JP2007124020A - Information processing system, information processor, and error prevention processing method of interconnect network and information processing system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce a system failure rate when a trouble occurs in an interconnect network without increasing traffic in the interconnect network. <P>SOLUTION: An error prevention processing method is used for an information processing system, where a plurality of information processors are connected via a plurality of interconnect networks. The error prevention processing method includes a step of monitoring a trouble symptom phenomenon, and detecting the trouble symptom phenomenon in data transmitted by the path of the interconnect network; a step of transmitting trouble symptom notification to the information processor of the issue source of the data; and a step of receiving the trouble symptom phenomenon notification before changing the path of a current active system interconnect network to that of other interconnect networks for further data transmission. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an information processing system and an error prevention processing method for an information processing system that realize high availability and high reliability of an inter-device transmission path, and more particularly to a duplex interconnect network information processing system and an error prevention processing method thereof.

  The duplex interconnect network computer to which the present invention relates is particularly used in computers that realize high availability of computers and high reliability of transmission paths between processing devices.

  With regard to this dual interconnect network computer, there is a method for improving the fault tolerance of the computer fault caused by the interconnect network fault. The fault tolerance of the inter-computer data transmission fault caused by the fault of the network between computers (the Internet in the typical example). There is a possibility that a method for improving the above can be adopted. Therefore, the prior art including a method for improving the fault tolerance of data transmission between computers will be considered.

  The following methods are available as methods for improving the fault tolerance of data transmission caused by an intra-computer interconnect network or inter-computer network failure.

  The first method is the method described in Patent Document 1. In this method, when transmitting data from the data issuing source device, the same data is transmitted to the two networks, and the data destination device receives the same data from the two networks, but picks one of the data, This is a method of discarding the other data. This method has an advantage that even if an unrecoverable failure occurs in one network, data transmission can be continued only in the other network.

JP 2002-229965 A

  However, the technique disclosed in Patent Document 1 has a problem that the traffic of the network increases twice because the same data is transmitted to two networks.

  Another method is to use one network as the active system and the other network as the standby system, and if an unrecoverable failure occurs in the active network, switch the standby network to the active system and continue data transmission. is there. In the case of this method, the data transmitted from the data issuing source device to the current active network is lost between the timing when the unrecoverable failure occurs in the active network and the time when the path is switched.

  Therefore, software such as a driver that controls the data transmission is rolled back to a processing point where it has been confirmed that data transmission has been normally performed, and data transmission is performed again from there. TCP / IP used for data transmission between computers such as the Internet adopts this method. However, when this method is used as a data transmission method inside a computer, there is a problem that data transmission cannot be retried by software, so that error recovery cannot be performed and the system goes down.

  For example, this problem occurs in data transmission for controlling data coherency in a computer. Consider a case in which a computer system having a plurality of CPU (Central Processor Unit) chips has a cache in each CPU. In this case, if data in one cache is updated, another cached data that has a copy of the updated data must also be updated to maintain data coherency. At this time, data is transmitted between the caches.

  Since such data transmission is data transmission that does not involve software such as an operating system, there is a problem that data transmission cannot be retried by software.

  SUMMARY OF THE INVENTION An object of the present invention is an information processing system and an error prevention processing method for solving a system down at the time of occurrence of an unrecoverable failure in an interconnect network without increasing the traffic of the interconnect network, which is the conventional problem described above. Is to provide.

  An error prevention processing method for an information processing system according to the present invention is an error prevention processing method for an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks. A step of detecting a failure predictor phenomenon of data transmitted through a network path, a step of transmitting a failure predictor notification to an information processing apparatus that issued the data, and a current active interconnect network after receiving the failure predictor notification And switching the other route to another interconnect network route and performing subsequent data transmission.

  In the error prevention processing method of the information processing system according to the present invention, the issuer information processing apparatus that has received the failure sign notification temporarily stops data transmission to the data destination information processing apparatus and is currently operating toward the data destination information processing apparatus. A route change notification is simultaneously sent to each of the system interconnect network and the current standby interconnect network, and the data destination information processing apparatus first receives the route change notification from the current standby interconnect network. In this case, the method may further include a step of waiting for data reception from the current standby interconnect network until a route change notification is received from the current active interconnect network.

  In the error prevention processing method of the present invention, the issuing information processing apparatus uses the current active interconnect network and the current standby interconnect network to simultaneously send a route change notification to each, and then the current standby interface. The method may further include a step of switching the connection network to be used as an operational interconnect network and restarting data transmission.

  According to the error prevention processing method of the information processing system of the present invention, when the data destination information processing apparatus receives a route change notification from the current active interconnect network, the current standby interconnect network is used as the active interconnect network. It may further include a step of switching and restarting the data reception.

