JPH0895931A - Faust detecting method for distributed computer system - Google Patents

Abstract

PURPOSE: To provide a fault detecting method with which fault resistance can be improved and the load on a network and a computer at normal time can be minimized by investigating whether each computer periodically receives a survival signal transmitted from the adjacent computer on a virtual ring or not, and judging the generation of any abnormality on a communication path to be used for transmitting the survival signal when no signal is received. CONSTITUTION: Plural computers 101-104 or the like are arranged on a virtual ring 10 and each of computers 101-104 or the like periodically transmit the survival signal showing its own survival to the adjacent computer in the specified direction on the virtual ring 10. Besides, each of computers 101-104 or the like investigates whether the survival signal transmitted from the adjacent computer on the virtual ring 10 is periodically received or not, when no signal is received, it is judged that any abnormality is generated on the communication path to be used for transmitting the survival signal, and the position of that fault is specified. Then, each of computers 101-104 or the like reports fault information on the discovered fault to all communicatable communication equipment.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a distributed computer system including a plurality of computers connected via a network, for reducing the load on the network and the computer to reliably detect a failure. The present invention relates to a failure detection method.

[0002]

2. Description of the Related Art FIG. 59 is a block diagram for explaining a failure detecting method for a computer and a network in a conventional distributed computer system.
Is a calculator. In the distributed computer system, these computers 101 to 105 are connected to each other via a local area network (hereinafter referred to as LAN) so that they can communicate with each other, as is well known.

Next, the operation will be described. One of these computers, for example, the computer 103, is assigned the role of managing operating information, and this computer 103 periodically sends a signal for failure detection, that is, a survival signal to all other computers. A signal for request is transmitted, and it is checked whether a response to this signal, that is, a survival signal is received within a certain time. When receiving the survival signal, the operation information management computer 103 determines that the computer that has transmitted the survival signal and the communication path with the computer are normal,
When the survivor signal is not received, it is determined that the computer that has transmitted the survivor signal transmission request signal or the communication path with the computer is in some failure state. Then, the operation information management computer 103 notifies all the computers of the detected failure information by some method.

By the way, in such a failure detection method,
Since the management of operation information is concentrated on one computer, the failure detection function is lost when the computer fails. To avoid such drawbacks, there is also a method of assigning similar functions to all computers.

[0005]

Since the conventional fault detection method for a distributed computer system is configured as described above, the fault detection function is concentrated on a specific computer especially in the former method. There is a problem that the failure detection function is lost when the device itself fails.

In the latter fault detection method, when the number of computers increases, the amount of signals transmitted / received for fault detection increases in proportion to the square of the number of computers, so that LAN is involved. There is a problem that the burden (the number of signals transmitted on the LAN per unit time) becomes large. That is, regarding the number of surviving signals sent out on the LAN, assuming that the number of computers is N, all computers send signals for requesting survival signal transmission other than themselves,
Since the survival signal for this is received, 2N
(N-1) signals will be transmitted on the LAN. Further, since the number of signals transmitted / received for each computer is also proportional to the number N of computers, the load on the computers becomes large.

Further, in the case of a distributed computer system connected to two or more LANs, the computer itself and some LANs
When the connection with the computer is disconnected, there is a problem that it is not possible to know the state of connection to another computer and its LAN. As a result, as shown in FIG. 60, when two computers, that is, the computers 102 and 103, are disconnected from different LANs, or when a trailing failure occurs, the computers 102 and 103 cannot communicate with each other. It is determined that At this time, since the computer 102 is disconnected from the LAN 401, the computer 103 may be out of order, or the connection between the computer 103 and the LAN 402 may be simply disconnected (that is, the computer 103 is normal). It cannot be determined whether communication is possible via the LAN 401.
Similarly, since the computer 103 is disconnected from the LAN 402, is it possible that the computer 102 is out of order?
Or, the connection between the computer 102 and the LAN 401 is only disconnected (that is, the computer 102 is normal), and the LAN 40
It is not possible to determine whether or not communication is possible via 2. Therefore, the computers 102 and 103 are
Even though there is a route for communication via 1, the problem is that the route cannot be found.

The invention of claim 1 has been made to solve the above-mentioned problems. It is excellent in fault tolerance without concentrating the fault detection function on a specific computer, and the network and It is an object of the present invention to obtain a fault detection method for a distributed computer system that can minimize the load on the computer.

According to a second aspect of the present invention, in addition to the first aspect of the invention, by periodically switching the transmission destination of the survival signal from the right adjacent to the left adjacent or vice versa, the communication amount can be minimized even when a failure occurs. It is an object of the present invention to provide a failure detection method for a distributed computer system that can detect a failure and minimize the number of surviving signals exchanged in normal times and when an abnormality occurs.

According to the third aspect of the present invention, by writing in the survival signal to be transmitted next whether or not the receiving computer itself has received the survival signal, whether or not the destination computer has received the survival signal and the transmission of the survival signal. The purpose of the present invention is to obtain a failure detection method for a distributed computer system that can minimize the number of surviving signals exchanged during normal times and during an abnormality by discovering a failure by combining whether or not the original computer has received a surviving signal. And

According to the invention of claim 4, by arranging each computer on a virtual tree having nodes as nodes, the failure can be found with a minimum communication amount even when the failure occurs, and the survivor is exchanged in normal times and when an abnormality occurs. It is an object of the present invention to obtain a fault detection method for a distributed computer system that can minimize the number of signals.

According to the invention of claim 5, by dividing the computer into a plurality of groups and arranging the representative computers in each group on the virtual ring, it is possible to find the fault with the minimum communication amount even when the fault occurs. An object of the present invention is to obtain a fault detection method for a distributed computer system that can minimize the number of surviving signals exchanged in normal times and in the event of an abnormality.

According to a sixth aspect of the present invention, by using a dual LAN, it is possible to check whether or not a surviving signal is received by a destination computer, and thereby, the number of surviving signals exchanged for detecting a failure in normal times. The objective is to obtain a fault detection method for a distributed computer system that can minimize

According to the seventh to tenth aspects of the invention, by using the duplicated LAN, the failure detection is performed by combining the receiving state of the live signal itself and the receiving state of the live signal at the transmission source computer. This minimizes the number of surviving signals that are exchanged during normal times and when an abnormality occurs, and it is possible to find a failure of one computer by multiple computers in the vicinity of the failure computer, resulting in a shorter delay time from the occurrence of the failure. The purpose of the present invention is to obtain a fault detection method for a distributed computer system that can detect faults in.

According to the eleventh to fourteenth aspects of the present invention, when a failure occurs, a failed computer is restored, or a new computer is added, the transmission destination of each computer is changed, and even if the system configuration changes, that It is an object of the present invention to obtain a fault detection method for a distributed computer system that can maintain the same fault detection capability as before.

In the inventions of claims 15 and 16, since the survival signal transmitted by each computer is used for the notification of the failure information, it is not necessary to transmit an extra signal for notification, and L
An object of the present invention is to obtain a failure detection method for a distributed computer system that can reduce the load on the AN.

In the inventions of claims 17 and 19, three or more LANs are grouped into two groups, and each group is defined by claim 6.
To claim 10, claim 12, claim 14, claim 16
It is an object of the present invention to obtain a failure detection method for a distributed computer system having an arbitrary number of LANs by applying the method of (1).

In the invention of claim 18, three or more LANs are grouped into two groups, and the methods of claims 1 to 5, claim 11, claim 13 and claim 15, for each group, Alternatively, claim 6 to claim 10, claim 12, claim 1
The object of the present invention is to obtain a failure detection method for a distributed computer system having an arbitrary number of LANs by applying the method of claim 4 or claim 16.

According to a twentieth aspect of the invention, in the failure detection method of the nineteenth aspect, L shared by two LAN groups is used.
In AN, the survival signal used in each set is 1
The purpose is to obtain a fault detection method for a distributed computer system that reduces the number of surviving signals to be exchanged.

It is an object of the present invention to provide a method for detecting a failure in a distributed computer system, which can reduce the effect of a failure occurrence on the original work.

It is an object of the inventions of claims 22 and 23 to obtain a failure detecting method for a distributed computer system capable of surely detecting a failure of a computer for which failure detection is highly necessary.

According to a twenty-fourth aspect of the present invention, it is possible to obtain a fault detection method for a distributed computer system in which the relationship between the transmission and reception times of the survival signal can be freely set according to the required characteristic of fault detection. To aim.

[0023]

According to a first aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system comprising: a virtual placement step of placing a plurality of computers on a virtual virtual ring; Check whether or not the survival signal transmitted from the adjacent computer has been regularly received. If not, it is determined that an abnormality has occurred in the communication path used to transmit the survival signal, and the failure location is specified. The failure detection step and the failure notification step in which each computer notifies failure information regarding the found failure to all computers with which it can communicate are executed.

According to a second aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein in a survival signal transmitting step in which a computer periodically transmits a survival signal, the computer alternately switches on a virtual ring at regular timings. It sends a survival signal to the computer next to or to the left.

According to a third aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein in the surviving signal transmitting step, it is determined whether or not the computer has received the surviving signal to be received within a predetermined time as a surviving signal to be transmitted. It is for writing and transmitting.

According to a fourth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein each computer is a node, and each node is arranged on a virtual virtual tree having two or more child nodes. The survival signal transmission step in which the computer periodically sends a survival signal to the parent computer located at the parent node on the virtual tree, and each computer is alive from the child computer located at the child node on the virtual tree. A failure detection step of checking whether or not a signal has been received, specifying a failure location by combining the results, and a failure notification step of notifying all computers with which each computer can communicate information regarding the discovered failure of each computer. It is what you do.

According to a fifth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein a computer is divided into M groups,
One computer for each group is set as a representative computer, and a virtual arrangement step of arranging M representative computers on a virtual virtual ring, and computers other than the representative computer survive on the representative computer of the group to which the computer belongs. The first survival signal transmission step of periodically transmitting a signal and each representative computer,
The second survival signal transmitting step of periodically transmitting the survival signal to the adjacent computer in the specific direction on the virtual ring, and checking whether or not each representative computer has received the survival signal transmitted to the representative computer The failure detection step of specifying the failure location by combining the results, and the failure notification step of notifying all the computers with which each representative computer can communicate the information about the found failure.

According to a sixth aspect of the present invention, there is provided a method of detecting a failure in a distributed computer system, which comprises a virtual placement step of placing a plurality of computers on a virtual virtual ring, and identifying each computer from a specific computer on the virtual ring. When dividing into an even number and an odd number according to the order in the direction of, the odd number computer periodically sends a survival signal to the adjacent computer on the virtual ring via the first LAN, and the even number computer , A survival signal transmitting step of periodically transmitting a survival signal to an adjacent computer on the virtual ring via the second LAN, and checking whether or not each computer has received the survival signal transmitted to the computer. , A failure detection step of specifying the failure location by combining the results, and a failure notification step of notifying all the computers with which each computer can communicate information on the discovered failure, It is intended to run.

According to a seventh aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, comprising a virtual placement step of placing a plurality of computers on a virtual virtual ring, and each computer placing a virtual ring via a first LAN. Alive signal is periodically transmitted to the computer adjacent to the above specific direction, and the second L
Through the AN, a survival signal transmitting step of periodically transmitting a survival signal to a computer adjacent to the adjacent computer in a direction opposite to the specific direction on the virtual ring, and each computer receiving the survival signal transmitted from the adjacent computer Check whether or not, and write the result to the survival signal to be transmitted to the adjacent computer as a response to the survival signal sent from the adjacent computer, and the survival signal response step from each computer on both sides on the virtual ring. By combining the presence / absence of the survival signal and the content of the response, a failure detection step of specifying the failure location and a failure notification step of notifying all computers with which the computer can communicate information regarding the discovered failure To do.

According to an eighth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein each computer transmits a survival signal periodically to an adjacent computer in a survival signal transmitting step. A copy of the response from another adjacent computer different from the adjacent computer is also written together with the response to.

According to a ninth aspect of the present invention, there is provided a method of detecting a failure in a distributed computer system, wherein a virtual placement step of placing a plurality of computers on a virtual virtual ring and each computer from a specific computer on the virtual ring. When dividing into even-numbered and odd-numbered computers depending on the order in a specific direction, the odd-numbered computers periodically send survival signals to the computers on both sides of the virtual ring via the first LAN, and the even-numbered computers However, the survival signal transmission step of periodically transmitting the survival signal to the computers on both sides of the virtual ring via the second LAN, and each computer is transmitted from both sides via the first or second LAN. Check whether or not a survivor signal has been received, and combine the results to identify the failure location, and each computer can communicate the information about the discovered failure. And it executes the failure notification step of notifying the computer.

According to a tenth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, which comprises a virtual placement step of placing a plurality of computers on a virtual virtual ring, and dividing each computer into a plurality of groups of three computers. In each group,
A survival signal in which the first computer periodically transmits a survival signal to the second computer via the first LAN and at the same time transmits a survival signal to the third computer via the second LAN. In the transmitting step and in each group, it is checked whether the second computer has received the survival signal from the first computer, and the result is periodically sent to the third computer via the second LAN. The first survival signal response step of writing in the survival signal to be transmitted, and whether the third computer has received the survival signal from the first computer, and the result is stored in the first computer to the first computer. The second survival signal response step of writing to the survival signal periodically transmitted via the LAN, the second computer checks the existence and contents of the survival signal transmitted from the first and third computers, and By combining the results, the failure location The first failure detection step for specifying, and the third computer for checking the existence and contents of the survival signal transmitted from the first and second computers, and for combining the results thereof, for specifying the failure location. The failure detection step of No. 2 and the failure notification step of notifying all the computers with which each computer can communicate of the information about the discovered failure.

According to an eleventh aspect of the present invention, there is provided a distributed computer system failure detection method, which further executes a reallocation step for newly setting a virtual layout of each computer when a failure occurs.

According to a twelfth aspect of the present invention, there is provided a distributed computer system failure detection method, further comprising a relocation step for re-setting a virtual layout of each computer when a failure occurs.

According to a thirteenth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, further comprising a relocation step for resetting a virtual layout of each computer when a failure occurs.

