JPH04281535A - Stand-by redundant type system - Google Patents

Abstract

PURPOSE:To efficiently enable an inside state transferring processing between a main system and a stand-by system according to the change of the inside state of the main system, and to simplify the detection of the fault of the main system. CONSTITUTION:When messages stored in a received message buffer of the stand-by system reaches a prescribed amount (401), when the messages from the other main systems are not received at all within a fixed time (402), or when the messages transmitted by the corresponding main system are not received at all within the fixed time, are defined as a timing in which a check point for holding the inside states of both the main system and the stand-by system to be the same is set. Thus, the total number of the check points can be reduced, and the check point can be efficiently set. Also, when a response to a check point request is not made even after the lapse of the fixed time, the abnormality of the main system is judged, and the succeeding processing is executed by the stand-by system.

Description

[Detailed description of the invention]

[Object of the invention]

0002

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a standby redundant system in which execution units such as computers and programs are multiplexed and arranged in a main system and a standby system.

0003

2. Description of the Related Art In recent years, as requirements in various fields of computer systems have diversified and technology has become more sophisticated, systems have tended to become larger and more complex. Therefore, demand for parallel/distributed processing systems that process jobs in a distributed manner is increasing today.

Such parallel/distributed processing system methods include a parallel multiprocessing method in which the same process is executed in parallel in multiple execution units, and a parallel multiprocessing method in which the same process is executed in a single execution unit as the main system. There is a standby redundancy control method in which a standby execution unit continues processing when the main execution unit fails.

[0005] The parallel multiprocessing method does not require operations such as starting up a standby system when the main system fails, and processing can be continued without stopping the system. However, since multiple execution units process the same process in parallel, there is a drawback that the redundancy of the computer increases.

On the other hand, in the standby redundancy control system, during normal operation, the standby system is in a standby state, so the processing load on each computer is small and computer resources can be used effectively.

By the way, in this method, it is necessary to set a checkpoint where the main system reports its internal state to the standby system and keeps the internal states of the main system and the standby system the same. Usually, these checkpoints are set by the system designer consciously specifying them in the program. Another method is for the main system to periodically send intermediate progress information to the standby system. However, these methods
Intermediate progress was transferred regardless of changes in the internal state of the main system, which could not be called efficient.

[0008]

[Problems to be Solved by the Invention] The present invention has been made in response to the above-mentioned circumstances, and is capable of efficiently transferring internal states between the main system and the standby system in response to changes in the internal state of the main system. The purpose of the present invention is to provide a standby redundant system that can perform processing and set checkpoints to ensure that failures in the main system can be easily detected.

[Configuration of the invention]

0010

[Means for Solving the Problems] In order to achieve the above-mentioned object, the standby redundant system of the present invention divides the execution units in the system into a main running execution unit and a standby standby execution unit. multiplexed, the main system executes work while exchanging messages with other main systems,
In a standby redundant system in which the standby system only receives the same messages as the main system but does not perform any work, and only when the main system fails, it is switched to the main system and continues the work. The execution unit of the standby system has an internal state transmitting means for transmitting, an internal state receiving means for receiving the internal state, and a monitoring means for monitoring the state of a reception message buffer for accumulating received messages, and the execution unit of the standby system has:
requesting means for requesting the main system to transmit an internal state when the monitoring means determines that the number of messages accumulated in the received message buffer has reached a predetermined amount; an updating unit that updates its own internal state and clears the received message buffer when the internal state is received within a certain period of time, and updates its own internal state and clears the received message buffer when the internal state is received within a certain time an updating means, and when the internal state is not received within a certain period of time, it is determined that the main system has failed, switches itself to the main system, and rolls based on its own internal state and the contents of the received message buffer. The present invention is characterized in that it has a takeover means that performs back processing and takes over work.

[0011]

[Operation] In the standby redundant system of the present invention, when the number of messages accumulated in the receive message buffer of the standby system reaches a predetermined amount, the standby system sends an internal state to the main system as a checkpoint timing. request. If the internal state is received within a certain period of time, it updates its own internal state and clears the received message buffer. In addition, if the internal state is not received within a certain period of time, it is determined that the main system has failed, switches itself to the main system, and performs rollback processing based on its own internal state and the contents of the received message buffer. The work will be taken over.

[0012] Furthermore, if there are few messages exchanged between the main systems, the standby system may not be able to tell whether the main system is operating normally. Therefore, the following two methods can be considered, which combine the above method with a method of confirming that the main system is operating normally.

One method is to think that if the main system becomes abnormal, there will be no execution units to send messages to the main system. A checkpoint is also taken when no messages are received from the server.