  In the error prevention processing method for an information processing system according to the present invention, when the data destination information processing apparatus does not receive a route change notification from the current standby interconnect network, a route change notification is sent from the current standby interconnect network. After receiving, the method may further include a step of switching the current standby interconnect network to use as the active interconnect network.

  The information processing system according to the present invention is an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks. The interconnect network monitors a predictive failure phenomenon, and uses an interconnect network path. A failure sign phenomenon detection unit that detects a failure sign phenomenon of transmitted data, and an interface control unit that transmits a failure sign notification to the information processing apparatus that issued the data. A path switching control unit is provided for switching the path of the current active interconnect network to a path of another interconnect network after receiving the failure sign notification and performing subsequent data transmission.

The information processing system according to the present invention is an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks. The interconnect network monitors a predictive failure phenomenon, and uses an interconnect network path. A failure sign phenomenon detection unit that detects a failure sign phenomenon of transmitted data, and an interface control unit that transmits a failure sign notification to the information processing apparatus that issued the data,
After receiving the failure sign notification, the issuing information processing device temporarily stops data transmission to the data destination information processing device, and sends the current active interconnect to the data destination information processing device. Send a route change notification to each of the network and the current standby interconnect network at the same time, and then switch the route of the current active interconnect network to the route of the other interconnect network to transmit subsequent data With a route switching control unit,
If the data destination information processing device first receives a route change notification from the current standby interconnect network, it receives data from the current standby interconnect network until it receives a route change notification from the current active interconnect network. A path switching control unit that waits for

  In the information processing system of the present invention, the path switching control unit in the issuing information processing apparatus sends a path change notification to each of the current active interconnect network and the current standby interconnect network at the same time, and then Data transmission may be resumed by switching the system interconnect network to be used as the operational interconnect network.

  In the information processing system of the present invention, when the path switching control unit in the data destination information processing apparatus receives a path change notification from the current active interconnect network, the current standby interconnect network is used as the active interconnect network. The data reception may be resumed by switching to.

  In the information processing system of the present invention, when the path switching control unit in the data destination information processing apparatus does not receive a path change notification from the current standby interconnect network, it sends a path change notification from the current standby interconnect network. After reception, the current standby interconnect network may be switched to be used as the active interconnect network.

An information processing apparatus according to the present invention is an information processing system in which information processing apparatuses are connected via a plurality of interconnect networks for monitoring a failure predictor phenomenon and detecting a failure predictor phenomenon of data transmitted through an interconnect network path Is an information processing apparatus that constitutes
A path switching control unit is provided for switching the path of the current active interconnect network to a path of another interconnect network after receiving the failure sign notification and performing subsequent data transmission.

  The information processing apparatus of the present invention temporarily stops data transmission to the data destination information processing apparatus after receiving the failure sign notification to the information processing apparatus of the issuing source, and the data destination information processing apparatus Send a route change notification to each of the current active interconnect network and the current standby interconnect network at the same time, and then change the route of the current active interconnect network to the route of the other interconnect network. A path switching control unit is provided to perform subsequent data transmission after switching, and when the data destination information processing apparatus first receives a path change notification from the current standby interconnect network, it changes the path from the current active interconnect network. Wait for data reception from the current standby interconnect network until a notification is received It may include a path switching control unit for.

  The interconnect network of the present invention is an interconnect network that constitutes an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks. Issuer information is provided with a failure sign phenomenon detection unit that detects a failure sign phenomenon of data transmitted through the path of the connect network, and an interface control unit that sends a failure sign notification to the information processing apparatus that issued the data. The processing apparatus is characterized in that after receiving the failure sign notification, the route of the current active interconnect network is switched to the route of another interconnect network and subsequent data transmission is performed.

  The interconnect network of the present invention is an interconnect network that constitutes an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks. A failure predictor detection unit that detects a failure predictor phenomenon of data transmitted through the path of the connect network, and an interface control unit that transmits a failure predictor notification to the information processing apparatus that issued the data; After receiving the failure sign notification, the information processing apparatus temporarily stops data transmission to the data destination information processing apparatus, and the current active interconnect network and the current standby are directed to the data destination information processing apparatus. Route change notifications simultaneously to each other After that, the data destination information processing apparatus includes a path switching control unit that switches the path of the current active interconnect network to the path of another interconnect network and transmits subsequent data. When a route change notification is received from the connect network, a route switching control unit is provided that waits for data reception from the current standby interconnect network until a route change notification is received from the current active interconnect network.

  The present invention has an effect of reducing the system down rate when a failure occurs in the interconnect network without increasing the traffic of the interconnect network. The reason is that when a failure predictor phenomenon is detected, the route of the current active interconnect network is switched to the route of another interconnect network, and subsequent data transmission is performed.

  Next, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

  Before describing an embodiment of a duplex interconnect network computer according to the present invention, general data transmission in a network, which is a premise of the present invention, will be described with reference to FIG. Data is transmitted between the processing devices in units called packets. The packet includes a DstID field (Destination ID Field) for identifying the destination processing device of the packet, a SrcID field (Source ID Field) for identifying the source processing device of the packet, and an operation type (read operation, write operation, etc.). A Cmd field (Command Field) that represents the packet, a Len field (Length field) that represents the length of the packet, a Data field (Data Field) that stores data to be transferred, and the like.

  The packet issuer processing device issues a packet with these fields set to appropriate values. The issued packet is routed to the packet destination processing device with reference to the DstID field in the data transmission path.

  For example, when a packet is issued from the processing device 100 in FIG. 1 to the processing device 105, the processing device 100 issues the packet with the DstID field as the processing device 105 and the SrcID field as its own processing device 100. Then, the interconnect network 102 or 103 routes the packet to the packet destination processing device 105 with reference to the DstID field.

  Next, the configuration of an embodiment of the duplex interconnect network computer according to the present invention will be described. Referring to FIG. 1, there is shown a duplex interconnect network computer as an embodiment of the present invention.

  The processing devices 100, 101, 104, and 105 are information processing devices, more specifically, a central processing device or an input / output processing device, and serve as a packet issuer or a packet destination processing device. In the present invention, since there is no need to distinguish between the central processing unit and the input / output processing unit, both are simply referred to as a processing unit. All these processing apparatuses have two interface control units 110, 111, 112, and 113 and one path switching control unit 120 or 121.

  The interface control units 110, 111, 112, and 113 manage interface control with the interconnect networks 102 and 103. The path switching control unit 120 or 121 controls packet transmission control to the two interface control units 110, 111, 112, and 113 and packet reception control from the interface control units 110, 111, 112, and 113. Since the configurations and operations in the processing apparatuses 100, 101, 104, and 105 related to the present invention are all the same, only the processing apparatus 100 will be used in the following description of the processing apparatus.

  The two interconnect networks 102 and 103 are networks independent from each other, and are interconnected with the processing apparatuses 100, 101, 104, and 105.

  One of the two interconnect networks is assigned to the active system and the other one to the standby system, and data transmission / reception is performed using only the active interconnect network at normal times when no failure sign phenomenon occurs. If a failure predictor phenomenon is monitored in the active interconnect network and the predictor phenomenon is detected, it is possible to change the route to the standby interconnect network autonomously and immediately by hardware alone without the need for software processing. Is possible.

  A computer configured by assigning one of the two interconnect networks to the active system and the other to the standby system is referred to as a duplex interconnect network computer.

  Since the interconnect networks 102 and 103 have the same configuration and operation, only the interconnect network 102 will be used in the following description of the interconnect network.

  Referring to FIG. 2, an interconnect network 102 is shown as one embodiment of the present invention. The interconnect network 102 includes one switch device 200 and includes four external interfaces 150, 152, 160, and 162. These four external interfaces are connected to the processing devices 100, 101, 104, and 105.

  If the number of processing devices connected to the interconnect network 102 increases, the interconnect network 102 is configured by connecting a plurality of the switch devices 200 as described in FIGS. 6, 7, and 8. The interconnect network configurations shown in FIGS. 6, 7, and 8 are also examples, and the present invention can be applied even when other configurations are adopted.

  Referring to FIG. 3, a switch device 200 as an embodiment of the present invention is shown. The interface control units 310, 311, 312, and 313 are responsible for interface control with the outside of the switch device, and have the same configuration and the same operation as the interface control units 110 and 111 in the processing apparatus 100 illustrated in FIG. 1. Therefore, hereinafter, when describing the inside of the interface control unit, the interface control unit 310 is used.

  The complete crossbar 300 has four interface controllers 310, 311, 312, and 313 connected by a complete crossbar (complete network). That is, it is simply a collection of connection lines. A complete crossbar means that any one terminal can be connected to any other terminal on a one-to-one basis.

  Referring to FIG. 4, an embodiment of an interface control unit 310 that is an important component of the present invention is shown. The failure control unit 400 has a function of monitoring a failure sign of the interface control unit 310 itself and a failure sign of the interfaces 150 and 350.

  A failure sign is a state in which only a correctable error has occurred, and is a state in the previous stage of reaching an unrecoverable failure that has not yet reached an unrecoverable failure. As an example of the failure sign detection method, it may be determined that error detection and error correction are performed using ECC (Error Correction Code). However, the method is not limited to this. Since the specific operation of error correction is a well-known technique for those skilled in the art, detailed description thereof is omitted.

  When the failure sign monitoring function detects a failure sign phenomenon in the path from the interface 350 to the interface 150, the failure sign monitoring function transmits a packet received from the interface 350 after the detection to the interface 150 and generates a failure sign notification from the packet information. The data is transmitted to the interface 350 via the selection circuit B411. Similarly, when a failure sign phenomenon is detected in the path from the interface 150 to the interface 350, a packet received from the interface 150 after the detection is transmitted to the interface 350, a failure sign notification is generated from the packet information, and the selection circuit A410. To the interface 150.

  It should be noted that at the stage where the sign phenomenon is detected, the operation where the sign phenomenon occurs can still be continued.

  In addition, when a path change notification is received from the interface 350 after the failure sign phenomenon is detected, it is transmitted to the interface 150, a path change notification reply is generated from the path change notification, and is transmitted to the interface 350 via the selection circuit B411. . Similarly, when a route change notification is received from the interface 150, it is transmitted to the interface 350, and a route change notification reply is generated from the route change notification and transmitted to the interface 150 via the selection circuit A410.

  FIG. 10 shows a method of generating a failure sign notification from a data transmission packet in the failure sign monitoring function. The DstID field value in the data transmission packet is embedded in the SrcID field in the failure sign notification, the SrcID field value in the data transmission packet is embedded in the DstID field in the failure sign notification, and the Cmd field in the failure sign notification is A value indicating that it is present is embedded, and an appropriate value is embedded in the Len field in the failure predictor notification.

  As a result, the failure sign notification can be sent to the processing apparatus that has issued the data transmission packet. Note that when a route change notification reply is generated from the route change notification, it is generated by the same method as described above.

  Referring to FIG. 5, there is shown an embodiment of a path switching control unit 120 that is an important component of the present invention.

  The transmission control unit 500 controls packet issuance from the inside of the processing apparatus 100 to a packet 180 or 181 that is an external interface of the processing apparatus 100. The packet issuance is transmitted to one of the interfaces for each destination processing apparatus. For example, when the packet is issued from the processing device 100 to the processing device 104 in FIG. 1, the packet is transmitted to the interface 180, and when the packet is issued to the processing device 105, the packet is transmitted (transmitted) to the interface 181. An interface that issues a packet is an operational interface for the destination processing apparatus.

  This packet transmission switching is controlled by two flags called a failure sign flag and a transmission path identification flag. These two flags exist for each packet destination processing device. The transmission path identification flag indicates an active interconnect network for the packet destination processing apparatus.

  The failure sign flag is a flag used for discarding the second and subsequent failure sign notifications from the same destination processing device, and indicates a corresponding failure processing device when this flag is 0. When the notification is received, it is determined that the failure sign notification is valid, and this flag is set to a value of 1. When the failure predictor notification indicating the corresponding destination processing apparatus when the flag is 1 is received, the failure predictor notification is determined to be invalid and discarded. This is to prevent the packet issuing source device from taking any action even if the second and subsequent failure predictor notifications are received on the same route.

  When a route change notification reply, which is a reply to the route change notification, is received, this flag corresponding to the destination processing apparatus is reset to 0.

  The reception control unit 501 controls packet reception from 180 or 181 which is an external interface of the processing apparatus 100. In the reception control, for the two interfaces 180 and 181, the operational interface can be switched for each issuer processing device. For example, in the processing apparatus 100 in FIG. 1, when the issuing source processing apparatus is the processing apparatus 104, the interface 180 is the active system, and when the issuing source processing apparatus is the processing apparatus 105, the interface 181 is the operating system. be able to.

  Which interface is to be used as an operational system must be consistent with the corresponding in-issuer processor transmission control unit 500. In this packet reception path switching, it is necessary to switch the reception path while guaranteeing (matching) the packet issuing order in the packet issuing source processing apparatus and the packet receiving order in the packet receiving processing apparatus. Therefore, it has two flags called a route change flag and a reception route identification flag, and operates as follows.

  1. If the route change flag has a value of 0 and the received route identification flag indicates an interface that has received the route change notification, the route change flag is set to a value of 1.

  2. When the route change flag has a value of 0 and the received route identification flag does not indicate the interface that has received the route change notification, the route change flag is set to a value of 1, and the route change is performed at the interface that has received the route change notification. The reception of the packet issued by the processing device indicated by the SrcID field in the notification is temporarily stopped.

  3. When the route change flag has a value of 1 and the received route identification flag indicates an interface that has received the route change notification, the route change flag is reset to a value of 0 and the received route identification flag indicates another interface. Update. Furthermore, the reception of the packet issued by the corresponding processing apparatus that is temporarily stopped by another interface is resumed.

  4). When the route change flag has a value of 1 and the received route identification flag does not indicate the interface that has received the route change notification, the route change flag is reset to a value of 0, and the reception route identification flag indicates another interface. Update.