According to a fourteenth aspect of the present invention, there is provided a distributed computer system failure detection method further comprising a relocation step for re-setting a virtual layout of each computer when a failure occurs.

According to a fifteenth aspect of the present invention, in a fault detecting method for a distributed computer system, in the fault notifying step of notifying the adjacent computer of the detected fault information, the fault signal is added to the surviving signal and the surviving signal is transmitted. The failure is notified by.

According to a sixteenth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein in the failure notification step of notifying the adjacent computer of the detected failure information, the failure information is added to the survival signal and the survival signal is transmitted. The failure is notified by.

According to a seventeenth aspect of the present invention, there is provided a distributed computer system failure detection method, comprising: a distributed system consisting of a plurality of computers connected by 2N LANs;
Two pairs of N are used and any one of the failure detection methods of claims 6 to 10, claim 12, claim 14 and claim 16 is used for each pair.

According to an eighteenth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein in a distributed system comprising a plurality of computers connected by (2N + 1) LANs, two LANs are paired, and each pair is paired. Claim 6
Any one of the failure detection methods of claim 10, claim 12, claim 14, and claim 16
The book uses any one of the failure detection methods of claims 1 to 5, claim 11, claim 13, and claim 15.

According to a nineteenth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein in a distributed system comprising a plurality of computers connected by (2N + 1) LANs, two LANs are paired, and (2N + 1) LANs are paired. The actual L
Claims 6 to 10 and claim 1 for each pair by further forming one pair with AN and any LAN
Any one of the failure detection methods of claim 2, claim 14 and claim 16 is used.

According to a twentieth aspect of the present invention, there is provided a distributed computer system fault detection method, wherein an LA shared by two pairs is used.
In N, the survival signals transmitted in each pair are combined.

In the failure detecting method for a distributed computer system according to the twenty-first aspect of the present invention, in the virtual arranging step, computers that frequently communicate with each other are arranged so as to be close to each other in the virtual arrangement.

According to a twenty-second aspect of the present invention, there is provided a distributed computer system failure detection method in which, in the virtual placement step, highly reliable computers and low reliability computers are alternately arranged in a virtual placement.

According to a twenty-third aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein in a virtual arrangement step, highly reliable computers and functionally important computers are arranged alternately in a virtual arrangement.

According to a twenty-fourth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein the transmission time or reception deadline of some or all of the surviving signals is based on the time at which each computer receives a specific surviving signal. Is set.

[0047]

According to the first aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein a plurality of computers are arranged on a virtual virtual ring, and each computer is connected to a computer adjacent in a specific direction on the virtual ring. , Periodically send a survival signal indicating its own survival. In addition, each computer periodically checks whether or not the survival signal transmitted from the adjacent computer on the virtual ring is received. If not, an abnormality has occurred in the communication path used to transmit the survival signal. Then, each computer notifies each of the computers with which it can communicate the fault information regarding the discovered fault. In other words, each computer periodically sends a signal to the adjacent destination on the virtual ring and sends a live signal periodically to check whether the live computer can receive the live signal. To do. The survivor signal is transmitted to a computer determined for each computer, and the destination computer checks whether the survivor signal can be received, whereby each computer detects a fault in a limited range. Since each computer sends a survival signal only to a designated partner, the amount of signals sent and received by all computers is proportional to the number of computers, and the amount of alive signals sent and received per computer is the number of computers. It is almost constant regardless of. This minimizes the number of surviving signals that are exchanged during normal times for fault detection. In addition, each computer notifies other computers of the information of the fault found within its own range, so that even if a strike failure occurs, each computer can obtain operating information of the entire system. You can

In the failure detection method for the distributed computer system according to the second aspect of the present invention, in the survival signal transmitting step in which the computer periodically transmits the survival signal, the computer is alternately switched on the virtual ring at every regular timing to the right of the neighbor. Or send a survival signal to the computer on the left. Thus, in addition to the periodical transmission of the survival signal, by devising the combination of the destination computers, a computer that receives the survival signals from two or more computers is created. Therefore, even when a failure occurs, it is possible to detect the failure with a minimum amount of communication, and it is possible to minimize the number of surviving signals that are exchanged in normal times and when an abnormality occurs.

According to a third aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein in the live signal transmitting step,
Whether or not the computer has received the survival signal to be received within a predetermined time is written in the survival signal to be transmitted and transmitted. That is, in addition to regular transmission of the survival signal and devising a combination with the destination computer, by writing in the survival signal to be transmitted next whether or not oneself has received the survival signal,
Whether the destination computer received the survival signal itself,
The source computer of the surviving signal detects whether or not the surviving signal is received in combination with the surviving signal. As a result, the number of surviving signals exchanged in normal times and when an abnormality occurs can be minimized.

According to a fourth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein each computer is arranged as a node and each node is arranged on a virtual virtual tree having two or more child nodes.
Each computer has a survival signal transmission step of periodically transmitting a survival signal to the parent computer located at the parent node on the virtual tree, and each computer sends a survival signal from the child computer located at the child node on the virtual tree. It is checked whether or not the survival signal is received, and the results are combined to identify the failure location. Further, each computer notifies all computers with which it can communicate of information regarding the discovered failure. That is, in addition to the periodical transmission of the survival signal according to claim 1, by devising a combination of transmission destination computers, a computer for receiving the survival signals from two or more computers is created. Therefore, even when a failure occurs, it is possible to detect the failure with a minimum amount of communication, and it is possible to minimize the number of surviving signals that are exchanged in normal times and when an abnormality occurs.

According to a fifth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein a computer is divided into M groups and one computer is set as a representative computer in each group.
M representative computers are arranged on a virtual virtual ring. Computers other than the representative computer periodically transmit the survival signal to the representative computers of the group to which the computer belongs, and each representative computer periodically transmits the survival signal to the adjacent computer in a specific direction on the virtual ring. . Further, each representative computer checks whether or not the survival signal transmitted to the computer has been received, and combines the results to identify the failure location, and each representative computer can communicate information regarding the found failure. Notify all computers. That is, in addition to the periodical transmission of the survival signal according to claim 1, by devising a combination of transmission destination computers, a computer for receiving the survival signals from two or more computers is created. Therefore, even when a failure occurs, it is possible to detect the failure with a minimum amount of communication, and it is possible to minimize the number of surviving signals that are exchanged in normal times and when an abnormality occurs.

In the fault detecting method for the distributed computer system according to the sixth aspect of the present invention, a plurality of computers are arranged on a virtual virtual ring. Depending on the order in a specific direction from a specific computer on the virtual ring, each computer is an even number,
Divided into odd-numbered computers, the odd-numbered computers periodically send survival signals to adjacent computers on the virtual ring via the first LAN, and the even-numbered computers send virtual signals to the virtual ring via the second LAN. The survival signal is periodically transmitted to the adjacent computer above. Further, each computer checks whether or not the survival signal transmitted to the computer is received, and combines the results to identify the failure location, and each computer communicates the information regarding the found failure to all the information that can be communicated. Notify the calculator. That is,
By applying a method in which each computer periodically sends a signal to a designated transmission partner and sends a live signal periodically, it is checked whether or not the live signal is received by the destination computer. To do. This minimizes the number of surviving signals that are exchanged during normal times for fault detection.

In the fault detecting method for the distributed computer system according to the seventh aspect of the present invention, a plurality of computers are arranged on a virtual virtual ring. Each computer periodically transmits a survival signal to a computer adjacent to a specific direction on the virtual ring via the first LAN, and at the same time, transmits a survival signal to the specific computer on the virtual ring via the second LAN. The survival signal is periodically transmitted to the computers adjacent in the opposite direction. In addition, each computer checks whether or not it has received the survival signal transmitted from the adjacent computer, and writes the result as a response to the survival signal transmitted from the adjacent computer in the survival signal transmitted to the adjacent computer. The computer identifies the failure location by combining the presence / absence of the survival signal from the computers on both sides on the virtual ring and the content of the response. Further, each computer notifies all computers with which it can communicate of information regarding the discovered failure. That is, by applying the transmission of the live signal periodically, the combination of the destination computers, and the device of the content of the live signal according to claim 3 to the duplicated LAN, the reception state of the live signal itself, Fault detection is performed by combining with the reception status of the surviving signal at the transmission source computer. This makes it possible to minimize the number of surviving signals exchanged in normal times and in the event of an abnormality, and it is possible to detect a failure of one computer by multiple computers in the vicinity of the failure computer, resulting in a shorter delay from the occurrence of the failure. The chance of finding a failure in time increases.

According to the eighth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein each computer transmits a survival signal to an adjacent computer at regular intervals in the survival signal transmitting step, with respect to the survival signal transmitted from the adjacent computer. Along with the response, a copy of the response from the other adjacent computer that is different from the adjacent computer is also written. In this way, by periodically transmitting the survival signal, devising the combination of the destination computers, and devising the content of the survival signal as claimed in claim 3 to the duplicated LAN, the reception state of the survival signal itself is obtained. And the reception state of the surviving signal at the transmission source computer are combined to detect the failure. This makes it possible to minimize the number of surviving signals exchanged in normal times and in the event of an abnormality, and it is possible to detect a failure of one computer by multiple computers in the vicinity of the failure computer, resulting in a shorter delay from the occurrence of the failure. The chance of finding a failure in time increases.

In the failure detecting method for the distributed computer system according to the ninth aspect of the present invention, a plurality of computers are arranged on a virtual virtual ring. Each computer is divided into an even-numbered computer and an odd-numbered computer according to the order in a specific direction from a specific computer on the virtual ring, and the odd-numbered computer is the first LA.
The survival signal is periodically transmitted to the computers on both sides of the virtual ring via N, and the even-numbered computer periodically transmits the survival signal to the computers on both sides of the virtual ring via the second LAN. . Each computer checks whether or not a survival signal transmitted from both sides is received via the first or second LAN, and by combining the results, the failure location is identified, and each computer finds the found failure. Informs all computers with which it can communicate of information about. In this way, the transmission of the survival signal on a regular basis, the combination of transmission destination computers, and the arrangement of the content of the survival signal as claimed in claim 3 are duplicated.
By applying to the AN, the failure detection is performed by combining the reception state of the survival signal itself and the reception state of the survival signal at the transmission source computer. This makes it possible to minimize the number of surviving signals exchanged in normal times and in the event of an abnormality, and it is possible to detect a failure of one computer by multiple computers in the vicinity of the failure computer, resulting in a shorter delay from the occurrence of the failure. The chance of finding a failure in time increases.

According to a tenth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein each computer is divided into a plurality of groups of three, and in each group, the first computer is divided into the second computer and the first computer. Sends a survival signal periodically via the LAN of the third computer and the second LAN to the third computer.
Send a live signal periodically via. Further, in each group, the second computer checks whether or not the survival signal from the first computer has been received, and periodically transmits the result to the third computer via the second LAN. Write to the survival signal. The third computer checks whether or not the survival signal from the first computer has been received, and writes the result in a survival signal that is periodically transmitted to the second computer via the first LAN. The second computer examines the existence and contents of the survival signal transmitted from the first and third computers, and by combining the results, identifies the failure location, and the third computer uses the first and third computers. The presence / absence and content of the survival signal transmitted from the second computer are checked, and the failure location is specified by combining the results. Then, each computer notifies all computers with which it can communicate of the information regarding the discovered failure. In this way, by periodically transmitting the survival signal, devising the combination of the destination computers, and devising the content of the survival signal as claimed in claim 3 to the duplicated LAN, the reception state of the survival signal itself is obtained. And the reception state of the surviving signal at the transmission source computer are combined to detect the failure. This makes it possible to minimize the number of surviving signals exchanged in normal times and in the event of an abnormality, and it is possible to detect a failure of one computer by multiple computers in the vicinity of the failure computer, resulting in a shorter delay from the occurrence of the failure. The chance of finding a failure in time increases.

In the fault detecting method for the distributed computer system according to the eleventh aspect of the present invention, when a fault occurs, a relocation step of newly setting the virtual placement of each computer is further executed. Each computer is equipped with a method of setting the transmission and reception destination of each computer according to the current operating status of each computer, so that when a failure occurs, when a failed computer is restored, or when a new computer is added, each computer Even if the destination is changed and the system configuration changes, the same failure detection capability as before is maintained.

According to the twelfth aspect of the present invention, in the method of detecting a failure in a distributed computer system, when a failure occurs, a relocation step of newly setting the virtual layout of each computer is further executed. Each computer is equipped with a method of setting the transmission and reception destination of each computer according to the current operating status of each computer, so that when a failure occurs, when a failed computer is restored, or when a new computer is added, each computer Even if the destination is changed and the system configuration changes, the same failure detection capability as before is maintained.

According to a thirteenth aspect of the present invention, there is provided a distributed computer system failure detection method, further comprising a relocation step of re-setting a virtual layout of each computer when a failure occurs. Each computer is equipped with a method of setting the transmission and reception destination of each computer according to the current operating status of each computer, so that when a failure occurs, when a failed computer is restored, or when a new computer is added, each computer Even if the destination is changed and the system configuration changes, the same failure detection capability as before is maintained.

In a distributed computer system failure detection method according to a fourteenth aspect of the present invention, when a failure occurs, a relocation step of newly setting the virtual layout of each computer is further executed. Each computer is equipped with a method of setting the transmission and reception destination of each computer according to the current operating status of each computer, so that when a failure occurs, when a failed computer is restored, or when a new computer is added, each computer Even if the destination is changed and the system configuration changes, the same failure detection capability as before is maintained.

According to a fifteenth aspect of the present invention, a distributed computer system fault detecting method uses a survival signal in a fault notifying step of notifying an adjacent computer of the detected fault information. Since the survival signal transmitted by each computer is used for notification of failure information, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

According to a sixteenth aspect of the present invention, there is provided a method for detecting a failure in a distributed computer system, wherein a live signal is used in a failure notification step of notifying the adjacent computer of the detected failure information. Since the survival signal transmitted by each computer is used for notification of failure information, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

According to the failure detecting method of the distributed computer system in the invention of claim 17, two LANs are paired, and each pair is defined by claim 6 to claim 10, claim 12, claim 14 and claim 16. Use one of the failure detection methods. Claims 6 to 10, claim 12, claim 1 for each group of three or more LANs in groups of two
4. By applying the method of claim 16, the invention can be applied to a system having an arbitrary number of LANs.