[0014] The other method is to consider that the main system is operating normally if the main system is sending messages to other execution units.In addition to the above method, the standby system also It also receives messages sent by the main system, and takes a checkpoint when no message is received from the main system within a predetermined time.

[0015] The present invention enables efficient internal state transfer processing between the main system and standby system in response to changes in the internal state of the main system, and also facilitates failure detection in the main system. becomes possible.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, details of embodiments of the present invention will be explained based on the drawings.

FIG. 1 is a diagram showing the configuration of a standby redundant system according to an embodiment of the present invention. In this example, the execution unit is a program module. In the same figure, 1a, 1b,
1c is a processor that executes program modules. Message exchange devices 2a, 2b, 2c for controlling communication between program modules are connected to the individual processors 1a, 1b, 1c, respectively, and the message exchange devices 2a, 2b, 2c are connected in a line. Although not shown in the figure, each message exchange device 2a
, 2b, 2c have receive message buffers for storing received messages.

Furthermore, each processor 1a, 1b, 1c includes:
For example, as shown in FIG. 2, a plurality of program modules, which are execution units, are registered in a multiplexed manner. For example, A's program module includes processors 1a, 1
b and 1c (each program module is represented by A1, A2, and A3). Here processor 1a
Only program module A1 registered in
The program modules A2 and A3, which are the main system that normally operates and are registered in the processors 1b and 1c, normally only receive the same messages as the main system, but do not perform any actual processing. It is in a standby state because it becomes the main system when module A1 fails. If there are two or more standby systems as described above, any one of the standby systems becomes the main system when the main system fails.

Program module B is also registered in processors 1a and 1b (B1 and B, respectively).
(represented by 2). The program module B1 of the processor 1b is the main system, and the program module B2 of the processor 1a is the standby system. Similarly, program module C
are registered in processors 1a, 1b, and 1c (represented by C1, C2, and C3, respectively). Program module C1 of processor 1c is the main system, processors 1a and 1
Program modules C3 and C2 of b are standby systems.

FIG. 3 shows an example of the configuration of each program module. In the figure, 3a is an individual program.
3b is an internal state transmitter that transmits the internal state; 3c is an internal state receiver that receives the internal state; and 3d is a receive message buffer monitor that monitors the state of the receive message buffer. be.

Next, the operation of taking a checkpoint between the main system and the standby system in the standby redundant system having the above configuration will be explained with reference to FIGS. 4 and 5. Note that FIG. 4 shows the flow of the operation of the standby system, and FIG. 5 shows the flow of the operation of the main system.

During the normal processing period between the main systems, the standby system (for example, B2) always receives messages from other main system groups (for example, A1, C1) as well as the main system (for example, B1), and responds accordingly. is stored in the received message buffer. During this time, the reception message buffer monitoring unit 3d of the standby system B2
constantly monitors the state of its receive message buffer.

[0023] When the number of messages accumulated in the reception message buffer reaches a certain amount (step 401
), or if no message has been received from the other main systems A1 and C1 for a certain period of time (step 402), the standby system B
2 determines that it is time to take a checkpoint, and requests the main system B1 to transmit the internal state (step 403).

When the main system B1 receives the request, it activates the internal state receiving section 3c, reads out the internal state of the main system B1 (step 501), and reads the internal state receiving section 3c of the standby system B2.
(step 502).

On the other hand, after the standby system B2 issues a checkpoint request to the main system B1, the internal state receiving section 3c waits for a certain period of time for the internal state to be sent from the main system B1. When the internal state from the main system B1 is received within a certain period of time (step 404), the standby system B2 updates its own internal state and clears the received message buffer (step 405).

If the internal status cannot be received even after waiting for a certain period of time (step 404), the standby system B2 determines that the main system B1 is in failure and switches itself to the main system. Then, rollback processing is performed based on the saved internal state and the contents of the message in the received message buffer (step 406), and subsequent processing is continued.

Therefore, according to the standby redundant system of this embodiment, depending on changes in the internal state of the main system, messages received from the main system can be stored and rolled back only at the limit of the possible range. By taking checkpoints, you can take checkpoints efficiently. At the same time, it becomes possible to set checkpoints that guarantee fault detection in the main system.

Next, another embodiment of the present invention will be described.

FIG. 6 shows an example of the configuration of a program module treated as an execution unit in this embodiment. In the figure, 6a is a main processing section that performs processing specific to each program module, 6b is an internal state transmitting section that transmits an internal state, 6c is an internal state receiving section that receives an internal state,
6d is a reception message buffer monitoring section that monitors its own reception message buffer, and 6e is a main system transmission message monitoring section that monitors transmission messages from the corresponding main system.