  Next, the operation in one embodiment of the present invention will be described with reference to FIG. 1, FIG. 4, FIG. 5, FIG.

  First, in describing the operation according to the present invention, the operation state (initial state) before that will be described. In the redundant interconnect network computer shown in FIG. 1, packets issued from the processing device 100 to the processing device 105 use the interconnect network 102 as an active system, and packets issued from the processing device 101 to the processing device 105 use the interconnect network 103. Assume that it is operating as an active system.

  Therefore, the transmission route identification flag corresponding to the processing device 105 in the transmission control unit 500 in the route switching control unit 120 in the processing device 100 indicates that the interconnect network 102 is an active system. Further, the reception route identification flag corresponding to the processing device 100 in the reception control unit 501 in the route switching control unit 121 in the processing device 105 also indicates that the interconnect network 102 is an active system. In this state, it is assumed that a failure sign phenomenon is detected at a location where a packet from the processing device 100 to the processing device 105 passes in the interconnect network 102.

Next, the operation of the present invention in the above operation state will be described by dividing it into the following four operations based on FIG. 11 and FIG.
1. 1. Operation in the interconnect network 102 that has detected a failure sign phenomenon 2. Operation in the processing apparatus 100 that has received the failure predictor notification 3. Operation of the failure sign phenomenon detection location in the interconnect network 102 that received the route change notification 4. Operation in the processing apparatus 100 that has received the route change notification reply Operation in Processing Device 105 Receiving Route Change Notification First, the operation in the interconnect network 102 that has detected the failure sign phenomenon will be described with reference to FIGS. 4 and 11. In the interconnect network 102, the interface control unit 310 shown in FIG. 4 through which a packet from the processing device 100 to the processing device 105 passes is used to detect a failure in the direction from the interface 150 to the interface 350 by the failure sign monitoring function in the failure control unit 400. A predictive phenomenon is detected (step 1111).

  A packet received from the interface 150 after detection is transmitted to the interface 350 (step 1112). Then, it is sent to the processing device 5 via the interface 162, and the processing device 105 receives the packet (step 1121).

  Together with these, the interface control unit 310 generates a failure sign notification from the packet information, and transmits it to the interface 150 via the selection circuit A410 (step 1113). The failure sign notification transmitted to the interface 150 is routed through the interconnect network 102 and transmitted to the processing device 100.

  Secondly, the operation in the processing apparatus 100 that has received the failure sign notification will be described with reference to FIGS. 1, 5, and 11. FIG. The failure sign notification sent from the interface 150 shown in FIG. 1 is sent to the path switching control unit 120 via the interface control unit 110 in the processing apparatus 100. In the path switching control unit 120 illustrated in FIG. 5, when a failure sign notification is received from the interface 180, the notification is sent to the transmission control unit 500 via the reception control unit 501.

  The transmission control unit 500 refers to the Cmd field in the packet received from the reception path unit 501 and knows that the received packet is a failure sign notification (step 1101). In addition, referring to the SrcID field, knowing that a failure sign phenomenon has been detected on the route to the processing device 105, the following operation is performed.

  1. A failure predictor flag corresponding to the processing device 105 is set to a value of 1. As a result, the subsequent failure sign notification in which the SrcID field indicates the processing device 105 is discarded (step 1102).

  2. Issuing packets addressed to the processing device 105 is temporarily stopped, and a route change notification is issued to both the interfaces 180 and 181 (step 1103). This route change notification uses the DstID field as the processing device 105 and the SrcID field as the processing device 100.

  3. After issuing the route change notification, the transmission route identification flag corresponding to the processing device 105 is rewritten so that the interconnect network 103 becomes the active system, and issuance of the packet addressed to the processing device 105 that has been suspended is resumed (step 1104). ).

  Third, the operation of the failure sign phenomenon detection location in the interconnect network 102 that has received the route change notification will be described with reference to FIGS. First, upon receiving a route change notification from the interface 150, the interface control unit 310 that has detected the failure sign phenomenon transmits the route change notification to the interface 350 (step 1114).

  At the same time, a route change notification reply is generated from the route change notification and transmitted to the interface 150 via the selection circuit A410 (step 1115). The route change notification reply sent to the interface 150 is routed through the interconnect network 102 and transmitted to the processing device 100.

  Fourth, the operation in the processing apparatus 100 that has received the route change notification reply will be described with reference to FIGS. 1, 5, and 11. FIG. First, in FIG. 1, the route change notification reply sent from the interface 150 is sent to the route switching control unit 120 via the interface control unit 110 in the processing apparatus 100.

  In the path switching control unit 120 illustrated in FIG. 5, when a path change notification reply is received from the interface 180, it is sent to the transmission control unit 500 via the reception control unit 501. The transmission control unit 500 refers to the Cmd field in the packet received from the reception path unit 501 and knows that the received packet is a path change notification reply (step 1105).

