JP2008146222A - Computer failure detection system and computer failure detection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that since the data transfer quantity of a common bus is limited in a conventional computer failure detection system, the use of the common bus for notifying the other processor of self-diagnosis information results in the deterioration of the data transfer quantity of the common bus, and since the common bus is used by the other processor, use permission is not available, and the output of self-diagnosis information is delayed, and it is not possible to correctly detect any failure. <P>SOLUTION: This computer failure detection system is provided with a plurality of self-diagnosis output computers, a failure detection computer, and a self-diagnosis information line for connecting a plurality of self-diagnosis output computers with the failure detection computer. The self-diagnosis output computer prepares self-diagnosis information by itself, and the self-diagnosis information line transmits the self-diagnosis information to be output from each self-diagnosis output computer, and the failure detection computer detects the failure of the self-diagnosis output computer which has transmitted self-diagnosis information based on the self-diagnosis information received from each self-diagnosis information line. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a system and method for detecting a failure of a computer in a multiprocessor parallel processing system without degrading the performance of a common bus connecting the computers mounted on the processor.

  In a conventional multiprocessor parallel processing system, as a method of detecting a failure occurring in a computer, each computer equipped with a processor periodically notifies self-diagnosis information to other processors via a common bus. A method of detecting a computer in which a failure has occurred by checking self-diagnosis information periodically notified from another processor is known (see, for example, Patent Document 1).

JP-A-7-219910 (pages 2-4, FIG. 3)

  However, in a multiprocessor parallel processing system, the common bus used for data communication between processors has a limited data transfer amount. Therefore, when a certain processor uses a common bus to periodically notify self-diagnosis information to other processors, the amount of data transfer that can be used for data communication between the processors decreases.

  In addition, the processor that should notify the self-diagnosis information cannot output self-diagnosis information because the use of the common bus cannot be obtained from the computer that manages the data communication on the common bus, for example, by using the common bus of another processor. Therefore, there is a possibility that output delay of self-diagnosis information occurs.

  Therefore, by notifying the self-diagnosis information using the common bus, the data transfer amount of the common bus decreases, and the reason that the self-diagnosis information is not updated in the computer that has received the self-diagnosis information is the output of the self-diagnosis information There is a problem in that it is impossible to determine whether it is not updated due to a delay or whether it is not updated because transmission is not possible due to a transmitter processor failure, and correct failure detection cannot be performed.

  The present invention has been made to solve such problems, and has an object to correctly detect a computer failure based on self-diagnosis information of each computer without reducing the data transfer amount of a common bus. .

The computer failure detection system according to the present invention is:
A plurality of self-diagnosis output computers, a failure detection computer, and a self-diagnosis information line connecting the plurality of self-diagnosis output computers and the failure detection computer,
The self-diagnosis output computer creates and outputs self-diagnosis information indicating its own self-diagnosis result,
The self-diagnosis information line is connected to the self-diagnosis output computer, and transmits self-diagnosis information output from the self-diagnosis output computer.
The fault detection computer is connected to the self-diagnosis information line, receives the self-diagnosis information transmitted from the self-diagnosis information line, and transmits the self-diagnosis information based on the received self-diagnosis information. It is characterized by detecting a failure of the output computer.

  By transmitting the self-diagnosis information using the self-diagnosis information line without using the common bus as in the present invention, the data transfer amount of the common bus is not reduced. In addition, since the self-diagnosis information line does not cause an output delay unlike the common bus, it is possible to reliably detect a failure.

Embodiment 1 FIG.
FIG. 1 shows a configuration example of a multiprocessor parallel processing system that implements the computer fault detection method according to the first embodiment.

  FIG. 1 includes two self-diagnosis output computers and one failure detection computer. However, it is not necessary to limit the number of self-diagnosis output computers in the multiprocessor parallel processing system to two as shown in FIG. 1, and it is needless to say that the parallel processing system may be configured by the number required for the parallel processing system. Absent.

  Each of the self-diagnosis output computers 11 includes a common bus I / F (interface) unit 12, a self-diagnosis output computer CPU (Central Processing Unit) 13, and a self-diagnosis information transmission unit 14. Each is internally connected by an internal bus 15.

  The failure detection computer 16 includes a common bus I / F unit 17, a failure detection computer CPU 18, a self-diagnosis information reception unit 19, and a reception memory unit 20, and each is connected via an internal bus 21. . The failure detection computer 16 includes a reception time measuring unit 22. Reception time data 23 output from the reception time measurement unit 22 is input to the self-diagnosis information reception unit 19.

  The self-diagnosis output computer 11 and the failure detection computer 16 of FIG. 1 are connected by a common bus 24 through the common bus I / F unit 12 of the self-diagnosis output computer 11 and the common bus I / F unit 17 of the failure detection computer 16. Yes. Data transfer is performed between the computers using the connected common bus 24.

  Further, the self-diagnosis information transmission unit 14 in each self-diagnosis output computer 11 and the self-diagnosis information reception unit 19 in the failure detection computer 16 are connected by a self-diagnosis information line 25 as shown in FIG. Each self-diagnosis output computer 11 transmits self-diagnosis information to the failure detection computer using the connected self-diagnosis information line 25. The self-diagnosis information line 25 may be either serial transmission or parallel transmission.

  FIG. 2 is a diagram illustrating an example of a table relating to self-diagnosis information output from the self-diagnosis output computer 11 and a data frame of self-diagnosis information in the first embodiment.

  Each self-diagnosis output computer 11 performs a self-diagnosis as to whether a function is processed in the self-diagnosis output computer 11 and a program or a device for realizing the function is operating normally or abnormally. . Here, as an example, self-diagnosis in the self-diagnosis output computer 11a will be described.

  The self-diagnosis output computer CPU 13a mounted on the self-diagnosis output computer 11a has functions to be processed inside and programs and devices for realizing the functions, such as ROM (Read Only Memory), RAM (Random Access Memory), etc. Memory, signal processing devices such as PLD (Programmable Logic Device), FPGA (Field Programmable Gate Array) and DSP (Digital Signal Processor) that perform information processing, etc. Run self-diagnostic program. In the self-diagnosis program, whether each function of the self-diagnosis output computer 11a is operating normally or abnormally is checked, and if necessary, for example, a function checking program is executed to detect abnormal operation.

  For example, if all the functions inside the self-diagnosis output computer 11a are operating normally as a result of self-diagnosis, the self-diagnosis output computer CPU 13a obtains self-diagnosis information 00 as shown in FIG. Further, for example, when a failure 2 occurs in the function 1 among the functions in the self-diagnosis output computer 11a as a result of the self-diagnosis, the self-diagnosis output computer CPU 13a similarly obtains the self-diagnosis information 02.

  In FIG. 2A, the self-diagnosis information is expressed in hexadecimal notation (hexadecimal, indicated as hex in FIG. 2A), and the number of bits is 8 bits. However, it is needless to say that the number of bits is not limited to 8 bits, and may be the number of bits that can display all the assumed failure contents.

  Based on the self-diagnosis information table of FIG. 2A, the self-diagnosis output computer CPU 13a creates a data frame of self-diagnosis information shown in FIG. 2B, for example, and outputs it to the self-diagnosis information line 25a. The constituent elements of the data frame of the self-diagnosis information to be transmitted are a self-diagnosis output computer ID (IDentifier) indicating which self-diagnosis output computer is self-diagnosis information and self-diagnosis information of the self-diagnosis output computer.

  The data frame of the self-diagnosis information may be created not by the self-diagnosis output computer CPU 13a but by the self-diagnosis information transmission unit 14a, or may be created by providing a functional block or circuit for creating a data frame.

  After creating the data frame of the self-diagnosis information, the self-diagnosis information of the self-diagnosis output computer 11a is output from the self-diagnosis information transmission unit 14a. The self-diagnosis information of the self-diagnosis output computer 11a output from the self-diagnosis information transmission unit 14a is input to the self-diagnosis information reception unit 19 of the failure detection computer 16 via the self-diagnosis information line 25a.

  The self-diagnosis output computer 11b performs the same processing. The same processing is performed when there are three or more self-diagnosis output computers 11.