Next, the operation of taking a checkpoint between the main system and the standby system in this embodiment system will be explained with reference to FIG. Note that FIG. 7 shows the flow of operation of the standby system, and the steps other than step 702 are the same as FIG. 4.

[0031] The standby system always receives messages sent from other main systems to the corresponding main system and stores them in the reception message buffer, and also receives messages sent by the corresponding main system. Receive. However, here, messages sent by the main system are not saved even if they are received, and only the arrival interval of the messages is checked by the main system transmission message monitoring section 6e. If no message is received from the corresponding main system within a certain period of time, the main system is requested to send its internal state. Since the subsequent steps are the same as those in FIG. 4, the explanation will be omitted.

[0032] As yet another embodiment, the above two embodiments are combined, and when the number of messages accumulated in the received message buffer reaches a predetermined amount, and at least one message is sent to the host or others within a certain period of time. It is also possible to take a checkpoint when no signal is received from the system.

Furthermore, in the above embodiments, the execution unit was described as a program module, but it is also possible to implement a similar standby redundancy control system using a computer as the execution unit as shown in FIG. In this case, multiple computers 8a1, 8a2, ..., 8b1, 8b2, ..., 8c1 as execution units
, 8c2, ..., processors 9a1 and 9a2, respectively.
,..., 9b1, 9b2,..., 9c1, 9c2,... and message exchange devices 10a1, 10a2,..., 10b1, 1
0b2,..., 10c1, 10c2... exist. Each message exchange device 10a1 to 10c2 is connected to all other message exchange devices in a communicable state. Therefore, each of the computers 8a1 to 8c2 can also communicate with any other computer. In addition, in the initial state, computers 8a1 and 8
a2,... are the main systems, and each is a computer 8b1, 8b
Assume that there are standby computers 2, . . . , 8c1, 8c2, .

[0034]

[Effects of the Invention] As explained above, according to the standby redundant system of the present invention, efficient internal state transfer processing can be performed between the main system and the standby system in response to changes in the internal state of the main system. Moreover, it becomes possible to easily detect failures in the main system.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a standby redundant system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a state in which program modules, which are execution units, are multiplexed and arranged on a plurality of processors.

FIG. 3 is a block diagram showing the configuration of a program module.

FIG. 4 is a flowchart showing the operation of the standby system in the standby redundant system of FIG. 1;

FIG. 5 is a flowchart showing the operation of the main system in the standby redundant system of FIG. 1;

FIG. 6 is a block diagram showing the configuration of a program module in another embodiment of the present invention.

FIG. 7 is a flowchart showing the operation of the standby system in the standby redundant system of the embodiment of FIG. 6;

FIG. 8 is a block diagram showing an example of a system configuration when the execution unit is replaced with a computer.

[Explanation of symbols]

1a, 1b, 1c...Processors 2a, 2b, 2c...Message exchange devices A1, A2,
A3, B1, B2, C1, C2, C3...Program module 3a...Main processing section 3b...Internal state transmitting section 3c...Internal state receiving section

Claims (3)

[Claims] Claim 1: Execution units in the system are multiplexed as a running execution unit as a main system and a standby execution unit as a standby system, and the main system performs work while exchanging messages with other main systems. On the other hand, in a standby redundant system, the standby system only receives the same message as the main system but does not perform any work, and only when the main system fails is switched over to the main system and continues the work. The unit has an internal state transmitting means for transmitting an internal state, an internal state receiving means for receiving the internal state, and a monitoring means for monitoring the state of a receiving message buffer for storing received messages, and The execution unit of the system includes a requesting means for requesting the corresponding main system to send an internal state when it is determined by the monitoring means that the number of messages accumulated in the receiving message buffer has reached a predetermined amount; updating means for updating its own internal state and clearing the received message buffer when the internal state is received within a certain time; and updating means for updating the main system when the internal state is not received within a certain time; a standby redundant type, characterized in that it has a takeover means that determines the status, switches itself to the main system, performs rollback processing based on its own internal state and the contents of the received message buffer, and takes over the work. system. [Claim 2] In the standby redundant system according to claim 1, when a standby execution unit does not receive a message within a predetermined time, the standby execution unit requests the corresponding main system to send an internal state. A standby redundant system further comprising a second requesting means. 3. In the standby redundant system according to claim 1, the execution unit of the standby system includes a message receiving means for receiving a message sent by a corresponding main system to another main system, and a message receiving means for receiving a message sent from a corresponding main system to another main system, and third requesting means for requesting the corresponding main system to transmit an internal state when the message is not received by the message receiving means;
A standby redundant system further comprising: