further, in the multi-computer system of the present invention, one of the plurality of tasks is processed in a redundant manner by a plurality of processor elements, and the plurality of processing results are received by the respective processor elements. When the plurality of processing results do not match, the processing result of the one task of the plurality of tasks is determined by majority vote.

Further, the multi-computer system according to the present invention may be arranged such that the processor element that processes the one of the plurality of tasks that has produced a processing result that does not match the processing result determined by the majority decision is provided. This task is stopped, and the task is processed by a processor element that has not processed the task.

Further, the multi-computer system according to the present invention may be arranged such that the processor element that processes the one task of the plurality of tasks that has produced a processing result that does not match the processing result determined by the majority decision is provided. The processing of the next cycle of the processing of this task is performed, and the processing result is compared with the processing results of a plurality of processor elements that have performed the processing of the next cycle of the processing of this task. The one task is stopped for the processor element that has issued a processing result that does not match the processing result determined by majority vote.

Further, the multi-computer system of the present invention connects a storage medium for storing the application program, the plurality of processor elements for processing the plurality of tasks representing the application program, and the plurality of processor elements. And an interface between the processor elements.

[0014]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing one embodiment of a multi-computer system according to the present invention.

In the embodiment shown in FIG. 1, a storage medium 1 for storing an application program, such as a magnetic disk or a semiconductor memory, a CPU 2 having a plurality of processor elements for processing a plurality of tasks representing the application program, It is constituted by an inter-processor element interface 3 for connecting a plurality of (for example, six) processor elements.

Next, the operation of the multi-computer system according to the present embodiment will be described in detail with reference to FIGS.

FIG. 2 is a diagram showing an example of the assignment of tasks to processor elements. In FIG.
When divided into 5), the five divided tasks and a control task arranged in each of a plurality of processor elements (for example, processor element # 1 to processor element # 6) and controlling each processor element are shown. I have. A plurality of tasks are arranged in each processor element. Tasks # 1 to # 3 are arranged so as to overlap processor elements # 1 to # 3, respectively.
4 and task # 5 are arranged so as to overlap processor elements # 4 to # 6, respectively. Here, each task is, for example, a periodic task that operates periodically, and notifies the processing result to the control task of the processor element in which the own task is arranged in each cycle,
The control task periodically notifies the control task on another processor element of the notified processing result via the inter-processor element interface 3. That is, these processing results are transmitted and received between the control tasks.

FIG. 3 is a flowchart showing an outline of the operation of the task. For example, it is assumed that the task is a periodic task that operates periodically, and a predetermined process is performed (S31). The control task of the arranged processor element is notified (S32), and this processing is repeated according to the instruction of the control task.

FIG. 4 is a flowchart showing an outline of the operation of the control task.

FIG. 5 is a flowchart showing the operation of fault isolation and system reconfiguration by the control task.

In FIG. 1, first, an application program to be executed is read from the storage medium 1 and divided into five predetermined tasks. It is arranged as shown, and the task and control task operate. Tasks # 1 to # 5 on each processor element continuously perform the operation shown in FIG. 3 during the system operation. As shown in FIG. 4, when the processing result is notified from each task, the control task performs a majority decision on the processing result of the same task (S41). It is checked whether or not all the processing results of each task match (S42). If they match, it is determined that no abnormality has occurred in each task, and control is continued to step S44. Step S42
As a result of the investigation, when there is a task that does not match, it is determined that an abnormality has occurred in a task that has a result different from the result of the majority decision in step S41, and fault isolation and system reconfiguration are performed as shown in FIG. (S43).

That is, each control task checks whether the task determined to be abnormal has been operating on its own processor element (S51). If the task determined to be abnormal is present on its own processor element, the task determined to be abnormal is added to the task determined to be abnormal in the next cycle as shown in FIG. Perform the processing and monitor the processing result. That is, if an abnormality occurs in the task # 1 of the processor element # 1, the control task of the processor element # 1
When the task # 1 in which the abnormality is detected is processed in the next cycle, and the processing result is compared with the task # 1 on another processor element having the normal task # 1 in the next cycle. The result of processing is compared with the result of majority decision (S5
2). It is checked whether or not the monitored processing result is abnormal (different from the result of the majority decision) (S53). If the processing result is abnormal, the task determined to be abnormal is stopped, and it is determined that the task has been stopped. The indicated task stop completion is transmitted to another control task via the inter-processor element interface 3 (S54), and the process proceeds to step S59. If the monitored processing result is not abnormal, an alternative task stop request for requesting to stop an alternative task indicating the same task as this task executed in place of the task determined to be abnormal is issued between the processor elements. The data is transmitted to another control task via the interface 3 (S55), and the process proceeds to step S59.

On the other hand, if the task determined to be abnormal in step S51 is operating on a processor other than its own processor element, it is checked whether the same task as the abnormal task is operating on its own processor element (S56). When the same task is found on the own processor element as a result of the check, the process proceeds to step S59. That is, if an abnormality occurs in the task # 1 of the processor element # 1, the task # 1 exists between the processor element # 2 and the processor element # 3. Control of the control task goes to step S59, and this processor element is excluded from the alternative task activation candidates.