  FIG. 3 is a diagram illustrating the use of the reception time data 23 for data processing in the self-diagnosis information reception unit 19 of the failure detection computer 16 in the first embodiment.

  FIG. 3 shows a case where a failure occurs in the self-diagnosis output computer 11a and the self-diagnosis output computer 11b operates normally. 3, (1) and (2) show the case where failure detection is performed without using the reception time data 23, and (3) and (4) show the case where failure detection is performed using the reception time data 23 together. Show.

  First, the case where the self-diagnosis information and the reception time data are not used together will be described with reference to (1) and (2) of FIG. As shown in FIG. 2B, the failure detection computer 16 is transmitted with the self-diagnosis output computer ID and the self-diagnosis information of the self-diagnosis output computer.

  Here, the failure of the self-diagnosis output computer 11a includes a case where the self-diagnosis output computer CPU 13a of the self-diagnosis output computer 11a has failed and a case where other than the self-diagnosis output computer CPU 13a has failed. Therefore, first, as in (1) of FIG. 3, a case where a failure other than the self-diagnosis output computer CPU 13a has failed, for example, a failure 2 in function 1 will be described.

  It is assumed that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally without a failure until the reception time t1 in the failure detection computer 16 after starting the operation. At this time, the self-diagnosis output computer 11a and the self-diagnosis output computer 11b input the self-diagnosis information “00” indicating normal operation to the self-diagnosis information receiving unit 19 of the failure detection computer 16 according to the table of FIG. . Therefore, as shown in (1) of FIG. 3, the self-diagnosis information of the self-diagnosis output computer 11a and the self-diagnosis output computer 11b received by the self-diagnosis output reception unit 19 at the reception time t1 is both “00”. is there.

  The failure detection computer CPU 18 of the failure detection computer 16 refers to the self-diagnosis information and determines that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally.

  Next, it is assumed that a failure 2 has occurred in the function 1 of the self-diagnosis output computer 11a between the reception time t1 and the reception time t2, which is the timing for receiving the next self-diagnosis information. In this case, since the self-diagnosis output computer 11b remains in normal operation, the self-diagnosis information is output as “00”. However, the self-diagnosis output computer 11a outputs “02” in accordance with the table of FIG. Accordingly, as shown in (1) of FIG. 3, the self-diagnosis information of the self-diagnosis output computer 11a received by the self-diagnosis information receiving unit 19 at the reception time t2 is “02”, and the self-diagnosis of the self-diagnosis output computer 11b. The information is “00”.

  The failure detection computer CPU 18 of the failure detection computer 16 refers to the self-diagnosis information, and the self-diagnosis output computer 11a determines that the failure 1 occurs in the function 1 and the self-diagnosis output computer 11b operates normally.

  However, with this method, when the CPU fails, failure detection cannot be performed correctly. In order to show this, a case will be described where the self-diagnosis output computer CPU 13a of the self-diagnosis output computer 11a fails as shown in (2) of FIG.

  It is assumed that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally without a failure until the reception time t1 in the failure detection computer 16 after starting the operation. At this time, the self-diagnosis output computer 11a and the self-diagnosis output computer 11b input the self-diagnosis information “00” indicating normal operation to the self-diagnosis information receiving unit 19 of the failure detection computer 16 according to the table of FIG. . Therefore, as shown in (2) of FIG. 3, the self-diagnosis information of the self-diagnosis output computer 11a and the self-diagnosis output computer 11b received by the self-diagnosis information reception unit 19 is “00” at the reception time t1. is there.

  The failure detection computer CPU 18 of the failure detection computer 16 refers to the self-diagnosis information and determines that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally.

  Next, it is assumed that a failure has occurred in the self-diagnosis output computer CPU 13a of the self-diagnosis output computer 11a between the reception time t1 and the reception time t2, which is the timing for receiving the next self-diagnosis information. In this case, since the self-diagnosis output computer 11b remains in normal operation, the self-diagnosis information is output as “00”. However, the self-diagnosis output computer 11a cannot transmit the self-diagnosis information because the failure has occurred in the self-diagnosis output computer CPU 13a, and the self-diagnosis output computer 11a cannot output the self-diagnosis information. Accordingly, as shown in (2) of FIG. 3, the self-diagnosis information of the self-diagnosis output computer 11b received by the self-diagnosis information receiving unit 19 at the reception time t2 is “00”. However, the self-diagnosis information of the self-diagnosis output computer 11a is not updated.