  Also, referring to the SrcID field, knowing that the reply is a route change notification reply to the processing device 105, resets the failure predictor flag corresponding to the processing device 105 to a value of 0, and subsequently enables the failure predictor notification to the processing device 105. It can be handled (step 1106). Note that after receiving the route change notification reply, no failure indication notification for the processing device 105 is received from the interconnect network 102 in which the failure indication phenomenon is detected.

  Fifth, the operation in the processing device 105 that has received the route change notification will be described with reference to FIGS. 1, 5, 11, and 12. First, in FIG. 1, the processing device 105 is in a state of receiving a packet issued by the processing device 100 from the interface 162.

  In this state, the route change notification issued by the processing device 100 is received from the interfaces 162 and 163. The route change notification received from the interface 162 or 163 is sent to the route switching control unit 121 via the interface control unit 112 or 113.

  Next, the operation of the path switching control unit 121 will be described with reference to FIGS. 5, 11, and 12. In FIG. 5, the path switching control unit 120 is described, but here it is read as the path switching control unit 121. In the path switching control unit 121, the reception control unit 501 receives the path change notification issued by the processing apparatus 100 from the interfaces 180 and 181 (step 1122).

  The route change notification via the interface 180 of the route switching control unit 120 of the processing device 100 is sent to the processing device 105 via the interconnect network 102, the interface 162, the interface control unit 112, and the interface 180 of the route switching control unit 120. . Similarly, a route change notification via the interface 181 of the route switching control unit 120 of the processing device 100 is sent to the processing device 105 via the interconnect network 103, the interface 163, the interface control unit 113, and the interface 181 of the route switching control unit 121. Will be sent.

  The reception control unit 501 refers to the Cmd field in the packet received from the interface 180 or 181 and knows that the received packet is a route change notification. Further, the SrcID field is referred to and it is known that this is a route change notification with the processing apparatus 100 (step 1122).

  Thereafter, it is determined which of the path change notifications from the interfaces 180 and 181 is received first, and one of the following two operations is performed depending on the destination (timing) (step 1202).

  First, the operation when the reception control unit 501 receives a route change notification from the interface 180 first (step 1202 YES) will be described. Before the route change notification is received from the interface 180, the route change flag has a value of 0, and the received route identification flag indicates the interface 180. That is, the interface 180 is an operational system.

  In this state, when the reception control unit 501 first receives a route change notification from the interface 180, the route change flag indicates the value 0 and the reception route identification flag indicates 180 that is the own interface. It is set to 1 (step 1203).

  Thereafter, when a route change notification is received from the interface 181 (step 1204), the route change flag has a value of 1 and the received route identification flag indicates 180 which is another interface. Therefore, the route change flag is reset to 0, and the reception route identification flag is updated to indicate the interface 181. That is, the interface 181 is set as an active system (step 1205).

  Thereby, the packet addressed to the processing device 105 issued from the processing device 100 after the reception path change notification can be received from the interface 181 (step 1206).

    Next, the operation when the route change notification is received first from the interface 181 (NO in step 1202) will be described. Before receiving a route change notification from the interface 181, the route change flag has a value of 0, and the received route identification flag indicates the interface 180. That is, the interface 180 is an operational system.

  When a route change notification is received first from the interface 181 in this state, the route change flag indicates the value 0, and the reception route identification flag indicates 180 which is another interface. Therefore, the route change flag is set to 1 (step 1207), and the interface 181 temporarily stops receiving the packet issued by the processing device 100 (step 1208).

  Here, when the packet reception is suspended, it is necessary to temporarily hold the packet. Specifically, for example, the packet is temporarily held in a buffer (not shown) in the interface control unit 113. Can do.

  Thereafter, if there is a packet that reaches the reception control unit 501 via the interface 180 before the route change notification is received from the interface 180, the packet is received. After receiving the route change notification from the interface 180, the packets temporarily stored in the buffer are sequentially received. In other words, when the interface 180 is delayed from the interface 181, the time axis is corrected by synchronizing both interfaces by temporarily waiting for the packet that has been advanced until the delayed route change notification arrives. .

  Further, by switching the packet reception path, it is possible to switch the reception path while guaranteeing a match between the packet issue order in the packet issuing source processing apparatus and the packet reception order in the packet receiving side processing apparatus.

  As a result, it is possible to prevent a packet to be received from being received before the path change notification is received from the interface 180 after being received from the interface 181 from being discarded.

  Thereafter, when a route change notification is received from the interface 180 (step 1209), the route change flag has a value of 1 and the received route identification flag indicates 180 that is the own interface. Therefore, the route change flag is reset to 0, and the reception route identification flag is updated to indicate the interface 181. That is, the interface 181 is set as the active system (step 1210).