According to the failure detecting method of the distributed computer system in the invention of claim 18, two LANs are paired, and each pair is defined by claim 6 to claim 10, claim 12, claim 14 and claim 16. Any one of the failure detection methods is used, and for the remaining one, claims 1 to 5,
Any one of the failure detection methods of claim 11, claim 13 and claim 15 is used. Two or more sets of three or more LANs are set, and for each set, the method of claim 1 to claim 5, claim 11, claim 13 or claim 15, or claim 6 to claim 10, Claim 12, Claim 14, Claim 1
By applying the method of No. 6, the invention can be applied to a system having an arbitrary number of LANs.

A fault detecting method for a distributed computer system according to a nineteenth aspect of the present invention comprises forming a pair of two LANs,
Any one of the fault detection methods of claims 6 to 10, claim 12, claim 14, and claim 16 is made for each pair by further forming a pair by the (2N + 1) th LAN and any one of the LANs. Use or. Claims 6 to 1 for each set, with two or more sets of three or more LANs
By applying the methods of 0, claim 12, claim 14, and claim 16, the invention can be applied to a system having an arbitrary number of LANs.

The fault detection method for a distributed computer system according to the twentieth aspect of the present invention is L shared by two pairs.
At the AN, the survivor signals transmitted in each pair are combined. In the failure detection method according to claim 19, in a LAN shared by two LAN groups, the survival signals used in each group are combined into one, thereby reducing the number of exchanged survival signals.

In the failure detecting method for the distributed computer system according to the twenty-first aspect of the invention, in the virtual arranging step, the computers that frequently communicate with each other are arranged so as to be close to each other in the virtual arrangement. By arranging computers that are highly likely to communicate with each other in their original business at positions that are logically close to each other, the failure information of a certain computer can be quickly transmitted to such a computer. As a result, the influence of the failure occurrence on the original work is reduced.

In the fault detecting method for the distributed computer system according to the twenty-second aspect of the present invention, the computer having high reliability and the computer having low reliability are alternately arranged in the virtual arrangement in the virtual arrangement step. By placing a computer with high need for fault detection adjacent to a computer with high reliability, the latter is arranged so that the latter checks the survival signal of the former. As a result, it is possible to reliably detect a failure in a computer that has a high need for failure detection.

In the fault detecting method of the distributed computer system according to the twenty-third aspect of the present invention, in the virtual arrangement step, the highly reliable computers and the functionally important computers are alternately arranged in the virtual arrangement. By placing a computer with high need for fault detection next to a computer with high reliability,
The latter is arranged to check the survival signal of the former. As a result, it is possible to reliably detect a failure in a computer that has a high need for failure detection.

According to a twenty-fourth aspect of the present invention, there is provided a distributed computer system failure detection method, wherein the transmission time or reception deadline of some or all of the surviving signals is based on the time when each computer receives a specific surviving signal. Set. The survival signal is exchanged synchronously by setting the transmission / reception time of the survival signal with reference to the time when the self receives the survival signal. As a result, the relationship between the transmission and reception time of the survival signal can be freely set according to the required characteristics of failure detection.

[0071]

【Example】

Example 1. An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a physical configuration of a distributed computer system according to an embodiment of the present invention. In the figure, 101 to 104 are computers. Each calculator is
It is connected to a LAN (local area network) 401 via communication interfaces 211 to 214 and cables 311 to 314, respectively.

FIG. 2 is a diagram showing a virtual virtual ring of the distributed computer system according to this embodiment, and 10 is a virtual ring. In the distributed computer system, in order to detect a failure, the computers 101 to 104 are arranged on the virtual virtual ring 10 as shown in FIG. The computers 101, 102, 103, 104 rotate clockwise on the virtual ring.
However, this order can be set regardless of the physical positional relationship. As a method of arranging computers on a virtual ring, a method of bringing computers with high frequency of mutual communication close to each other, a method of arranging computers with high reliability and computers with low reliability, and a computer with high reliability are important. A possible method is to arrange the computers in charge of the functions alternately.

By the way, as will be described later in detail in the description of the operation, not only this embodiment but the fault detecting method for the distributed computer system according to the present invention is configured to have the following basic features. .

(1) A failure of a computer is found by the adjacent computer in the virtual arrangement relationship. (2) The failure information is transmitted using the survival signal. Therefore, the failure information is transmitted earlier to the computer that is closer to the computer that has found the failure computer in the virtual arrangement relationship.

The above-mentioned method of bringing the computers with high frequency of mutual communication close to each other utilizes these two basic features, whereby the original work is performed when a computer fails. For this reason, a computer that frequently communicates with the computer can receive the failure information at an early stage, and can prevent the original work from being stopped by continuing the communication without knowing the failure. . Therefore, it is possible to reduce the influence of the failure on the original work of the computer.

A method of alternately arranging a computer with high reliability and a computer with low reliability, and a method of alternately arranging a computer with high reliability and a computer in charge of important functions have the characteristics of the above (1). Therefore, a highly reliable computer is expected to reliably detect a failure of an adjacent unreliable computer or a computer in charge of an important function. Therefore, by constructing such a virtual layout relationship, it is possible to reliably detect the failure of an adjacent computer with low reliability, or a computer with a high need for failure detection, such as a computer in charge of important functions. it can.

Next, the operation will be described. FIG. 3 is a flowchart showing the operation of the distributed computer system, and FIG. 4 is a block diagram for explaining the transmission / reception of the survival signal of the computers arranged on the virtual ring as shown in FIG. 6 is a diagram showing the operation of the distributed computer system when the cable 312 of the computer 102 fails, and FIG.
FIG. 3 is a diagram showing a reconfigured distributed computer system after a failure. Below, referring to these figures,
While corresponding to each step of the flowchart shown in
The operation of the computer of the distributed computer system will be described.

As shown in FIG. 4, computers 101 to 104
Respectively periodically transmit a survival signal to the computer located on the right side on the virtual ring via the LAN 401. Specifically, it is checked whether or not the transmission timer for periodically transmitting the survival signal is 0 (step ST1), and if the transmission timer is 0, that is, if a predetermined time-out time has elapsed. For example, the survival signal is transmitted, the timeout time is set, and the survival signal transmission timer is set (step ST2).

Normally, the computer is operating in the normal mode (step ST3), and each computer checks whether or not the survival signal from the computer on the left is received at regular intervals (step ST4), and receives it. In this case, a predetermined timeout time is set and the survival signal reception timer is set or reset (step ST5). Further, it is checked whether failure information is added to the survival signal (step ST6).

In step ST4, when the computer does not receive the survival signal from the computer on the left, there is a possibility that the following failure has occurred.

1) Failure of the computer on the left side 2) Failure of the communication interface of the computer on the left side 3) Failure of the cable connecting the computer on the left side to the LAN 4) Failure of the cable connecting itself to the LAN 5) Failure of own communication interface

Hereinafter, as shown in FIG. 5, the operation of the following distributed computer system will be described assuming that the cable 312 connecting the communication interface 212 of the computer 102 and the LAN 401 is out of order.

The computers 102 and 103 cannot receive the survival signal from the computers 101 and 102 on the left side, respectively. When the survival signal reception fails, the computers 102 and 103 first determine whether the survival signal reception timer is 0, that is, whether the survival signal has not been received for a predetermined time (step ST10), and receive the signal. If the timer is 0, above 1)
It is determined that any of the failures in 5) to 5) has occurred, the mode shifts to the failure detection mode, the computers 102 and 103 select a computer other than the computer on the left, and transmit a survival signal to the computer. Is sent, and a survival signal reception timer is reset by setting a predetermined timeout time (step ST11). The case where the reception timer is 0 corresponds to the case where the reception timer has timed out. After step ST11, step ST1
Returning to step ST3, since the survival signal transmission timer is not 0 but in the failure detection mode, the steps ST3 to ST14 are performed.
Move to.

The computer 102 cannot receive the live signal for this request due to the failure of the cable 312. Therefore, the computer 102 determines that the failure of 4) or 5) described above has occurred, determines whether the reception timer of the survival signal is 0 in step ST19, and then takes measures such as restarting itself. Perform (step ST20).

On the other hand, the computer 103 shown in FIG. 5 can receive a response from a computer other than the computer on the left side selected in step ST11 by sending a request to transmit a survival signal, and therefore 1) to 3). It is judged that the failure of 1) has occurred (step ST14). Next, the computer 103 creates a new virtual ring from which the computer 102 has been removed.
In the new virtual ring, as shown in FIG. 6, the computers are arranged clockwise in the order of 101, 103, 104. In this way, when the configuration of the computer changes due to a failure or restoration, the process of changing the virtual placement of the computer is called reconfiguration. By performing the reconfiguration, the distributed computer system can exhibit the same failure detection capability as before the failure occurred. Further, as a result, since the destination computer of the survival signal of the computer 101 is changed from the computer 102 to the computer 103, the computer 103 transmits a request to change the destination of the survival signal to the computer 101 (step ST15).

Further, the computer 103 needs to notify all the other computers with which it can communicate of the failure information of the computer 102. The survival signal is used for notification. Each computer writes the failure information in the survival signal transmitted by itself, so that the failure information is sequentially transferred to the adjacent computers on the ring and notified to all the computers. The operation for notifying the failure information will be described below. The computer 103 writes the failure information indicating the failure occurrence of the computer 102 into the survival signal to be transmitted next (step ST16), saves the failure information in the memory in preparation for retransmission, and sets the failure information timer (step ST17). ), And returns to the normal mode (step ST18).

On the other hand, in step ST6, when each computer finds failure information indicating a failure occurrence in the received survival signal and the source of the failure information is not its own (step ST7), it is the same as the computer 103. And the similar failure information is written in the survival signal to be transmitted next (step ST9).

In step ST7, it is checked whether or not the failure information contained in the survival signal is output by itself, that is, whether or not the message related to the failure information has reached itself by traveling around the ring within a predetermined time. If received, the message is deleted from the memory and the failure information timer is deleted (step ST8). If it cannot be received, the failure information is written in the survival signal again and the retransmission is attempted. That is, the computer 103 that has found the failure transmits the failure information together with the survival signal to the computer 104 on the right side on the virtual ring in accordance with step ST16, as described above. The computer 104 transmits the failure information to the computer 101 on the right side in the same manner. Repeating this, if the failure information makes one round in the virtual ring, it is confirmed that the failure information has been notified to all computers, and as shown in step ST8, the message relating to the failure information is deleted from the memory, Delete the failure information timer. However, if the computer on the way fails while the failure information has completed a cycle, the failure information may be lost. In order to prevent this, in step ST4, the survival signal is not received,
If the reception timer is not 0 in step ST10, after transmitting the failure information, it is checked whether or not the failure information has returned to itself within the time-out time set in step ST17, making one round in the loop of the virtual ring. (Step ST12), if the time-out period has been exceeded, the failure information is added to the survival signal again and transmitted to the adjacent computer, and the failure information timer is reset (step ST13).

Incidentally, steps ST1 and ST2 are survival signal transmission steps, and steps ST3, ST4, ST10 and S
T11 and ST14 are failure detection steps, step ST5
~ ST9 and ST16 to ST18 correspond to the failure notification step, and steps ST15 and ST9 correspond to the reconfiguration step.

The distributed computer system according to this embodiment is
Since the failure detection function is distributed to each computer, the failure detection function is not lost due to the failure of a specific computer. Further, in this distributed computer system, each computer only transmits one survival signal every cycle in order to notify the survival of itself. Therefore, the number of surviving signals transmitted and received by each computer in normal times can be minimized, and the load on the computer is reduced. Also, since the total number of live signals transmitted on the LAN is proportional to the number of computers, the LAN
Load, that is, the number of signals transmitted on the LAN per unit time can be reduced. Further, in this distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be further reduced.

Example 2. FIG. 7 is a flow chart showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention. The distributed computer system according to this embodiment has a physical configuration as shown in FIG. 1 as in the first embodiment. Further, the respective computers are arranged on the virtual virtual ring 10 shown in FIG. 2 for detecting the failure. At this time, as described in the first embodiment, an arrangement method focusing on some attributes of the computer can be considered.

Next, the operation will be described. The operation of each computer will be described below in correspondence with each step of the flowchart shown in FIG.

Each computer checks whether the existence signal transmission timer to the adjacent computer X is 0 (step S
T21), while transmitting the survival signal periodically, set a predetermined timeout time and set the survival signal transmission timer, and change the transmission destination from right adjacent to left adjacent or left adjacent to right adjacent Yes (step ST22). That is, the transmission destinations are the right adjacent computer, the left adjacent, the right adjacent, the left adjacent, and so on for each cycle, such that the odd adjacent cycle is the right adjacent computer and the even cycle is the left adjacent computer. Next, each computer checks whether or not a survival signal has been received from the left and right adjacent computers at a predetermined time (step ST23), and if not, further reception of the survival signal is within a predetermined time. (Step ST34), the reception timer is reset (step ST37), and step S37.
Return to T21. FIG. 8 shows how the surviving signals of each computer are transmitted and received at this time.

Hereinafter, as shown in FIG. 9, it is assumed that the cable 312 connecting the communication interface 212 of the computer 102 and the LAN 401 is out of order, and the distributed computer is in the order of failure detection, ring reconfiguration, and failure information notification. The operation of the system will be described.

Due to the occurrence of a failure, the computer 101 is on the right side,
Computers 102 cannot receive signals from both computers, and computer 103 cannot receive signals from the computer on the left (step ST3).
4). When each computer detects the first failure in receiving the live signal, it further checks whether a live signal from another adjacent computer can be received (step ST35). Since the computers 101 and 103 can receive the survival signal from the computer 104, they determine that the computer 102 has failed. That is, in FIG. 7, steps ST35 to ST
After 37, the process returns to step ST21 and step ST22.
It is possible to change the transmission destination from the computer 102 to the computer 104 and receive the survival signal from the computer 104 (step ST23), but since the survival signal from the computer 102 times out without being received, (Step ST24), the computer 102 determines that there is a failure.