  The failure detection computer CPU 18 of the failure detection computer 16 refers to the self-diagnosis information and correctly determines that the self-diagnosis output computer 11b is operating normally. However, since the self-diagnosis output computer 11a does not update the self-diagnosis information, when the failure detection computer CPU 18 refers to the self-diagnosis information of the self-diagnosis output computer 11a, it is updated as shown in (2) of FIG. The self-diagnosis information “00” is not referred to, and it is erroneously determined that it is operating normally.

  Therefore, if only the self-diagnosis information transmitted from each self-diagnosis output computer 11 is used, it is possible to detect a failure when the self-diagnosis output computer CPU 13 is operating. It can be seen that the failure cannot be detected when it stops.

  Therefore, as shown in (3) and (4) of FIG. 3, the reception time data 23 output from the reception time measurement unit 22 is input to the self-diagnosis information reception unit 19 and transmitted from each self-diagnosis output computer 11. Fault detection is performed using the self-diagnosis information and the reception time data together. Note that t1 and t2 described in (3) and (4) of FIG. 3 are data representing the reception time t1 and the reception time t2.

  The reception time measurement unit 22 in the failure detection computer 16 measures time using a timer such as a counter, for example, and always outputs the time to the self-diagnosis information reception unit 19. The self-diagnosis output computers 11 and the timers mounted on the failure detection computer 16 are synchronized with each other, and the reception timing of the self-diagnosis information output from each self-diagnosis output computer 11 to the failure detection computer 16 is predetermined. If it is known, the reception time data 23 output from the reception time measuring unit 22 need not always be output, and may be output only in accordance with the reception timing.

  When receiving the self-diagnosis information from each self-diagnosis output computer 11, the self-diagnosis information receiving unit 19 adds the reception time data 23 output from the reception time measuring unit 22 to the self-diagnosis information.

  As shown in (3) of FIG. 3, when a failure 2 occurs in the function 1 of the self-diagnosis output computer 11a, the data including the self-diagnosis information and the reception time data 23 is “00t1” at the reception time t1, and the reception time t2 Then, since the data combining the self-diagnosis information and the reception time data 23 is “02t2,” the failure detection computer CPU 18 refers to and compares only the self-diagnosis information as in (1) of FIG. It is possible to correctly determine that the failure 2 has occurred in the function 1 of 11a.

  Next, a case where the self-diagnosis output computer CPU 13a of the self-diagnosis output computer 11a fails in (4) of FIG. 3 as in (2) of FIG. 3 will be described.

  It is assumed that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally without a failure until the reception time t1 in the failure detection computer 16 after starting the operation. At this time, the self-diagnosis output computer 11a and the self-diagnosis output computer 11b input the self-diagnosis information “00” indicating normal operation to the self-diagnosis information receiving unit 19 of the failure detection computer 16 according to the table of FIG. . Therefore, as shown in (4) of FIG. 3, the data obtained by combining the self-diagnosis information of the self-diagnosis output computer 11a and the self-diagnosis output computer 11b and the reception time data 23 at the reception time t1 is “00t1”.

  The failure detection computer CPU 18 of the failure detection computer 16 refers to the self-diagnosis information, and although not shown in FIG. 3, compares the data obtained by combining the self-diagnosis information one hour before and the reception time data 23, It is determined that the self-diagnosis output computer 11a and the self-diagnosis output computer 11b are operating normally.

  It is assumed that a failure has occurred in the self-diagnosis output computer CPU 13a of the self-diagnosis output computer 11a between the reception time t1 and the reception time t2, which is the timing for receiving the next self-diagnosis information. In this case, since the self-diagnosis output computer 11b remains in normal operation, the self-diagnosis information is output as “00”. However, since the self-diagnosis output computer 11a has failed in the self-diagnosis output computer CPU 13a, the self-diagnosis output computer 11a cannot output self-diagnosis information.