  Further, the interface 181 restarts the reception of the packet issued by the processing apparatus 100 that has been suspended (step 1211). Thus, the packet addressed to the processing device 105 issued from the processing device 100 to the interconnect network 102 is processed first, and then the packet addressed to the processing device 105 issued from the processing device 100 to the interconnect network 103 is processed. be able to.

  Therefore, before the packet A addressed to the processing device 105 issued from the processing device 100 to the processing device 102 arrives before the reception path change notification issuance timing, the address addressed to the processing device 105 issued from the processing device 100 to the interconnect network 103 is reached. As a result of arrival of the packet B, it is possible to prevent a situation where a packet addressed to the processing device 105 issued to the processing device 102 issued from the processing device 100 to the interconnect network 103 cannot arrive and is discarded. That is, there is an effect that data loss can be prevented.

  Next, an operation when the reception control unit 501 receives a route change notification from the interface 180 and the interface 180 simultaneously will be described.

  Before receiving a route change notification from the interface 180 or 181, the route change flag has a value of 0, and the received route identification flag indicates the interface 180. That is, the interface 180 is an operational system.

  When a route change notification is received from the interface 180 in this state, the route change flag is set to the value 1 because the route change flag indicates the value 0 and the received route identification flag indicates 180 that is the own interface.

  At the same time (at the same timing), since the route change notification is also received from the interface 181, the route change flag indicates the value 1 and the reception route identification flag indicates 180 which is another interface. Therefore, the route change flag is reset to 0, and the reception route identification flag is updated to indicate the interface 181.

  In other words, the interface 181 is used as an operational system. Thereby, the packet addressed to the processing device 105 issued from the processing device 100 after the reception path change notification can be received from the interface 181.

  According to the operation described above, when a failure sign phenomenon is detected in the interconnect network, it is possible to change the route autonomously and immediately using only hardware without requiring software processing. For this reason, this operation can be positioned as error prevention processing, and system down due to an interconnect network failure can be reduced.

  It can be used in information processing equipment.

1 is a configuration diagram of a duplexed interconnect network system of the present invention. The block diagram of the interconnect network of this invention. The block diagram of the switch apparatus of this invention. The block diagram of the interface control part of this invention. The block diagram of the path | route switching control part of this invention. The block diagram of the other interconnect network of this invention. The block diagram of the other interconnect network of this invention. The block diagram of the other interconnect network of this invention. The block diagram of the packet transmitted in a network. The figure which shows the method of producing | generating the failure sign notification from the data transmission packet in a failure sign monitoring function. The flowchart which shows operation | movement of the Example of this invention. The flowchart which shows operation | movement of the processing apparatus which received the path | route change notification among operation | movement of the Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Processing apparatus 105 Processing apparatus 102 Interconnect network 103 Interconnect network 110 Interface control part 111 Interface control part 112 Interface control part 113 Interface control part 120 Path switching control part 121 Path switching control part 200 Switch apparatus 300 Complete crossbar 310 Interface control Unit 311 interface control unit 400 fault control unit 410 selection circuit A
411 Selection circuit B
500 Transmission control unit 501 Reception control unit

Claims (14)