On the other hand, the computer 102 itself, which has failed to receive the live signal from one adjacent computer in step ST34, also fails to receive the live signal from the other adjacent computer (step ST35).
It is determined that the connection with 401 has been disconnected, and measures such as restarting are performed (step ST36). The above judgment is
It is based on the assumption that the probability of two computers failing at the same time is very low.

If the computer 103 is the adjacent computer 102,
Suppose you first discovered the breakdown. The computer 103 deletes the computer 102 from the ring and sets a new arrangement on the virtual ring. Such processing is called ring reconstruction. As a result, the computers adjacent to the computer 103 are changed to the computers 101 and 104 (step ST25). Figure 10
FIG. 4 is a diagram showing a virtual ring of the distributed computer system after reconfiguration. Even if a failure occurs due to the reconfiguration, it is possible to maintain the same level of failure detection capability as before the failure occurred.

Further, the computer 103 needs to notify all computers with which it can communicate of failure information of the computer 102. The operation for notifying the failure information will be described below.

First, the computer 103 executes step ST2.
In 5, the failure information is written in the survival signal to be transmitted to the computers on both sides thereafter. Calculator 101, 10
4 receives the surviving signal and discovers that failure information is written therein (step ST26), reconfigures the ring accordingly (step ST27), and then from the other adjacent computer. It is confirmed whether the failure information has already been received (step ST28), and if it has not been received yet, the failure information is written in the survival signal to one of the adjacent computers (step ST29). In addition, a reception confirmation for notifying the sender that the failure information has been received is added to the survival signal to be sent to the computer 103 (step S).
T31).

On the other hand, when the computer 103 finds the reception confirmation in the surviving signal (step ST32), thereafter, the failure information is not added to the surviving signal to the sender computer of the surviving signal (step ST33).

The same processing as above is repeated on other computers. As a result, the failure information is transferred clockwise and counterclockwise on the ring, starting from the computer 103.
The failure information transferred clockwise and counterclockwise is simultaneously received by a computer. At this time, since the computer determines in step ST28 that the failure information has already been received, the failure information is not further transferred,
Only reception confirmation is performed (step ST30).

Incidentally, steps ST21, ST22, ST3
7 is a survival signal transmission step, steps ST23 to ST
25 and ST34 to ST36 correspond to the failure detection step, steps ST25, ST26 and ST28 to ST33 correspond to the failure notification step, and steps ST25 and ST27 correspond to the reconstruction step.

The distributed computer system according to this embodiment is
Since the failure detection function is distributed to each computer, the failure detection function is not lost due to the failure of a specific computer. In addition, the distributed computer system can minimize the amount of communication for fault detection in normal times. Further, unlike the first embodiment, it is not necessary to transmit / receive an additional signal in order to specify the failure location, so the communication amount at the time of failure is small. Furthermore, in the distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

Example 3. FIG. 11 is a flow chart showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention. The distributed computer system of this embodiment has the physical configuration shown in FIG. 1, and each computer is virtually arranged on a virtual ring 10 as shown in FIG. Similar to the first embodiment, for the arrangement of the computers on the virtual ring 10, several methods that consider various attributes of the computers can be considered.

Next, the operation will be described. Below, FIG.
While corresponding to each step of the flowchart shown in
The operation of each computer will be described.

Each computer periodically checks whether or not the survival signal transmission timer is 0 so as to transmit the survival signal to the computer on the right (step ST41), and transmits the survival signal. The survival signal transmission timer is set by setting a predetermined timeout time (step ST4).
2). Next, it is checked whether or not the survival signal from the computer on the left side is regularly received (step ST43), and if it is not received, the failure information added to the survival signal is received for a predetermined time. It is checked whether or not it has been performed within (that is, whether or not it has timed out) (step ST57), and if it has not timed out, the process returns to step ST41.

At step ST43, each computer
If a live signal is received within a predetermined time, a preset timeout time is set to set or reset a live signal reception timer, and ACK is written to the live signal to be transmitted to the right (step ST44). . Otherwise, the process proceeds to step ST53, and if the survival signal reception timer is 0, NAK is written in the survival signal (step ST54). In FIG. 12, each computer is ACK or N
It is a block diagram of the distributed computer system which showed a mode that AK was written in a survival signal.

As shown in FIG. 13, as in the first embodiment,
Assuming that a failure has occurred in the cable 312 that connects the communication interface of the computer 102 and the LAN 401, the operation of the following failure detection method will be described.

The computers 102 and 103 cannot receive the survival signals from the computers 101 and 102, respectively. Therefore, as described above, the computer 102 transmits the survival signal including the NAK to the computer 103 and the computer 103 to the computer 104 according to steps ST53 and ST54. Therefore, the contents of the survival signal transmitted by each computer are as shown in FIG. 2 including NAK
Of the two survival signals, the survival signal transmitted by the computer 102 is the computer 10 because the cable 312 has a failure.
3 is not reached. Therefore, only the computer 104 is NAK
The survival signal to which the failure information including is added is received (step ST45). Therefore, the computer 104 is the computer 10
It is determined that a failure has occurred in No. 2, and the destination of the survival signal of the computer 101 is changed in the same manner as in the first embodiment (step ST50). Further, the computer 104 next writes the failure information in the survival signal to be transmitted to transfer the failure information (step ST51), stores the failure information in the memory in preparation for retransmission, and sets a predetermined timeout time. Set and set failure information timer (step S
T52).

On the other hand, if it is found in step ST45 that the surviving signal includes ACK, each computer checks whether or not the received surviving signal includes failure information indicating a failure occurrence (step S45). In ST46), when the failure information is found and the transmission source of the failure information is not its own (step ST47), the same failure information is written in the survival signal to be transmitted next (step ST49).

In step ST47, it is checked whether or not the failure information in the survival signal is issued by itself, that is, whether or not the message related to the failure information has reached itself by going around the ring within a predetermined time. If received, the message is deleted from the memory and the failure information timer is deleted (step ST48). On the other hand, if the survival signal cannot be received and the reception of the failure information exceeds the timeout time set in step ST52 (step S
T57), similarly to the first embodiment, the failure information is written into the next survival signal again, the failure information timer is reset, and the retransmission is attempted (step ST58).

Note that steps ST41 and ST42 are survival signal transmission steps, and steps ST43, ST44 and ST
53 to ST56 are failure detection steps, step ST45
-ST52, ST57, and ST58 correspond to the failure notification step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function will not be lost due to the failure of a specific computer. In addition, the distributed computer system can minimize the number of live signals transmitted and received by each computer during normal operation, which reduces the load on the computer. Moreover, since the total number of live signals transmitted on the LAN is proportional to the number of computers, the load on the LAN, that is, the number of signals transmitted on the LAN per unit time can be reduced. . Furthermore, in the distributed computer system, since the survival signal is used to notify the failure information, there is no need to send an extra signal for notification,
The load on the LAN can be reduced.

Example 4. FIG. 14 is a flow chart showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention, and FIG. 15 is a block diagram showing a virtual arrangement of the distributed computer system according to this embodiment. Further, the distributed computer system of this embodiment is shown in FIG.
It has the physical configuration shown in.

In the distributed computer system according to this embodiment, as shown in FIG. 15, computers 101 to 107 are arranged on the nodes of a virtual tree, except for the computers 103, 104, 106 and 107 at the bottom of the tree. , Configure the tree so that each computer has multiple child computers. Hereinafter, when viewed from a certain computer on the tree, the computer at the position of the parent node is called a parent computer, the computer at the position of the child node is called a child computer, and the computers at the same level are called sibling computers. Similar to the first embodiment, for arranging the computers on the tree, several arrangement methods can be considered by paying attention to various attributes of the computers.

Next, the operation will be described. Below, FIG.
While corresponding to each step of the flowchart shown in
The operation of each computer will be described.

Each computer checks whether the transmission timer of the survival signal is 0 in order to periodically transmit the survival signal to the parent computer (step ST61), and if the transmission timer is 0, the survival signal is 0. And the survival signal transmission timer is set (step ST62). Each computer checks whether or not the survival signal from a plurality of child computers has been received (step ST63), and when it does not receive the survival signal, it checks whether or not the survival signal reception timer is 0. (Step ST66), these results are combined to determine the failure location.

The computer corresponding to the root of the tree detects a failure of the root computer by transmitting a survival signal to the computer at the bottom of the tree. FIG. 16 is a diagram showing how a live signal is transmitted and received in the distributed computer system having the virtual tree arrangement shown in FIG.

Next, as shown in FIG. 17, the computer 105
The operation of the failure detection method according to this embodiment when the cable 315 connecting the communication interface 215 and the LAN 401 is disconnected will be described.

Due to this failure, the computer 105 cannot receive the survival signals from all the child computers, so that it is determined that the survival signals of all the child computers have timed out in steps ST66 and ST67, and the computer itself and the LAN.
It is assumed that a failure has occurred during 401, and measures such as restarting itself are performed (step ST68).

On the other hand, the computer 101 detects that the survival signal from the computer 105 cannot be received within a predetermined time (step ST66). After a while, when the computer 101 detects the survival signal from the computer 102 (step ST63), the survival signal of the child computer 105 has timed out, causing the computer 105 to fail, or between the computer 105 and the LAN 401. It is determined that a failure has occurred (step ST64). The above judgment is based on the assumption that the probability of simultaneous failure of two computers is extremely low. Computer 101
Reconfigures the tree as shown in FIG. 18 by using the child computer (grandchild computer) of the child computer 105 as its own child computer. As a result, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred. The computer 101 notifies the new child computer of the failure information and the new configuration information. The computer that has received the failure information and the configuration information further notifies the child computer of this (step ST65).

Note that steps ST61 and ST62 are survival signal transmission steps, and steps ST63, ST64 and ST
66 to ST68 are failure detection steps, step ST65
Corresponds to the failure notification step and the reconstruction step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function is not lost due to the failure of a specific computer. Furthermore, since the distributed computer system only sends one survival signal in each cycle in order to inform each computer of its own survival, it is possible to minimize the number of survival signals transmitted and received by each computer during normal operation. The load on the computer is reduced. Also, LA
Since the total number of surviving signals transmitted on N is proportional to the number of computers, the load on the LAN, that is, the number of signals transmitted on the LAN per unit time can be reduced.

Example 5. FIG. 19 is a flow chart showing the operation of the fault detecting method for the distributed computer system according to another embodiment of the present invention, and FIG. 20 is a block diagram showing the virtual configuration of the distributed computer system according to this embodiment ( Hereinafter, such a configuration is called a chain). Similar to the first embodiment, for the virtual arrangement of the computers in the distributed computer system, several methods that pay attention to various attributes of the computers can be considered.

As shown in FIG. 20, in the distributed computer system according to this embodiment, first, the computer 101 and the computer 1
05 and group 1001, computer 102 and computer 10
3 and the computer 104 form a group 1002. Also,
The computers 101 and 104 are group 1001 and group 1001, respectively.
This is a representative computer representative of 1002, and these representative computers 101 and 104 are arranged on a virtual virtual ring. Computers 102, 103, 10 other than the representative computer
5 is configured to periodically transmit a survival signal to the representative computer of the group to which it belongs. Figure 21
FIG. 3 is a block diagram showing transmission / reception of a survival signal of a distributed computer system having such a chain structure.

Next, the operation will be described. Below, FIG.
The operation of the representative computer will be described with reference to the flowchart of FIG.

The representative computers 101 and 104 arranged on the virtual ring periodically transmit survival signals to the representative computers on the right side, and the computers other than the representative computer periodically transmit the survival signal to the representative computer. In order to do so, it is checked whether the transmission signal of the survival signal is 0 (step ST7).
1) The transmission of the survival signal is started, a predetermined timeout time is set, and the survival signal transmission timer is set (step ST72). Therefore, each representative computer receives survival signals from other computers in the same group and from the representative computer on the left side. For example, the representative computer 104
Receives the survival signals from the computers 102 and 103 of the group 1002 and the survival signal from the representative computer 101 on the left side.

Each representative computer checks whether or not the survival signal from the computer on the left side is received at regular intervals (step ST73), and when received, sets a predetermined time-out time to set the survival signal. The reception timer is set or reset (step ST74). When the representative computer fails to receive the survival signal, first, the representative computer determines whether the survival signal reception timer is 0, that is, whether the survival signal has not been received for a predetermined time (step ST83), Detect failure. That is, each representative computer estimates the failure location by combining the reception statuses of the signals transmitted to itself.

As an example, the operation of the failure detecting method continued when the computer 102 as shown in FIG. 22 fails will be described below.

The representative computer 104 cannot receive the surviving signal of the computer 102 (step ST83), but can receive the surviving signals from the representative computer 101 and the computer 103. Therefore, the process proceeds from step ST73 to step ST74.
Since failure information is not added to the received survival signal, the process proceeds to step ST79, it is determined whether or not the survival signal reception of the transmission computer 102 has timed out, and it is determined that a failure has occurred in the computer 102. Such a judgment is based on the assumption that two computers have a very low probability of simultaneously failing.

The computer 104 removes the computer 102 from the group, forms a new chain as shown in FIG. 23, and with respect to the other computers 103 of the group 1002,
The failure information is notified by a signal different from the survival signal (step ST80). As a result, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred.

Next, the computer 104 transfers the fault information to another representative computer by adding the detected fault information to the survival signal (step ST8).
1). The transfer of the failure information using the survival signal is performed in the first embodiment.
In the same manner as above, the failure information is stored in the memory and the failure information timer is set in preparation for retransmission (step ST82).
Then, in step ST75, when the representative computer finds failure information indicating the occurrence of a failure in the received survival signal, and the source of the failure information is not its own (step ST76), the representative computer sends the failure information to the next survival signal. And a signal different from the survival signal is sent to other computers in the group (step ST78). Also,
In step ST76, it is checked whether or not the failure information in the survival signal is the one issued by itself, that is, whether or not the message regarding the failure information has reached itself by going around the ring within a predetermined time. If so, the message is deleted from the memory and the failure information timer is deleted (step ST77).