  Therefore, the combined data of the self-diagnosis information of the self-diagnosis output computer 11b and the reception time data 23 at the reception time t2 is “00t2.” However, since the self-diagnosis information of the self-diagnosis output computer 11a is not received by the self-diagnosis information receiving unit 19, the combined data of the self-diagnosis information and the reception time data 23 remains “00t1” without being updated.

  As will be described later, the failure detection computer CPU 18 of the failure detection computer 16 compares the data of the reception time t2 with the data of the reception time t1 one time earlier with respect to the data obtained by combining the self-diagnosis information and the reception time data. That is, the self-diagnosis output computer 11b correctly determines that the operation is normal by comparing the data “00t1” at the reception time t1 with the data “00t2” at the reception time t2.

  On the other hand, since the combined data of the self-diagnosis information and the reception time data is not updated, the self-diagnosis output computer 11a compares the data “00t1” at the reception time t1 with the data “00t1” at the reception time t2, and receives the data. Detect that time data is not updated.

  Since the self-diagnosis information line 25 connects the failure detection computer 16 and each self-diagnosis output computer 11, the self-diagnosis output computer 11 and the failure detection computer 16 are connected to each other as in the case where the common bus 24 is used. There is no delay in output of self-diagnosis information due to data transmission / reception. Therefore, a data update delay due to an output delay does not occur.

  Therefore, the reason why the data combining the self-diagnosis information of the self-diagnosis output computer 11a and the reception time data 23 is not updated at the reception times t1 and t2 is not due to the output delay of the self-diagnosis information. It can be determined that the operation has been stopped due to the occurrence of a failure in the self-diagnosis output computer CPU 13a of the computer 11a. Therefore, the failure detection computer CPU 18 can correctly detect the failure of the self-diagnosis output computer 11a.

  FIG. 4 is a diagram illustrating an example of data output from the self-diagnosis information reception unit 19 to the reception memory unit 20 in the first embodiment.

  FIG. 4A is an example of data relating to self-diagnosis information and reception time of the self-diagnosis output computer. For example, no. 1 represents that the self-diagnosis information “00” (normal) of the self-diagnosis output computer a is received at 00:01:00 (t1). No. 2 indicates that self-diagnosis information “02” (fault 2 occurred in function 1) of the self-diagnosis output computer a was received at 00:01:30 (t2).

  FIG. 4B shows an example of a data format to be output to the reception memory unit 20 when the self-diagnosis information receiving unit 19 of the failure detection computer 16 receives the self-diagnosis information based on the data in FIG. . For example, as shown in FIG. 2B, the self-diagnosis information receiving unit 19 includes a self-diagnosis output computer ID and self-diagnosis information transmitted from the self-diagnosis output computer, and a reception time output from the reception time measuring unit 22. Data 23 is input. From these inputs, data is created as in the self-diagnosis information receiving unit data format example shown in (1) of FIG.

  Next, for temporary storage in the reception memory unit 20, the reception memory address and reception memory data are divided. As shown in (2) of FIG. 4B, the reception memory address is composed of a self-diagnosis output computer ID unit and a TIME unit that is a bit representing the order of the reception time.

  As described in FIG. 3, the TIME unit compares the self-diagnosis information at the reception time t2 with the self-diagnosis information at the reception time t1 one time before, and performs fault detection. It is a bit for distinguishing the self-diagnosis information of the diagnostic output computer. For convenience of explanation, an address based on the reception time t2 is called a NEW address, and an address based on the reception time t1 one time before is called an OLD address.

  As shown in (3) of FIG. 4B, the reception memory data is composed of self-diagnosis information and reception time. Thus, for example, “00t1” is received when the self-diagnosis information operating normally at the reception time t1 is received, and “02t2” is received when the self-diagnosis information in which the failure 2 occurs in the function 1 at the reception time t2. It becomes. For convenience of explanation, in the received memory data, data corresponding to the NEW address is referred to as NEW data, and data corresponding to the OLD address is referred to as OLD data. The reception memory unit 20 stores the NEW data output from the self-diagnosis information reception unit 19.