An error prevention processing method for an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks,
Monitoring a failure predictor and detecting a failure predictor of data transmitted through the interconnect network path;
Transmitting a failure sign notification to the information processing apparatus that issued the data;
An error prevention processing method for an information processing system, comprising: a step of switching a route of a current active interconnect network to a route of another interconnect network after receiving the failure sign notification and performing subsequent data transmission . An error prevention processing method according to claim 1,
The issuer information processing apparatus that has received the failure sign notification temporarily stops data transmission to the data destination information processing apparatus, and the current active interconnect network and the current standby interconnect network are directed to the data destination information processing apparatus. Sending a route change notification to each simultaneously using and
In the data destination information processing apparatus, when a route change notification is received from the current standby interconnect network first, data is received from the current standby interconnect network until a route change notification is received from the current active interconnect network. An error prevention processing method for an information processing system, further comprising a step of waiting for the information processing system. An error prevention processing method according to claim 2,
After the step in which the issuing information processor uses the current active interconnect network and the current standby interconnect network to send a route change notification to each simultaneously,
8. The error prevention processing method for an information processing system according to claim 7, further comprising the step of switching the current standby interconnect network to be used as the active interconnect network and restarting data transmission. An error prevention processing method according to claim 2 or 3,
When the data destination information processing apparatus receives a route change notification from the current active interconnect network, the data destination information processing apparatus further includes a step of switching the current standby interconnect network to be used as the active interconnect network and restarting the data reception. An error prevention processing method for an information processing system. An error prevention processing method according to claim 2 or 3,
If the data destination information processing device does not receive a route change notification from the current standby interconnect network first, after receiving the route change notification from the current standby interconnect network, the current destination interconnect network An error prevention processing method for an information processing system, further comprising a step of switching to use as an interconnect network. An information processing system in which a plurality of information processing devices are connected via a plurality of interconnect networks,
In the interconnect network, a failure sign phenomenon detection unit that monitors a failure sign phenomenon and detects a failure sign phenomenon of data transmitted through the route of the interconnect network;
An interface control unit for transmitting a failure sign notification to the information processing apparatus that issued the data,
The information processing apparatus of the issuing source includes a path switching control unit that switches a path of the current active interconnect network to a path of another interconnect network after receiving the failure sign notification and performs subsequent data transmission An information processing system characterized by this. An information processing system in which a plurality of information processing devices are connected via a plurality of interconnect networks,
In the interconnect network, a failure sign phenomenon detection unit that monitors a failure sign phenomenon and detects a failure sign phenomenon of data transmitted through the route of the interconnect network;
An interface control unit for transmitting a failure sign notification to the information processing apparatus that issued the data,
After receiving the failure sign notification, the issuing information processing device temporarily stops data transmission to the data destination information processing device, and sends the current active interconnect to the data destination information processing device. Send a route change notification to each of the network and the current standby interconnect network at the same time, and then switch the route of the current active interconnect network to the route of the other interconnect network to transmit subsequent data With a route switching control unit,
When the data destination information processing apparatus previously receives a route change notification from the current standby interconnect network, data from the current standby interconnect network is received until a route change notification is received from the current active interconnect network. An information processing system comprising a path switching control unit that waits for reception. The information processing system according to claim 7,
The path switching control unit in the issuing information processing device uses the current active interconnect network and the current standby interconnect network to send route change notifications to the active standby network and the active standby interconnect network. An information processing system, wherein data transmission is resumed by switching to use as an interconnect network. An information processing system according to claim 7 or 8,
When the path switching control unit in the data destination information processing apparatus receives a path change notification from the current active interconnect network, the path switching control unit switches the current standby interconnect network to be used as the active interconnect network and receives the data. An information processing system characterized by restarting. An information processing system according to claim 7 or 8,
If the path switching control unit in the data destination information processing apparatus does not receive the path change notification from the current standby interconnect network, the path switching control unit receives the path change notification from the current standby interconnect network, An information processing system characterized by switching to use an interconnect network as an operational interconnect network. It is an information processing device that constitutes an information processing system in which information processing devices are connected via a plurality of interconnect networks that monitor the failure predictor phenomenon and detect the failure predictor phenomenon of data transmitted through the route of the interconnect network ,
An information processing apparatus comprising: a path switching control unit configured to switch a path of a current active interconnect network to a path of another interconnect network after receiving the failure sign notification and perform subsequent data transmission. An information processing apparatus according to claim 11,
After receiving the failure sign notification, the issuing information processing device temporarily stops data transmission to the data destination information processing device, and sends the current active interconnect to the data destination information processing device. Send a route change notification to each of the network and the current standby interconnect network at the same time, and then switch the route of the current active interconnect network to the route of the other interconnect network to transmit subsequent data With a route switching control unit,
If the information processing apparatus of the data destination has previously received a route change notification from the current standby interconnect network, the information processing device from the current standby interconnect network until receiving a route change notification from the current active interconnect network. An information processing apparatus comprising a path switching control unit that waits for data reception. An interconnect network constituting an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks,
The interconnect network monitors a failure predictor phenomenon and detects a failure predictor phenomenon of data transmitted through the interconnect network path,
An interface control unit for transmitting a failure sign notification to the information processing apparatus that issued the data,
The information processing apparatus of the issuing source performs subsequent data transmission by switching the path of the current active interconnect network to a path of another interconnect network after receiving the failure sign notification network. An interconnect network constituting an information processing system in which a plurality of information processing apparatuses are connected via a plurality of interconnect networks,
The interconnect network monitors a failure predictor phenomenon and detects a failure predictor phenomenon of data transmitted through the interconnect network path,
An interface control unit for transmitting a failure sign notification to the information processing apparatus that issued the data,
After receiving the failure sign notification, the issuing information processing device temporarily stops data transmission to the data destination information processing device, and sends the current active interconnect to the data destination information processing device. Send a route change notification to each of the network and the current standby interconnect network at the same time, and then switch the route of the current active interconnect network to the route of the other interconnect network to transmit subsequent data With a route switching control unit,
If the data destination information processing apparatus previously receives a route change notification from the current standby interconnect network, it receives data from the current standby interconnect network until it receives a route change notification from the current active interconnect network. An interconnect network comprising a path switching control unit that waits for

Images