Also, as in the first embodiment, if the computer in the middle fails while the failure information goes around the virtual ring, the failure information may be lost. In order to prevent this, if the survival signal is not received in step ST73 and the reception timer is not 0 in step ST83, the failure information is transmitted within the timeout period set in step ST82 after the failure information is transmitted. It is checked whether or not it has returned to itself (step ST86), and if the timeout time has been exceeded, the failure information is added to the survival signal and transmitted to the adjacent computer again, and the failure information is returned. The timer is reset (step ST87).

Next, as another example, the operation of the failure detecting method according to this embodiment when the communication interface 314 of the representative computer 104 fails as shown in FIG. 24 will be described.

The representative computer 104 is the other representative computer 10.
1 and the computers 102 and 103 of the same group 1002
Since it is not possible to receive any of the surviving signals, the process goes to step ST84 through step ST83, and the receiving of the surviving signals from all the transmission computers exceeds the timeout time. For this reason, the computer 104 considers that the connection between itself and the LAN has been disconnected, and takes measures such as restarting itself (step ST85).

Also, the representative computer 101 is the representative computer 1
Since the live signal 04 of the computer 104 cannot be received and the live signal of the computer 105 can be received, the failure of the computer 104 can be found (step ST79). Next, computer 10
1 constructs a new chain as shown in FIG. 25, and notifies other computers of failure information in the same manner as the previous example (steps ST80 to ST82, ST76, ST77,
ST86, ST87).

Incidentally, steps ST71 and ST72 are survival signal transmission steps, and steps ST73, ST74 and ST
83 to ST85 are failure detection steps, step ST75
~ ST82, ST86 to ST87 are failure notification steps,
Step ST80 corresponds to the reconstruction step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function is not lost due to the failure of a specific computer. In addition, since the distributed computer system only sends one survival signal in each cycle in order to notify each computer of its own survival, failure detection in normal times and identification of a failure location at the time of failure occurrence are possible. Can minimize the amount of communication. Furthermore, in the distributed computer system, since the survival signal is used to notify the failure information between the representative computers, the load on the LAN can be reduced.

Example 6. FIG. 26 is a flow chart showing the operation of the fault detecting method for a distributed computer system according to another embodiment of the present invention, and FIG. 27 is a physical configuration diagram of the distributed computer system according to this embodiment.
01 to 104 are computers, and 401 and 402 are LANs. The computers 101 to 104 are connected to the LAN 401 by communication interfaces 211 to 214 and cables 311 to 314, respectively. Also, the computer 10
1 to 104 are communication interfaces 221 to 22 respectively
4 and the cables 321 to 324, the LAN 402
It is connected to the.

In the distributed computer system according to this embodiment, since the faults are detected, the computers 101 to 1 are used as in the first embodiment.
04 are arranged in a virtual virtual ring shape as shown in FIG.
Similar to the first embodiment, for the virtual arrangement of computers, several methods that consider various attributes of the computers can be considered.

Next, the operation will be described. Hereinafter, FIG.
The operation of each computer will be described with reference to the flowchart of FIG.

Each computer transmits a survival signal to the computer located on the right side of the virtual arrangement. For convenience of explanation, when assigning a number to each computer according to the arrangement order on the virtual ring, the odd-numbered computer is LAN 401.
Send a survival signal to the computer on the right using
Further, the even-numbered computer uses the LAN 402 to check whether the transmission timer for periodically transmitting the survival signal is 0 in order to regularly transmit the survival signal to the computer on the right side (step ST91). ), If the transmission timer is 0, the transmission of the survival signal is started, a predetermined timeout time is set, and the survival signal transmission timer is set (step ST92). The survival signal from the computer on the left side is received through the LAN 402 by the odd-numbered computer and is received through the LAN 401 by the even-numbered computer. FIG. 28 is a block diagram showing transmission / reception of a survival signal of each computer.

Normally, the computer is operating in the normal mode (step ST93), and each computer checks whether the survival signal from the computer on the left is received at regular intervals (step ST94), and receives it. In this case, a predetermined timeout time is set and the survival signal reception timer is set or reset (step ST95). Further, it is checked whether failure information is added to the survival signal (step ST96).

In the following, as a failure example 1, a cable 3 for connecting the computer 102 and the LAN 402 as shown in FIG.
The operation of the fault detection method for the distributed computer system according to this embodiment when the switch 22 is disconnected will be described.

When the computer 103 detects that the surviving signal from the computer 102 cannot be received within a predetermined time (step ST100), it shifts to the failure detection mode, and the computer 103 itself A signal requesting itself to transmit a survival signal is sent using LAN2 (step ST101).

After that, in the computer 103, step ST
Returning to step 91, the process proceeds to step ST93, and since it is in the failure detection mode, the process branches to step ST104. On the other hand, if this request signal is sent to the computer 104, the computer 1
04 transmits a survival signal to the computer 103 in response to the request signal. In step ST104, the computer 103 determines that a failure has occurred in the connection between the computer 102 and the LAN 402 since the response to the request signal has been obtained. Further, the computer 103 reconfigures and resets the transmission destination of the survival signal of each computer as shown in FIG. 30 (step ST105). By performing such reconfiguration, even if a failure occurs, it is possible to maintain the same level of failure detection capability as that before the failure. At this time, the computer 103 directly reports the failure information to the computer whose destination is changed (step ST106), then saves the failure information in the memory and sets the failure information timer in preparation for retransmission (step ST106). ST107) and returns to the normal mode (step ST108). Further, other computers are notified of the failure information by adding the failure information to the survival signal and sequentially transferring it to the adjacent computers (step ST
106, ST107, ST97, ST99). Example 1
Similarly, the computer 103 checks whether the failure information is transferred to itself and returned within a fixed time (steps ST96 and ST97), and if received, deletes the message from the memory. Then, the failure information timer is deleted (step ST98). On the other hand, when the failure information is lost during one cycle and the failure information does not return, retransfer is performed (steps ST102 and ST103).

Next, as another failure example, the communication interface of the computer 103 and the LAN 40 as shown in FIG.
The operation of the failure detecting method according to this embodiment when the cable 323 connecting the cable 2 and the cable 2 fails will be described.

When the computer 103 detects that the survival signal from the computer 102 cannot be received within the predetermined time (step ST100), the computer 103 shifts to the failure detection mode, and the computer 103 survives itself to the appropriate computer. A signal requesting transmission of a signal is sent using LAN2 (step ST101). After that, the computer 103 returns to step ST91 and, since it is in the failure detection mode, branches to step ST104 via step ST93. Since the request signal does not reach the destination computer, the computer 103
Cannot receive a response to the request signal.
Therefore, the computer 103 shifts from step ST104 to step ST109, and determines that a failure has occurred between itself and the LAN 402 because the response cannot be received within the predetermined time (step ST105). The computer 103 resets the transmission destination of the survival signal of each computer as shown in FIG. 32, notifies each computer of the failure information in the same manner as in failure example 1, saves the failure information in the memory, and saves the failure information timer. Is set, and the process returns to the normal mode (steps ST105 to ST108). Further, as in the failure example 1, the computer 103 checks whether the failure information is transferred to itself and returned within a fixed time (step ST
96, ST97), if received, the message is deleted from the memory and the failure information timer is deleted (step ST98). On the other hand, when the failure information is lost during one cycle and the failure information does not return, retransfer is performed (steps ST102 and ST103).

Note that steps ST91 and ST92 are survival signal transmission steps, and steps ST93, ST94 and ST
100, ST101, ST104, and ST109 are failure detection steps, steps ST95 to ST99, and ST106.
~ ST108 is a failure notification step, step ST10
5, ST99 corresponds to the reconstruction step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function is not lost due to the failure of a specific computer. In addition, the distributed computer system can minimize the amount of communication for detecting a failure during normal operation and for specifying a failure location when a failure occurs. Furthermore, in the distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

Example 7. FIG. 33 is a flow chart showing the operation of the failure detecting method for the distributed computer system according to another embodiment of the present invention. The distributed computer system according to this embodiment has the same physical configuration as that of the sixth embodiment. In the distributed computer system, the computers 101 to 104 are arranged in a virtual virtual ring shape shown in FIG. 2 for detecting a failure. Has been done. Further, as in the first embodiment, for the virtual arrangement of computers, several methods that pay attention to various attributes of the computers can be considered.

Next, the operation will be described. Below, FIG.
The operation of each computer will be described with reference to the flowchart of FIG.

Each computer checks whether or not the transmission signal of the survival signal to the computer x on the right side or the left side is 0 in order to transmit the survival signal to the adjacent computer in the virtual arrangement ( Step ST111), the transmission timer is 0,
That is, if the survival signal has not been transmitted within the predetermined time, the survival signal is transmitted and the transmission destination is changed from right adjacent to left adjacent or left adjacent to right adjacent (step ST1).
12). Then, the process returns to step ST11. At this time,
In LAN401, the computer on the right side is the destination,
In 02, the computer on the left is set as the transmission destination. As a result, the two adjacent computers use different LANs to transmit the survival signal to the other computer. Such a path for exchanging a survival signal between adjacent computers is called a loop. FIG. 34 is a block diagram showing transmission / reception of a survival signal via a loop in the failure detection method according to this embodiment.

By using the loop, each computer determines whether or not the survival signal from the adjacent computer can be received.
It is possible to respond to this neighboring computer by using the survival signal transmitted to the neighboring computer. If adjacent computer x
The transmission signal of the survival signal to the
If the surviving signal from the computer can be received within the predetermined time (step ST113), the computer writes ACK in the surviving signal as a response to the adjacent computer x (step ST113).
114). When the survival signal cannot be received within the defined time (step ST122), the computer writes NAK in the survival signal as a response to the adjacent computer (step S122).
T123). On the other hand, if the computer that fails to receive the live signal from one adjacent computer in step ST122 also fails to receive the live signal from the other adjacent computer (step ST124), then it is said that itself is isolated. Judge and take action such as restart (step ST1)
25).

As a result, each computer receives the survival signal from the adjacent computer, and the content of the response is ACK or NA.
It is either K or cannot receive the survival signal itself. Hereinafter, the response when the survival signal cannot be received is expressed as “No Msg.”.

In FIG. 35, the hatched computer indicates that the computer on the left has ACK, NAK, NoMsg. FIG. 6 is a diagram showing a range of possible failures when each response of FIG. In the case of ACK, both routes to the two computers and the two LANs are normal. On the other hand, in the case of NAK, the computer on the left cannot receive the survival signal within the time specified by the computer on the right indicated by the diagonal line, and at this time the computer on the left is to receive this survival signal. There is a possibility that the route for sending data is broken. Also, No Ms
g. In the case of, there may be a failure in the computer on the left or the path of the survival signal from the computer on the left.

Since each computer can receive the responses from the computers on both sides, it is possible to specify the range where the failure exists by combining the two responses. FIG. 36 shows
The range of the failure specified from the combination of the survival signals received by the computer indicated by the diagonal lines is shown. Of the nine cases shown in FIG. 36, the failure location can be identified in 20 cases.
02, 2004, 2005 (step S
T116, ST118). In addition, cases 2003 and 20
In 07, the failure range extends over a plurality of locations, but in this case, it is determined that a failure has occurred at the location indicated by the cross mark in the figure (step ST117).

In the distributed computer system, the loop causes
When exchanging the survival signal between the adjacent computers, the survival signal from the partner computer is received, and then the survival signal to the partner computer is transmitted after an appropriate waiting time. By repeating this mutually, transmission / reception of the survival signal of a constant cycle is realized by a synchronous method (step ST126).

As an example of setting the waiting time, when the survival signal of the computer on the left side is received, the waiting time is set to 0, and when the survival signal of the computer on the right side is received, the waiting time is set to T. There is a way. In this method, the response to the surviving signal sent to the computer on the right is obtained without delay, so the surviving signal from the computer on the right always accurately reflects the state of the distributed computer system. There are advantages.

Next, as a failure example, the operation of each computer will be described by using the case where the cable 323 connecting the communication interface of the computer 103 and the LAN 402 is disconnected as shown in FIG.

By disconnecting the cable 323, the computer 10
The survival signal transmitted by the computer 3 to the computer 102 using the LAN 402 and the survival signal transmitted by the computer 104 to the computer 103 using the LAN 402 are not received by the target computer (step ST122). Therefore, the computer 102 transmits a survival signal including NAK to the computer 103 and the computer 103 to the computer 104 using the LAN 401, respectively (step ST123). As a result, the computer 10
2 receives ACK and No from the computers on both sides
Msg. (Step ST117). The response received by the computer 103 is NAK and No Msg. The response received by the computer 104 is NAK and ACK (step ST116). Therefore, as shown in FIG. 37, both the computer 102 and the computer 104 determine that a failure has occurred in either the cable 323 or the communication interface 223 (step ST119). Calculator 10
3 can detect that a failure has occurred around itself from the combination of responses (case 2008), but cannot specify its position. For this reason,
It waits for the adjacent computer 102 or 104 to identify it and notify the failure content (step ST120).

The computer 102 or 104 reconfigures the ring as shown in FIG. With the new configuration, Calculator 1
02 and 104 send a response to each other via the LANs 401 and 402. Further, the computer 103 can send and receive signals using the LAN 401, so that
In order to confirm the connection of No. 1, the survival signal is periodically received from the computer 102, and the survival signal is transmitted to the computer 104 to notify the survival of itself.
As a result, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred.

In order to notify the failure information as described above to all the computers, a method similar to that of the second embodiment is used, and the failure information is sequentially transferred to the adjacent computer while confirming the delivery of the failure information between the adjacent computers. (Step ST121).

Incidentally, steps ST111, ST112, S
T126 is a survival signal transmission step, step ST11
3, ST114, ST122, ST123 are survival signal response steps, steps ST115 to ST120, ST1.
24, ST125 is a failure detection step, step ST1
21 is a failure notification step, steps ST119 and ST1
20 corresponds to the reconstruction step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function is not lost due to the failure of a specific computer. Also, in the distributed computer system, the communication amount for detecting a failure during normal operation requires twice as much as that in the sixth embodiment, but since the order of the communication amount is the same, the communication amount increases as the number of computers increases. Little increase. In addition, even when an abnormality occurs, an extra signal for identifying a failure location is not required, so that the communication amount can be small. Further, in the distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to transmit an extra signal for notification, and the total number of survival signals transmitted on the LAN is Since the number is proportional to the number of computers, the load on the LAN can be reduced.