  FIG. 5 is a diagram showing failure detection by comparing data with respect to the self-diagnosis output computer a stored in the reception memory unit 20. FIG. 6 is an example of a flowchart showing that failure detection is performed using data of the self-diagnosis output computer a temporarily stored in the reception memory unit 20. The processing contents will be described with reference to FIGS. The description is given only for the self-diagnosis output computer a, but the same applies to the self-diagnosis output computer b.

  First, failure detection is started when the multiprocessor parallel processing system is activated (step ST101). At this time, the data in the reception memory unit 20 is set to an appropriate initial value such as all bits 0 or all bits 1.

  Next, as shown in FIG. 5, the NEW data at the reception time one time before in the reception memory is overwritten as OLD address data (OLD data) and temporarily stored (step ST102 in FIG. 6).

  Next, the data obtained by combining the received self-diagnosis information of the self-diagnosis output computer a and data indicating the reception time is overwritten and updated in the NEW data of the reception memory unit 20 (step ST103 in FIG. 6).

  Next, the temporarily stored NEW data and the OLD data which is the NEW data one hour before are read from the reception memory unit 20 (step ST104 in FIG. 6).

  Next, the OLD data read from the reception memory unit 20 and the NEW data are compared (step ST105 in FIG. 6).

  Here, the case where the self-diagnosis output computer a is operating normally as shown in (1) of FIG. 5 will be described. When the self-diagnosis output computer a is operating normally, the OLD data and the NEW data (for example, “00t1” and “00t2”) are different as shown in (1) of FIG. 5, and therefore step ST105 of FIG. 6 is NO. .

  In that case, it is determined whether the NEW data indicates an abnormality (step ST106 in FIG. 6). As shown in FIG. 4, the presence / absence of an abnormality can be found from self-diagnosis information among elements constituting NEW data.

  When the self-diagnosis output computer a is operating normally, the NEW data indicates normality (for example, “00” of “00t2”), so step ST106 in FIG. 6 is NO. In that case, the process returns to step ST102 of FIG. Thereafter, the same processing loop is repeated.

  Next, as shown in (2) of FIG. 5, a case where a failure 2 occurs in the function 1 in the self-diagnosis output computer a will be described. In this case, for example, since the OLD data is “00t1” and the NEW data is “02t2”, the process up to the point where NO is selected in step ST105 of FIG. 6 is as described above.

  (2) in FIG. 5 shows a case where the failure 2 occurs in the function 1 between the reception times t1 and t2. Accordingly, since the NEW data at the reception time t2 includes information (for example, “02” of “02t2”) in which the failure 1 occurred in the function 1, step ST106 in FIG. 6 is YES.

  From the self-diagnosis information composing the NEW data, which function is causing what kind of abnormality is read (step ST107 in FIG. 6), and the fact that an abnormality has occurred is displayed on the self-diagnosis output computer a (step ST108 in FIG. 6). . Further, for the sake of accuracy, the malfunction function of the self-diagnosis output computer a and the contents of the malfunction may be displayed.

  When the occurrence of abnormality in the self-diagnosis output computer a is displayed, the failure detection process is terminated by stopping all the computers, for example (step ST109 in FIG. 6).

  Finally, as shown in (3) of FIG. 5, a case where the CPU fails in the self-diagnosis output computer a between the reception time t1 and the reception time t2 will be described. In that case, since the data at the reception time t2 is not updated as shown in (3) of FIG. 5, the OLD data and the NEW data match (for example, “00t1” and “00t1”), and step ST105 of FIG. Select YES.

  In this case, it is determined that the self-diagnosis information has not been updated because the CPU of the self-diagnosis output computer a is out of order or the like (step ST110 in FIG. 6).

  If it is determined that the CPU of the self-diagnosis output computer a has failed, a display indicating that an abnormality has occurred in the self-diagnosis output computer a is displayed (step ST108 in FIG. 6). For further accuracy, the CPU failure of the self-diagnosis output computer a may be displayed. When the occurrence of abnormality in the self-diagnosis output computer a is displayed, the failure detection process is terminated by stopping all the computers, for example (step ST109 in FIG. 6).