If the same task as the abnormal task does not exist on the own processor element in the check in step S56, for example, the number of the processor element obtained in advance when the task is allocated to each processor element is Based on information indicating, for example, the number of tasks arranged in the processor element of this number, it is determined whether the load of the own processor element (for example, the number of arranged tasks) is the smallest and the processor element number is the smallest. Then, it is compared with another processor element in which the same task as the abnormal task is not arranged (S57). If the load is the smallest and the processor element number is the smallest, the alternative task is started on its own processor element (S58), and the process proceeds to step S59. As a result of the check in step S57, if the processor element number is not the smallest among the processor elements having the smallest load even if the load is not the smallest or the load is the smallest, the processor element is excluded from the alternative task activation candidates and the process proceeds to the step S59. . That is, if an abnormality occurs in the task # 1 of the processor element # 1, an alternative task is activated on the processor element # 4 of the processor elements # 4 to # 6.

Next, in step S59, if an abnormality occurs in the task # 1 of the processor element # 1, the control tasks of the processor elements # 2 to # 6 complete the task stop from the processor element # 1 or Waiting for the alternative task stop request, investigating whether the alternative task stop request or the task stop completion has come (S59). If the alternative task stop request has been received, the alternative task is stopped and deleted (S60), and the fault isolation and The system reconfiguration processing ends. As a result of the investigation in step S59, if the task stop has been completed, each control task checks the load status of its own processor element including the alternative task, and for example, a remarkable imbalance such as a difference in the number of tasks of 3 or more occurs. Investigate if there is any (S61), and if there is a significant imbalance,
Task relocation is performed so that the difference in the number of tasks does not become 3 or more (S62). If there is no significant imbalance, the process is terminated.

That is, as described above, the control task autonomously controls the start, stop, and deletion of the alternative task and the reconfiguration of the system.

After the fault isolation and the system reconfiguration are performed as shown in FIG. 5 as described above, as shown in FIG. 4, the processing results of the tasks match in step S44 as shown in FIG. The control task on another processor element is notified of the fact or the majority decision result and the fault isolation / system reconfiguration result via the inter-processor element interface 3 (S44). This operation is continued during the operation of the system.

[0029]

As described above, according to the multi-computer system of the present invention, an application program is divided into a plurality of tasks, and these tasks are executed on a plurality of processor elements in an overlapping manner. Is determined by majority rule, the task that has produced a processing result different from the result of majority rule is stopped, and the same task as this task is executed as an alternative task on another processor element, so that the task is a unit of redundancy management Since the unit of failure detection / isolation / reconfiguration can be reduced, even if an error occurs in a task, the processor element that executed this task does not need to be separated, so that resources such as the processor element can be used effectively and degeneration operation can be performed. Can be realized. In addition, when a task is executed in duplicate on a plurality of processor elements, the redundancy can be increased by increasing the degree of redundancy, so that it is not necessary to increase the number of processor elements when increasing the redundancy. The cost and the size of the system can be reduced. Further, since the abnormality detection is performed by a majority decision using software on a plurality of processor elements, the risk of a system down due to a single point failure can be reduced as compared with the conventional majority decision using hardware.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of a multi-computer system according to the present invention.

FIG. 2 is a diagram illustrating an example of an arrangement of tasks in processor elements.

FIG. 3 is a flowchart showing an outline of a task operation;

FIG. 4 is a flowchart showing an outline of an operation of a control task.

FIG. 5 is a flowchart showing the operation of fault isolation and system reconfiguration by a control task.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Storage medium 2 CPU 3 Interface between processor elements

Claims (8)

[Claims] 1. In a multi-computer system having a plurality of processor elements, when an application program is divided into a plurality of tasks and these plurality of tasks are processed by the plurality of processor elements,
A multi-computer system characterized in that the task is a unit of redundancy management so as to have fault tolerance. 2. The multi-computer system according to claim 1, wherein one of the plurality of tasks is processed by a plurality of processor elements in an overlapping manner. 3. The multi-computer system according to claim 1, wherein a plurality of said tasks are processed by one processor element. 4. The system according to claim 1, wherein the plurality of processor elements are connected by an interface between the processor elements, and information is transmitted and received between the plurality of processor elements via the interface between the processor elements. 4. The multi-computer system according to 2, 3 or 4. 5. One of the plurality of tasks is processed in a redundant manner by a plurality of processor elements, and the plurality of processing results are received by the respective processor elements, and the plurality of processing results are identical. 5. The multi-computer system according to claim 1, wherein a processing result of said one task of said plurality of tasks is determined by a majority decision when there is no task. 6. The processor element that has processed the one task among the plurality of tasks that has issued a processing result that does not match the processing result determined by the majority decision, suspends this task, 6. The multi-computer system according to claim 5, wherein said task is processed by a processor element which has not processed the task. 7. The processor element, which has processed the one task among the plurality of tasks that have produced a processing result that does not match the processing result determined by the majority decision, is provided next to the processing of this task. Cycle processing, comparing this processing result with the processing results of a plurality of processor elements that have executed the processing of the next cycle of the processing of this task, and when the processing results are different, the processing result determined by the majority decision 7. The multi-computer system according to claim 5, wherein the one task is stopped for the processor element that has output a processing result that does not match. 8. A storage medium for storing the application program; a plurality of processor elements for processing the plurality of tasks representing the application program; an interface between the processor elements for connecting the plurality of processor elements; Claims 1, 2, 3, 4, 5,
8. The multi-computer system according to 6 or 7.