Example 8. FIG. 39 is a block diagram showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention. The distributed computer system of this embodiment has the same physical and virtual configuration as that of the seventh embodiment. Also, as in the case of the seventh embodiment, a survival signal is exchanged between adjacent computers using a loop.

In the distributed computer system, the content of the live signal includes two fields. The first field is a response indicating to each adjacent computer whether or not each computer was able to receive the survival signal of the adjacent computer, as in the case of the seventh embodiment. FIG. 39
As shown in, the second field corresponds to the computer copying the first field (response) of the live signal received from the other adjacent computer, which is different from the destination of the live signal.

As in the case of the seventh embodiment, each computer combines the responses included in the survival signals from both sides to specify the fault location. Further, by combining the contents of the above two fields included in one survival signal, the failure location can be specified. FIG. 40 shows a failure location that can be identified from the content of the survival signal received by the computer, which is indicated by hatching.

Next, the operation will be described. As an example of failure, a cable 323 connecting the communication interface 223 of the computer 103 and the LAN 402 as shown in FIG.
The operation of each computer will be explained taking the case of disconnection.

Due to a cable failure, computers 103, 1
The live signal 04 transmitted to the left by using the LAN 402 is not received by the destination computer. Therefore, computer 1
The response to the computer on the right of 02 and 103 is NAK.
That is, the first field of the survival signal transmitted to the right of the computers 102 and 103 using the LAN 401 is NAK. Therefore, the second field of the survival signal transmitted by the computers 103 and 104 to the right of the LAN 401 is NA.
It becomes K. The second field of the survival signal transmitted from the computer 102 to the computer on the left using the LAN 402 is No Msg. Becomes The second field of other survival signals becomes ACK. Therefore, the computer 101 determines from the survival signal received from the computer 104 that the case 3003
Or the case 3005 can be detected from the survival signal received from the computer 102 to detect the failure. In addition, the computers 102 and 104 combine two survival signals received from adjacent computers,
A failure can be detected. In addition, the computer 103
From the survival signal received from the computer 102, the case 300
By applying 2, the failure can be detected.

Reconstruction after failure detection and notification of failure information are the same as in the seventh embodiment. By the reconfiguration, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred.

The distributed computer system according to this embodiment is
Since the failure detection function is distributed to each computer, the failure detection function is not lost due to the failure of a specific computer. Further, in the distributed computer system, each computer only sends one survival signal in every cycle in order to notify the existence of its own, so failure detection in normal times and identification of a failure location at the time of failure occurrence are possible. Communication volume can be minimized. Further, in the distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to send an extra signal for notification,
Since the total number of surviving signals sent to N is proportional to the number of computers, the load on the LAN can be reduced.

Example 9. FIG. 42 is a flow chart showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention. The distributed computer system of this embodiment has the same physical configuration as that of the sixth embodiment shown in FIG. In the distributed computer system, each computer is arranged in a virtual virtual ring shape as shown in FIG. 2 in order to detect a failure. Similar to the first embodiment, for the virtual arrangement of computers, several methods that consider various attributes of the computers can be considered.

Next, the operation will be described. Below, FIG.
The operation of each computer will be described with reference to the flowchart of FIG.

Among the computers constituting the virtual ring, whether the even-numbered computer has a live signal transmission timer to the adjacent computer x is 0 in order to periodically transmit the live signal to both adjacent computers using the LAN 401. It is checked whether or not (step ST131), the survival signal is transmitted, the predetermined timeout time is set, the survival signal transmission timer is set, and the transmission destination is changed from one adjacent computer x to the other adjacent computer x ′. Is changed to (step ST132).
Similarly, the odd-numbered computers periodically transmit the survival signal to the computers on both sides using the LAN 402. Each computer checks whether or not it has received a live signal (step ST133), and if it has not received, it further checks whether or not the reception of a live signal from an adjacent computer has exceeded the timeout time (step ST133). ST144) If the timeout has not occurred, the reception timer is reset (step ST147) and the process returns to step ST131. FIG. 43 shows
FIG. 7 is a block diagram showing how a live signal is transmitted and received by each computer in the distributed computer system according to this embodiment. As shown in FIG. 43, two adjacent computers use different LANs to transmit a survival signal to the other computer. Such a survival signal exchange path is called a loop.

Hereinafter, as a failure example 1, as shown in FIG. 44, the communication interface 213 of the computer 103 and the LAN
The operation of each computer will be described by using the case where the cable 313 connecting to 401 is broken.

Due to the failure, the survival signal transmitted by the computer 313 is not received by the destination computers 102 and 104 (step ST144). Next, it is checked whether or not the reception of the survival signal from the other computer different from the computer 103 has exceeded the timeout time (step ST145).
03, since a survival signal is received from the other adjacent computer 101 different from 03, steps ST147, ST131,
After ST132, the operation proceeds to step ST133. Therefore,
In the computers 102 and 104, it is determined that any of the computer 103, the communication interface 213, and the cable 313 has failed (steps ST133, ST134,
ST135). Such a judgment is based on the assumption that two computers have a very low probability of simultaneously failing.

As a failure example 2, as shown in FIG. 45, the communication interface 223 of the computer 103 and the LAN 402 are connected.
Using the case where the cable 323 connecting to and fails,
The operation of each computer will be described.

Due to a failure, the survival signal transmitted by the computers 102 and 104 cannot be received by the computer 103 (step ST144). Therefore, the computer 103
When it is determined that the surviving signals of the computers on both sides cannot be received (step ST145), the self and the LAN
It is determined that the connection with 402 is disconnected (step ST146).

In failure example 1, the ring is reconfigured as shown in FIG. Further, in failure example 2, the ring is reconfigured as shown in FIG. As a result, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred. In order to notify all computers of the above failure information,
Using the same method as in the second embodiment, the failure information is sequentially transferred to the adjacent computer while confirming the delivery of the failure information between the adjacent computers (steps ST136 to ST143).

Incidentally, steps ST131, ST132, S
T147 is a survival signal transmission step, step ST13
3, ST134, ST135, ST144 to ST146
Is a failure detection step, steps ST136 and ST138
~ ST143 and ST146 correspond to the failure notification step, and steps ST135 and ST137 correspond to the reconstruction step.

Since the distributed computer system distributes the failure detection function to each computer, the failure detection function is not lost due to the failure of a specific computer. Further, in the distributed computer system, each computer only sends one survival signal in every cycle in order to notify the existence of its own, so failure detection in normal times and identification of a failure location at the time of failure occurrence are possible. Communication volume can be minimized. Further, in the distributed computer system, since the survival signal is used to notify the failure information, it is not necessary to transmit an extra signal for notification, and the total number of survival signals transmitted on the LAN is Since the number is proportional to the number of computers, the load on the LAN can be reduced.

Example 10. FIG. 48 is a flow chart showing the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention. The distributed computer system of this embodiment has the same physical configuration as that of the sixth embodiment shown in FIG.

In the distributed computer system according to this embodiment, three computers are grouped in groups of three as shown in FIG. 49 for detecting a failure. When constructing such a group, as in the case of the first embodiment, several methods that consider various attributes of the computer can be considered.

Next, the operation will be described. Below, FIG.
Corresponding to the flow chart of
The operations of 03, 104 and 106 will be described.

Each computer in the group is called A, A ', B depending on its role in detecting a failure. In FIG. 49, the computers 101 and 104 are A, and the computers 102 and 105.
Corresponds to B, and the computers 103 and 106 correspond to A ′. B is L
Send a live signal to A periodically using AN401, and
The AN 402 is used to periodically send a live signal to A '. Using the LAN 402, A checks whether or not the transmission timer of the survival signal is 0 in order to periodically transmit the survival signal to A '(step ST151), and if the transmission timer is 0, The survival signal is transmitted and the survival signal transmission timer is reset (step ST152). At this time,
No survival signal is received from A '(step ST153),
If the survival signal from B can be received (step ST1
61), write ACK to the survival signal (step ST1)
62), reception of the survival signal from A ′ has not exceeded the predetermined timeout time (step ST168), and if the survival signal from B cannot be received, NAK is written (steps ST172 and ST173). A ′, like A, writes the presence / absence signal from B to the survival signal, and A
To periodically send a survival signal to.

On the other hand, A determines the failure location by using the existence and contents of the survival signal from A'and B, and A'determines the failure location by using the existence and content of the survival signal from A and B. to decide. FIG. 50 shows the range of failure locations estimated from the combination of survival signals received by A. In some cases, there may be one anomalous failure at multiple locations. In the distributed computer system, a candidate for a failure location is searched for by using the assumption that there is one failure location at the same time. Furthermore, among the candidate failure locations, it is considered that the failure has occurred in the candidate closest to the computer that determines the failure. The X mark in FIG. 51 indicates a failure location determined by such a criterion.

Next, as a failure example 1, a communication interface 212 of the computer 102 and a LAN as shown in FIG.
The operation of the distributed computer system according to the present embodiment when the cable 312 connecting to 401 is disconnected will be described.

By cutting the cable 312, the computer 10
The survival signal transmitted by 2 (B) is not received by the computer 101 (A) (step ST172). Therefore, the computer 101 transmits NAK to the computer 103 (A ′) (step ST173). When the computer 101 (A) receives the survival signal ACK from the computer 103 (A ′),
Applying the case 5004 shown in FIG.
Is determined to have failed (steps ST153 to ST15)
5, ST157, ST159, ST177). Such a judgment is based on the assumption that two computers have a very low probability of simultaneously failing. Calculator 103 (A ')
Receives a survival signal from the computer 102 (B) and receives a survival signal NAK from the computer 101 (A). This corresponds to the case 5002 in FIG. 50, but in this case the failure location cannot be specified (steps ST154 to ST1).
56, ST160, ST177).

As Failure Example 2, the communication interface 211 of the computer 101 and the LAN 401 as shown in FIG.
Take the case where the cable 311 connecting to and is disconnected.

By disconnecting the cable 311, the computer 10
1 (A) cannot receive all the survival signals. For this reason, the computer 101 (A) determines that the cable 311 has failed by applying the case 5006 of FIG. 50 (steps ST168 to 171, ST177, or step ST1) when it is determined that both survival signals cannot be received.
72-ST177). The computer 103 (A ′) receives the survival signal from the computer 102 (B) and
The survival signal NAK is received from (A). This is shown in FIG.
In this case, the failure location cannot be identified (steps ST154 to ST156, ST
160, ST177).

As Failure Example 3, as shown in FIG. 53, the communication interface 221 of the computer 101 and the LAN 402 are connected.
Take the case where the cable 321 connecting to and is disconnected.

When the cable 321 is cut, the computer 10
The survival signal transmitted by 1 (A) is the computer 103 (A ').
Is not received (step ST168). Computer 103
When the survival signal ACK is received from the computer 102 (B), (A ′) applies the case 5003 of FIG. 50 and determines that the cable 321 has failed (step ST161).
~ ST166, ST177). Computer 101 (A)
Since all the survival signals are normal, it is not possible to detect a failure (steps ST153, ST154, ST
177 or steps ST161 to ST163, ST1
77).

As Failure Example 4, the case where the computer 102 (B) fails as shown in FIG. 54 will be taken up.

Due to the failure of the computer 102 (B), both the computers 101 (A) and 103 (A ') are
The survival signal from (B) cannot be received (step ST172). Computer 101 (A), 103
(A ′) are computers 103 (A ′) and 101, respectively.
The survival signal NAK is transmitted to (A) (step ST17).
3). The computers 101 (A) and 103 (A ′), when receiving the survival signal NAK, apply the case 5005 of FIG. 50 and determine that the computer 102 (B) has failed (steps ST153 to ST159, ST177). .

The steps ST151 and ST152 are
Survival signal transmission step, steps ST161 and ST16
2, ST173 and ST174 are first and second survival signal response steps, steps ST154 to ST158 and ST.
160, ST169, ST170 are the first failure detection step, steps ST163 to ST165, ST167,
ST174 and ST175 are second failure detection steps, steps ST159, ST166, ST171 and ST17.
6 is a failure notification step, steps ST159 and ST16
0, ST166, ST167, ST171, ST176
Corresponds to the reconstruction step. The computer A corresponds to the second computer, the computer B corresponds to the first computer, and the computer A ′ corresponds to the third computer.

When a computer, communication interface, or cable failure is found, the found computer replaces the failed computer with a computer having the same role in another group to perform reconfiguration. The replaced computer will belong to two groups at the same time. As an example, if the computer 102 fails, the computer 105
FIG. 55 shows how the computer is replaced with. As a result, even if a failure occurs, it is possible to maintain the same level of failure detection capability as before the failure occurred. The result of the reconfiguration is notified to all computers by appropriate broadcast communication. In the distributed computer system according to this embodiment, the failure detection function is distributed to each computer, so that the failure detection function is not lost due to the failure of a specific computer. Further, in the distributed computer system, similarly, the communication amount for detecting a failure during normal times and for specifying a failure location when a failure occurs can be minimized.

Example 11. The distributed computer system according to this embodiment has the same physical configuration as that of the seventh embodiment and a virtual configuration. The operation for detecting a failure of the distributed computer system according to this embodiment is also almost the same as that of the seventh embodiment.

Next, the operation will be described. Hereinafter, parts different from the seventh embodiment will be described. In the distributed computer system, when each computer transmits a survival signal to the adjacent computer on the right using the LAN 401, each computer periodically transmits the survival signal by a timer managed by each computer. When each computer receives the survival signal from the neighbor on the left, it immediately executes AC
The survival signal in which K is written is transmitted. If the survival signal from the computer on the left is not received within the predetermined time, the survival signal in which NAK is written is transmitted.

The fault detecting method in this embodiment has the same advantages as the method of setting the transmission / reception timing of the live signal shown in the seventh embodiment. Further, this embodiment corresponds to the inventions of claim 7, claim 12, claim 16, and claim 21 to claim 24.

Example 12 56 is a block diagram showing the configuration of a distributed computer system according to another embodiment of the present invention. In the figure, 101 and 102 are computers, 211, and
212, 221, 222, 231, 232, 241, 2
42 is a communication interface, 311, 312, 32
1, 322, 331, 332, 341, and 342 are cables, and 401 to 404 are LANs.