  In this way, it is possible to detect a failure of the self-diagnosis output computer by performing a reference comparison using the self-diagnosis information and the reception time together.

  After detecting a failure, it is possible to replace the failed computer by stopping the processing of all computers, or to remove the failed computer from the multiprocessor parallel processing system and continue the processing using the remaining computers. It is.

  As described above, by transmitting the self-diagnosis information using the self-diagnosis information line 25 without using the common bus 24, the data transfer amount of the common bus 24 is not reduced. Further, since the self-diagnosis information line 25 does not generate an output delay unlike the common bus 24, it is possible to reliably detect a failure.

  Further, a reception time measuring unit 22 is provided in the failure detection computer 16 to add the reception time data 23 to the received self-diagnosis information, and the failure detection computer CPU 18 performs reference comparison so that only the function failure of the self-diagnosis output computer is detected. In other words, it is possible to detect a failure when the self-diagnosis information is not updated due to the occurrence of a processor failure.

1 is a diagram illustrating a configuration example of a multiprocessor parallel processing system that implements a computer fault detection method according to Embodiment 1. FIG. It is the figure which showed an example of the table regarding the self-diagnosis information output from the self-diagnosis output computer in Embodiment 1, and the data frame of self-diagnosis information. 6 is a diagram for explaining the use of reception time data for data processing in the self-diagnosis information receiving unit of the failure detection computer according to Embodiment 1. FIG. 6 is a diagram illustrating an example of data output from a self-diagnosis information reception unit to a reception memory unit in the first embodiment. FIG. 6 is a diagram illustrating failure detection by comparing data with respect to a self-diagnosis output computer stored in a reception memory according to Embodiment 1. FIG. 3 is an example of a flowchart showing that failure detection is performed using data of a self-diagnosis output computer temporarily stored in a reception memory unit in the first embodiment.

Explanation of symbols

11. Self-diagnosis output calculator 12. Common bus I / F section 13. Self-diagnosis output computer CPU
14 Self-diagnosis information transmission unit 15. Internal bus 16. Failure detection computer 17. Common bus I / F section 18. Failure detection computer CPU
19. Self-diagnosis information receiving unit 20. Reception memory unit 21. Internal bus 22. Reception time measuring unit 23. Receive time data 24. Common bus 25. Self-diagnosis information line

Claims (4)

A plurality of self-diagnosis output computers, a failure detection computer, and a self-diagnosis information line connecting the plurality of self-diagnosis output computers and the failure detection computer,
The self-diagnosis output computer creates and outputs self-diagnosis information indicating its own self-diagnosis result,
The self-diagnosis information line is connected to the self-diagnosis output computer, and transmits self-diagnosis information output from the self-diagnosis output computer.
The fault detection computer is connected to the self-diagnosis information line, receives the self-diagnosis information transmitted from the self-diagnosis information line, and transmits the self-diagnosis information based on the received self-diagnosis information. A computer fault detection system characterized by detecting a fault in an output computer. The failure detection computer includes a memory unit and a CPU,
When the memory unit receives the self-diagnosis information output from the self-diagnosis output computer, the memory unit stores data in which the reception time of the self-diagnosis information and the self-diagnosis information are associated with each other.
2. The computer failure detection system according to claim 1, wherein the CPU detects a failure of a self-diagnosis output computer that has transmitted the self-diagnosis information with reference to the data stored in the memory unit. Creating and outputting self-diagnosis information in each computer;
Detecting a failure of the computer that transmitted the self-diagnosis information by referring to and comparing data in which the self-diagnosis information is received and the self-diagnosis information based on the output self-diagnosis information; and
A computer fault detection method comprising:   The reference comparison data includes data that associates the reception time of the self-diagnosis information with the self-diagnosis information, the reception time when the self-diagnosis information is received one hour before the reception time, and the self before the one time. 4. The computer fault detection method according to claim 3, wherein the data is associated with diagnostic information.

Images