In the distributed computer system according to this embodiment, each computer is connected to four LANs 401 to 404. In the distributed computer system, among the four LANs, LAN401 and LAN402, LAN403 and LA
N404 is set as each set, and each set is used to detect a failure by the same method as in the sixth embodiment.

As a result, the inventions of claims 1 to 16 can be applied to a general distributed computer system having an even number of LANs.

Example 13 57 is a block diagram showing the configuration of a distributed computer system according to another embodiment of the present invention. In the figure, the same reference numerals as those in FIG. 56 designate the same or corresponding components.

In the distributed computer system according to this embodiment, each computer is connected to three LANs 401-403. In the distributed computer system, of the three LANs,
Failure detection is performed using the LANs 401 and 402 in the same manner as in the sixth embodiment. Further, using the remaining LAN 403, failure detection is performed by the same method as in the first embodiment.

As a result, the inventions of claims 1 to 16 can be applied to general distributed computer systems having an odd number of LANs.

Example 14. 58 is a block diagram showing the configuration of a distributed computer system according to another embodiment of the present invention. In the figure, the same reference numerals as those in FIG. 56 designate the same or corresponding components.

In the distributed computer system according to this embodiment, each computer is connected to three LANs 401-403. In the distributed computer system, of the three LANs,
LAN401 and LAN402, LAN403 and LAN4
02 as a set. In each group, failure detection is performed by the same method as in the sixth embodiment, but in the LAN 402, two survival signals for failure detection executed in each group are set to 1
By putting them together, the transmission frequency of the survival signal is reduced.

Thus, the inventions of claims 1 to 16 can be applied to general distributed computer systems having an odd number of LANs.

[0210]

As described above, according to the invention of claim 1, the virtual placement step of placing a plurality of computers on a virtual virtual ring, and each computer is adjacent in a specific direction on the virtual ring. A survival signal transmitting step of periodically transmitting a survival signal indicating the survival of oneself to the computer
Each computer periodically checks whether or not it has received the survival signal transmitted from the adjacent computer on the virtual ring. If not, it is determined that an error has occurred in the communication path used to transmit the survival signal. Then, the failure detection step of identifying the failure location and failure information regarding the failure discovered by each computer are
Since it is configured to execute the failure notification step of notifying all computers that can communicate, the number of surviving signals exchanged during normal times to detect the failure of the computers, communication interfaces or cables that make up the distributed computer system. There is an effect that can minimize. In addition, each computer notifies other computers of the information of the fault found within its own range, so that even if a plow failure occurs,
There is an effect that each computer can obtain operation information of the entire system.

According to the invention of claim 2, in the live signal transmitting step in which the computer regularly transmits the live signal,
The distributed computer system is configured with a minimum amount of communication even when a failure occurs, because it is configured to switch alternately on the virtual ring at regular timings and send a survival signal to the computer to the right or left. There is an effect that a failure of a computer, a communication interface or a cable that operates can be detected, and the number of surviving signals exchanged in normal times and when an abnormality occurs can be minimized.

According to the invention of claim 3, in the survival signal transmitting step, whether or not the computer has received the survival signal to be received within a predetermined time is written in the survival signal to be transmitted and transmitted. Therefore, by sending and receiving the minimum survival signal during normal times, it is possible to detect failures in the computers, communication interfaces, or cables that make up the distributed computer system, and to detect the survival signals that are exchanged during normal times and when an error occurs. This has the effect of minimizing the number.

According to the invention of claim 4, a virtual arranging step of arranging each computer on a virtual virtual tree having each node as a node and having two or more child nodes, and each computer on the virtual tree Survival signal transmission step that periodically transmits a survival signal to the parent computer located at the parent node, and whether or not each computer has received the survival signal from the child computer located at the child node on the virtual tree. And the failure detection step of identifying the failure location by combining the results, and the failure notification step of notifying each computer of the information about the found failure to all computers with which it can communicate. Even when a failure occurs, the failure of the computer, communication interface or cable that constitutes the distributed computer system can be detected and the failure can be found with a minimum amount of communication. There is an effect of minimizing the number of viable signals exchanged when an abnormality occurs.

According to the invention of claim 5, the computer is divided into M groups, one computer is set as a representative computer in each group, and the M representative computers are arranged on a virtual virtual ring. Virtual allocation step, a first survival signal transmission step in which a computer other than the representative computer periodically transmits a survival signal to the representative computer of the group to which the computer belongs, and each representative computer has a specific direction on the virtual ring. The second survival signal transmitting step for periodically transmitting the survival signal to the adjacent computer, and whether or not each representative computer has received the survival signal transmitted to the representative computer, and combines the results to determine the failure. Each representative computer is configured to execute a failure detection step of identifying a location and a failure notification step of notifying all computers that can communicate of information regarding the found failure. The number of surviving signals exchanged during normal times and in the event of an abnormality can be detected with a minimum amount of communication even when a failure occurs, the failure of the computer, communication interface or cable that constitutes the distributed computer system can be detected. There is an effect that can minimize.

According to the invention of claim 6, a virtual arranging step of arranging a plurality of computers on a virtual virtual ring,
When dividing each computer into an even number and an odd number according to the order in a specific direction from a specific computer on the virtual ring, the odd numbered computer regularly sends to the adjacent computer on the virtual ring via the first LAN. To send a survival signal,
Survival signal transmission step in which the even-numbered computer periodically transmits the survival signal to the adjacent computer on the virtual ring via the second LAN, and each computer receives the survival signal transmitted to the computer. Whether or not the failure detection step of identifying whether or not the failure location is determined by combining the results, and the failure notification step of notifying all computers with which the computer can communicate the information regarding the discovered failure, are executed. Since it is configured, by sending and receiving the minimum survival signal during normal times, it is possible to detect failures in the computers, communication interfaces or cables that make up the distributed computer system. The effect is to minimize the number of.

According to the invention of claim 7, a virtual placement step of placing a plurality of computers on a virtual virtual ring,
Each computer periodically transmits a survival signal to a computer adjacent to a specific direction on the virtual ring via the first LAN, and the specific direction on the virtual ring via the second LAN. Survival signal transmission step to periodically transmit the survival signal to the adjacent computer in the opposite direction, and check whether each computer has received the survival signal transmitted from the adjacent computer, and the result is sent to the adjacent computer. Combining the survival signal response step to write in the survival signal to be transmitted as a response to the survival signal transmitted from the adjacent computer, and the presence or absence of the survival signal from the computers on both sides on the virtual ring and the content of the response To execute the failure detection step of identifying the failure location and the failure notification step of notifying all computers with which the computer can communicate the information about the discovered failure. Since the configuration is adopted, the number of surviving signals exchanged in normal times and in the occurrence of an abnormality can be minimized, and a failure of one computer can be found by a plurality of computers in the vicinity of the failure computer, resulting in a shorter failure time. There is an effect that a failure can be found in the delay time.

According to the invention of claim 8, each computer has a response to the survival signal transmitted from the adjacent computer in the survival signal transmitting step of transmitting a periodic survival signal to the adjacent computer and the adjacent computer. Since it is configured to write a copy of the response from the other adjacent computer that is different from, it is possible to minimize the number of surviving signals exchanged in normal times and during an abnormality, and
A failure of one computer can be found by a plurality of computers near the failure computer, and there is an effect that the failure can be found with a shorter delay time from the occurrence of the failure.

According to the invention of claim 9, a virtual arrangement step of arranging a plurality of computers on a virtual virtual ring,
When dividing each computer into an even number and an odd number according to the order in a specific direction from a specific computer on the virtual ring, the odd numbered computers become the adjacent computers on both sides of the virtual ring via the first LAN. Send a survival signal regularly,
An even-numbered computer periodically transmits a survival signal to adjacent computers on both sides of the virtual ring via the second LAN, and each computer has a first or second LA.
A failure detection step of checking whether or not a survival signal transmitted from both sides is received via N and specifying the failure location by combining the results, and each computer communicates information on the discovered failure. Since it is configured to execute the failure notification step of notifying all the obtained computers, it is possible to minimize the number of surviving signals exchanged at normal times and at the time of occurrence of an abnormality, and at the same time the failure of one computer Discovered by multiple computers in the vicinity,
There is an effect that a failure can be found with a shorter delay time from the occurrence of the failure.

According to the tenth aspect of the invention, a virtual placement step of placing a plurality of computers on a virtual virtual ring, and dividing each computer into a plurality of groups of three,
In each group, the first computer periodically sends the survival signal to the second computer via the first LAN, and at the same time sends the survival signal to the third computer via the second LAN. The survival signal transmitting step of transmitting, and in each group, it is checked whether the second computer has received the survival signal from the first computer, and the result is transmitted to the third computer via the second LAN. First survival signal response step of writing to the survival signal to be periodically transmitted by the third computer, and whether or not the third computer has received the survival signal from the first computer, and the result is used by the second computer. A second liveness signal response step for writing in a liveness signal to be periodically transmitted via the first LAN, and the presence / absence and contents of the liveness signal transmitted from the first and third computers by the second computer. Examine and combine those results According to the first failure detection step of identifying the failure location, the third computer checks the existence and contents of the survival signal transmitted from the first and second computers, and combines the results, Since the second failure detection step of identifying the failure point and the failure notification step of notifying each computer of the information about the found failure to all computers with which the computer can communicate are executed in normal times and abnormalities. The effect that the number of surviving signals exchanged at the time of occurrence can be minimized, and the failure of one computer can be found by multiple computers near the failed computer, and the failure can be found in a shorter delay time from the occurrence of the failure. There is.

According to the eleventh aspect of the present invention, when the failure occurs, the rearrangement step for resetting the virtual arrangement of each computer is further executed. Therefore, when the failure occurs, the failure computer is restored. At the same time, or when a new computer is added, the transmission destination of each computer can be changed, and even if the system configuration changes, the same failure detection capability as before can be maintained.

According to the twelfth aspect of the present invention, when the failure occurs, the relocation step for resetting the virtual placement of each computer is further executed. Therefore, when the failure occurs, the failure computer is restored. At the same time, or when a new computer is added, the transmission destination of each computer can be changed, and even if the system configuration changes, the same failure detection capability as before can be maintained.

According to the thirteenth aspect of the present invention, when a failure occurs, the rearrangement step for resetting the virtual arrangement of each computer is further executed. Therefore, when a failure occurs, the failure computer is restored. At the same time, or when a new computer is added, the transmission destination of each computer can be changed, and even if the system configuration changes, the same failure detection capability as before can be maintained.

According to the fourteenth aspect of the present invention, when a failure occurs, the rearrangement step for resetting the virtual arrangement of each computer is further executed. Therefore, when the failure occurs, the failure computer is restored. At the same time, or when a new computer is added, the transmission destination of each computer can be changed, and even if the system configuration changes, the same failure detection capability as before can be maintained.

According to the fifteenth aspect of the invention, in the failure notifying step of notifying the adjacent computer of the detected failure information, the failure information is added to the survival signal and the failure signal is notified by transmitting the survival signal. Since it is configured, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

According to the sixteenth aspect of the present invention, in the failure notifying step of notifying the adjacent computer of the detected failure information, the failure information is added by adding the failure information to the survival signal to notify the failure. Since it is configured, it is not necessary to transmit an extra signal for notification, and the load on the LAN can be reduced.

According to the invention of claim 17, 2N LAs are included.
In a distributed system consisting of a plurality of computers connected by N, two LANs are paired, and the failure detection method according to claim 6 to claim 12, claim 14, claim 14 and claim 16 for each pair. Since any one of the above is used, there is an effect that the above invention can be applied to a system having an arbitrary number of LANs.

According to the eighteenth invention, (2N + 1)
In a distributed system consisting of a plurality of computers connected by two LANs, two LANs are paired, and each pair has a failure according to claim 6, claim 12, claim 14, or claim 16. Any one of the detection methods is used, and the remaining one is configured to use any one of the failure detection methods of claims 1 to 5, claim 11, claim 13, and claim 15. Therefore, there is an effect that the above invention can be applied to a system having an arbitrary number of LANs.

According to the invention of claim 19, (2N + 1)
In a distributed system consisting of multiple computers connected by two LANs, two LANs are paired, and (2
N + 1) One LAN and one of the LANs further form one pair, and each pair forms claim 6 to claim 1.
Since any one of the 0, claim 12, claim 14, and claim 16 failure detection methods is used, the invention can be applied to a system having an arbitrary number of LANs. effective.

According to the twentieth aspect of the invention, in the LAN shared by two pairs, the survival signals transmitted in each pair are combined into one, so that the number of exchanged survival signals is increased. There is an effect that can be reduced.

According to the twenty-first aspect of the invention, in the virtual arrangement step, the computers that frequently communicate with each other are arranged close to each other in the virtual arrangement, so that the occurrence of a failure affects the original work. There is an effect that the influence can be reduced.

According to the twenty-second aspect of the invention, in the virtual arrangement step, the highly reliable computers and the less reliable computers are arranged alternately in the virtual arrangement. There is an effect that it is possible to reliably detect a high computer failure.

According to the twenty-third aspect of the present invention, in the virtual placement step, the highly reliable computers and the computers that are functionally important are arranged alternately in the virtual placement. There is an effect that a failure of a high-performance computer can be reliably detected.

According to the invention of claim 24, the transmission time or the reception deadline of some or all of the surviving signals is set on the basis of the time when each computer receives the specific surviving signal. Therefore, there is an effect that the relationship between the transmission and reception time of the survival signal can be freely set in accordance with the required characteristic of failure detection.

[Brief description of drawings]

FIG. 1 is a block diagram showing a physical configuration of a distributed computer system according to an embodiment of the present invention.

2 is a diagram showing computers arranged on a virtual virtual ring in the distributed computer system shown in FIG.

FIG. 3 is a flow chart for explaining the operation of the failure detection method in the distributed computer system shown in FIG.

4 is a block diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method in the distributed computer system shown in FIG.

5 is a diagram showing a failure example for explaining the operation of the failure detection method in the distributed computer system shown in FIG.

6 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for removing the influence of the fault in the fault detecting method for the distributed computer system shown in FIG. 1;

FIG. 7 is a flow chart for explaining the operation of a failure detection method in a distributed computer system according to another embodiment of the present invention.

8 is a diagram showing a state of transmission and reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG.

9 is a diagram showing a failure example for explaining the operation of the failure detection method of the distributed computer system shown in FIG.

FIG. 10 is a diagram showing a virtual ring after reconstruction is performed in order to eliminate the influence of a failure in the failure detection method for the distributed computer system shown in FIG.

FIG. 11 is a flow chart for explaining the operation of the failure detection method in the distributed computer system according to another embodiment of the present invention.

FIG. 12 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG. 11.

13 is a diagram showing a failure example for explaining the operation of the failure detection method of the distributed computer system shown in FIG.

FIG. 14 is a flow chart for explaining the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention.

15 is a block diagram showing a state in which each computer is virtually arranged on a tree in order to perform the failure detection method of the distributed computer system shown in FIG.

16 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG.

FIG. 17 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG.

18 is a diagram showing the configuration of a virtual tree after reconfiguration for removing the influence of a fault in the fault detection method for the distributed computer system shown in FIG.

FIG. 19 is a flow chart for explaining the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention.

20 is a block diagram showing a state in which each computer is virtually arranged in a chain in order to perform the failure detection method of the distributed computer system shown in FIG.

FIG. 21 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG.

22 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG.

23 is a diagram showing a configuration of a virtual tree after reconfiguration for removing the influence of the failure shown in FIG. 22 in the failure detection method for the distributed computer system shown in FIG. .

FIG. 24 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG.

25 is a diagram showing a configuration of a virtual tree after reconfiguration in order to remove the influence of the failure shown in FIG. 24 in the failure detection method for the distributed computer system shown in FIG. .

FIG. 26 is a flow chart for explaining the operation of a failure detection method for a distributed computer system according to another embodiment of the present invention.

27 is a configuration diagram showing a physical configuration of a distributed computer system to which the failure detection method for the distributed computer system shown in FIG. 26 is applied.

28 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG. 26.

29 is a block diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG.

FIG. 30 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the failure shown in FIG. 29 in the failure detection method for the distributed computer system shown in FIG. 26. Is.

31 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG.

32 shows a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the fault shown in FIG. 31 in the fault detection method for the distributed computer system shown in FIG. It is a figure.

FIG. 33 is a flow chart for explaining the operation of a failure detection method for a distributed computer system according to another embodiment of the present invention.

FIG. 34 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG. 33.

FIG. 35 is a table showing the relationship between the content of a live signal received by a computer and the possible range of a fault in the fault detection method for the distributed computer system shown in FIG. 33.

FIG. 36 is a table diagram showing the relationship between the content of a live signal received by a computer from adjacent computers and the range in which a fault may exist in the fault detection method for the distributed computer system shown in FIG. 33.

FIG. 37 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 33.

38 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the fault of FIG. 37 in the fault detection method for the distributed computer system shown in FIG. 33. Is.

FIG. 39 is a block diagram for explaining the operation of the failure detecting method for the distributed computer system according to another embodiment of the present invention.

40 is a table showing the relationship between the content of a live signal received by a computer from an adjacent computer and the range in which a fault may exist in the fault detection method for the distributed computer system shown in FIG. 39.

FIG. 41 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 39.

FIG. 42 is a flow chart for explaining the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention.

43 is a block diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG. 42.

FIG. 44 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 42.

45 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 42.

FIG. 46 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the failure of FIG. 44 in the failure detection method of the distributed computer system shown in FIG. 42. is there.

47 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the failure of FIG. 45 in the failure detection method of the distributed computer system shown in FIG. 42. is there.

FIG. 48 is a flow chart for explaining the operation of a failure detecting method for a distributed computer system according to another embodiment of the present invention.

FIG. 49 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system shown in FIG. 48.

50 is a table showing the relationship between the combination of survival signals received by a computer and the range in which a fault may exist in the fault detection method for the distributed computer system shown in FIG. 48.

51 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 48.

52 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 48.

FIG. 53 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 48.

FIG. 54 is a diagram showing a failure example for explaining the operation of the failure detection method for the distributed computer system shown in FIG. 48.

55 is a diagram showing a state of transmission / reception of a survival signal of each computer after reconfiguration for eliminating the influence of the fault of FIG. 54 in the fault detection method of the distributed computer system shown in FIG. 48. Is.

FIG. 56 is a diagram showing a state of transmission / reception of a survival signal of each computer for performing the failure detection method of the distributed computer system according to another embodiment of the present invention.

FIG. 57 is a block diagram showing a state of transmission / reception of a survival signal of each computer for performing a failure detection method for a distributed computer system according to another embodiment of the present invention.

FIG. 58 is a block diagram showing a state of transmission / reception of a survival signal of each computer for performing a failure detection method for a distributed computer system according to another embodiment of the present invention.

FIG. 59 is a block diagram showing an outline of a conventional fault detection method for a distributed computer system.

FIG. 60 is a diagram showing a strike failure in a conventional distributed computer system having a multiplexed LAN.

[Explanation of symbols]

10 virtual rings, 101-104 computers, 401-
404 LAN (Local Area Network), 10
01,1002 groups.

Claims (24)

[Claims] 1. A failure detection method for a distributed computer system including a plurality of computers connected to each other via at least one local area network, the virtual placement step of placing the plurality of computers on a virtual virtual ring. , A survival signal transmitting step of periodically transmitting a survival signal indicating the survival of itself to a computer adjacent in a specific direction on the virtual ring, and each computer on the virtual ring. Check whether the survival signal transmitted from the adjacent computer is regularly received, and if not, determine that an abnormality has occurred in the communication path used to transmit the survival signal, and identify the failure point. A detection step and a failure notification step in which each computer notifies failure information related to the found failure to all computers with which it can communicate. A method for detecting failures in distributed computer systems. 2. The survival signal transmitting step transmits the survival signal to the computer adjacent on the right side or the left side by alternately switching on the virtual ring at regular timings. Detection method for distributed computer systems in Japan. 3. The survival signal transmitting step writes and transmits to a survival signal to be transmitted whether or not the computer has received the survival signal to be received within a predetermined time. Item 2. A failure detection method for a distributed computer system according to item 2. 4. A failure detection method for a distributed computer system including a plurality of computers connected to each other through at least one local area network, wherein each computer has a node and each node has two or more child nodes. Virtual placement step to place on a virtual tree, each computer, to the parent computer located at the parent node on the virtual tree, a survival signal transmission step of periodically transmitting a survival signal, each computer, It is checked whether or not a survival signal from a child computer located at a child node on the virtual tree is received, and a failure detection step of identifying a failure location by combining the results, and each computer provides information on the found failure. , A failure notification step of notifying all computers with which communication is possible, a failure detection method for a distributed computer system. 5. A failure detection method for a distributed computer system including a plurality of computers connected to each other via at least one local area network, wherein M is a computer.
Into a group, one computer for each group is used as a representative computer, and M representative computers are arranged on a virtual virtual ring. A first survival signal transmitting step of periodically transmitting a survival signal to a representative computer of a group to which each of the groups belongs, and each representative computer periodically transmitting a survival signal to a computer adjacent in a specific direction on the virtual ring. A survival signal transmission step 2 and a failure detection step of checking whether or not each representative computer has received the survival signal transmitted to the computer and combining the results to identify a failure location;
A failure detection method for a distributed computer system, wherein each representative computer executes a failure notification step of notifying all computers with which it can communicate of information about the discovered failure. 6. A failure detection method for a distributed computer system including a plurality of computers connected to each other via first and second local area networks, wherein the plurality of computers are arranged on a virtual virtual ring. According to the arranging step and dividing each computer into an even number and an odd number according to an order in a specific direction from a specific computer on the virtual ring, an odd number computer may be the virtual computer via the first local area network. Survival signal that periodically transmits a survival signal to an adjacent computer on the ring, and an even-numbered computer that regularly transmits the survival signal to an adjacent computer on the virtual ring via the second local area network. The transmitting step and each computer checks whether or not the survival signal transmitted to the computer has been received, and based on the result, the failure location is identified. A failure detection method for a distributed computer system, comprising: performing a failure detection step to specify; and a failure notification step in which each computer notifies all computers with which it can communicate information about a discovered failure. 7. A failure detection method for a distributed computer system including a plurality of computers connected to each other via first and second local area networks, wherein the plurality of computers are arranged on a virtual virtual ring. An arranging step, in which each computer periodically transmits a survival signal to a computer adjacent in a specific direction on the virtual ring via a first local area network, and via a second local area network, A survival signal transmitting step of periodically transmitting a survival signal to the adjacent computer in the opposite direction to the specific direction on the virtual ring, and whether each computer has received the survival signal transmitted from the adjacent computer. And the result is written as a response to the survival signal sent from the adjacent computer to the survival signal sent to the adjacent computer. Answer step, each computer, the failure detection step for identifying the failure point by combining the presence or absence of the survival signal from both adjacent computers on the virtual ring and the content of the response, and the failure found by each computer And a failure notification step of notifying information to all computers that can communicate with the failure detection method of the distributed computer system. 8. Each computer is different from the adjacent computer together with a response to the survival signal transmitted from the adjacent computer in the survival signal transmitting step of transmitting a periodic survival signal to the adjacent computer. The failure detection method for a distributed computer system according to claim 7, wherein a copy of a response from one adjacent computer is also written. 9. A failure detection method for a distributed computer system including a plurality of computers connected to each other via first and second local area networks, wherein the plurality of computers are arranged on a virtual virtual ring. When the arrangement step and each computer are divided into an even number and an odd number according to the order in a specific direction from the specific computer on the virtual ring, the odd number computer is the first computer via the first local area network. Survival signal is transmitted periodically to the computers on both sides of the virtual ring, and the even-numbered computer periodically transmits the survival signal to the computers on both sides of the virtual ring via the second local area network. The signal transmission step and each computer transmits the survival signal transmitted from both sides via the first or second local area network. The failure detection step of checking whether or not received, and specifying the failure location by combining the results, and the information on the failure found by each computer,
And a failure notification step of notifying all computers with which communication is possible, a failure detection method for a distributed computer system. 10. A failure detection method for a distributed computer system including a plurality of computers connected to each other via first and second local area networks, wherein the plurality of computers are arranged on a virtual virtual ring. Arrangement step and dividing each computer into a plurality of groups of three, and in each group, the first computer periodically transmits a survival signal to the second computer via the first local area network. In addition, the survival signal transmitting step of periodically transmitting the survival signal to the third computer via the second local area network, and in each group, the second computer transmits the survival signal from the first computer. It is checked whether or not it has been received, and the result is sent to the third computer as the second
A first liveness signal response step of writing to the liveness signal that is periodically transmitted via the local area network of
The third computer checks whether or not the surviving signal from the first computer is received, and writes the result in the surviving signal that is periodically transmitted to the second computer via the first local area network. A second survival signal response step, and a second
The first computer checks the existence and contents of the survival signal transmitted from the first and third computers, and by combining the results, the first failure detection step of identifying the failure location and the third computer , A second failure detection step of identifying the failure location by checking the existence and contents of the survival signal transmitted from the first and second computers and combining the results, and the failure found by each computer And a failure notification step of notifying information to all communicable computers, a failure detection method for a distributed computer system. 11. The method according to any one of claims 1 to 3 and 5, further comprising a reallocation step for resetting a virtual layout of each computer when a failure occurs.
A failure detection method for a distributed computer system according to any one of the above. 12. The method according to claim 6, further comprising executing a rearrangement step of newly setting a virtual arrangement of each computer when a failure occurs. Detection method for distributed computer systems in Japan. 13. The method of detecting a failure in a distributed computer system according to claim 4, further comprising executing a rearrangement step of newly setting a virtual arrangement of each computer when a failure occurs. 14. The method of detecting a failure in a distributed computer system according to claim 10, further comprising executing a rearrangement step of newly setting a virtual arrangement of each computer when a failure occurs. 15. The failure notification step of notifying the adjacent computer of the detected failure information notifies the failure by adding the failure information to the survival signal and transmitting the survival signal. A failure detection method for the distributed computer system described. 16. The failure notification step of notifying the adjacent computer of the detected failure information notifies failure by adding failure information to the survival signal and transmitting the survival signal. A failure detection method for the distributed computer system described. 17. A distributed system consisting of a plurality of computers connected by 2N local area networks, wherein two local area networks are paired, and each pair is paired with one of claims 6 to 10, and
A method for detecting a failure in a distributed computer system, which uses any one of the failure detecting methods according to claim 2, claim 14 and claim 16. 18. In a distributed system comprising a plurality of computers connected by (2N + 1) local area networks, two local area networks are paired, and each pair is defined by any one of claims 6 to 10,
Any one of the failure detection methods of claim 12, claim 14, and claim 16 is used, and the remaining one is defined as claim 1 to claim 5, claim 11, claim 13, and claim. A failure detection method for a distributed computer system using any one of the failure detection methods of 15. 19. In a distributed system comprising a plurality of computers connected by (2N + 1) local area networks, two local area networks are paired, and a local area network of (2N + 1) th and any one of the local areas are paired. Further, one pair is formed by the area network, and each pair is defined by claims 6 to 10, claim 12, claim 14, and claim 16.
A fault detection method for a distributed computer system using any of the above fault detection methods. 20. A failure detecting method for a distributed computer system according to claim 19, wherein, in a local area network shared by two pairs, the survival signals transmitted in each pair are combined into one. . 21. In the virtual arranging step, computers that frequently communicate with each other are arranged so as to be close to each other in the virtual arrangement. A method for detecting a failure in a distributed computer system according to the item. 22. A computer having high reliability and a computer having low reliability are alternately arranged in a virtual arrangement in the virtual arrangement step. A method for detecting a failure in a distributed computer system according to the item. 23. A computer having high reliability and a computer which are functionally important are alternately arranged in a virtual arrangement in the virtual arrangement step. A method for detecting a failure in a distributed computer system according to one item. 24. For some or all of the survival signals,
The distributed computer system according to any one of claims 1 to 23, characterized in that the transmission time or the reception deadline is set based on the time when each computer receives a specific survival signal. Failure detection method.