JP3860587B2 - Multiprocessor device - Google Patents

Description

  The present invention relates to a multiprocessor apparatus that enables data transmission / reception processing from an application process in each processor in a loosely coupled multiprocessor system that provides a data transmission / reception function with a terminal coupled to the system for an application process.

  In a conventional loosely coupled multiprocessor system, a table defining the relationship between a terminal and a processor coupled to the system is created during system generation. Then, the table is loaded into the main storage device of each processor when the system is started, and the processor is determined by referring to the load information when the transmission system call issued by the application process is processed. For this reason, the processor to which the terminal is coupled cannot be dynamically changed, for example, for the purpose of load distribution of the processor.

  There has also conventionally been a loosely coupled multiprocessor system configured to dynamically change a processor to which a terminal is coupled. However, in this case, application processes that can perform transmission / reception processing with a terminal coupled to the processor are limited to those that operate on the processor.

  When realizing terminal use in a loosely coupled multiprocessor system, the performance of terminal coupling processing and the performance of data transmission / reception processing cannot be improved at the same time. Therefore, a trade-off according to the business operation is required. The problem is whether this choice is done by the system or by the user.

  First, as a method for the system to select one of two methods, there are a method based on terminal operation definition and a method in which an application process designates a parameter at the time of combination. However, it is not realistic in that the processing logic of the system program is complicated and that the system program is complicated because it consists of two systems.

  On the other hand, when the user chooses between the two, there is little problem because only one is selected in accordance with the operation of his / her business. As described above, it is difficult for the system to realize terminal sharing while satisfying the operational requirements of the user's business.

  Furthermore, when a failure occurs in a processor to which a terminal is coupled, there is no function that can take over the terminal to a backup processor and share the terminal taken over from each processor.

Conventionally, a technique for dynamically converting a network address (see Patent Document 1) and a technique for solving the problem of how to distribute input messages from a terminal to a plurality of hosts (see Patent Document 2) and so on.
Japanese Patent Laid-Open No. 03-254540 Japanese Patent Laid-Open No. 02-028866

However, in a loosely coupled multiprocessor system, when a terminal is coupled to the system, the decision logic of the processor to be coupled is determined by a plurality of methods such as a method defined during generation and a method selected according to the load status of each processor during operation. It is desirable to be able to use a combination of methods to improve the flexibility of business design.

  Similarly, from the viewpoint of improving the flexibility of business design, it is desirable that a terminal coupled to the system can be shared by application processes in each processor constituting the system.

  Further, when a loosely coupled multiprocessor system is used as a fault tolerant computer, it is desirable that the interface of the transmission / reception system call executed by the application process is unchanged before and after the coupled terminal is taken over from the failed processor to the backup processor.

  However, none of the conventional multiprocessor systems satisfy all the above requirements. Accordingly, an object of the present invention is to allow data transmission / reception processing to be performed from an application process in each processor, and to dynamically change a processor to which a terminal is coupled, so that the terminal is shared from the application process in each processor. An object of the present invention is to provide a multiprocessor device that can be used at all times and can be suitably used for destination management in network devices, troubleshooting, and business processing.

  In order to achieve the above object, the present invention provides a loosely coupled multiprocessor system in which a plurality of multiprocessors having an application process unit, a communication processing server process unit, and a communication access method process unit are coupled in parallel to an input / output bus. In the above, the communication processing server process unit is designated corresponding to the destination terminal when executing the system call for data transmission by the application process, and the array identification number in the terminal management table stored in the processor and the processor identification signal The processor identification number is extracted from the communication identifier that is unique within the multiprocessor system, and the transmission processing to the terminal is performed by the communication processing server process in the processor for the processor determined by the processor identification number. Send to execute By transferring the data, it is possible to perform data transmission / reception processing from the application process in each processor and to dynamically change the processor to which the terminal is coupled. And a multiprocessor device that can be operated at all times.

  FIG. 2 shows a principle configuration of a multiprocessor system according to the present invention. The system shown in FIG. 2 includes terminals 1A, 1B, 1C, and 1D and a loosely coupled multiprocessor system 6. The loosely coupled multiprocessor system 6 includes processors 9A, 9B, and 9C. Each of the processors 9A, 9B, and 9C includes an application process execution unit, a communication processing server process execution unit, and a communication access method process execution unit, respectively. Have. The terminals 1A, 1B, 1C and 1D are connected to the processors 9A, 9B and 9C via the switching network 2.

  The communication processing server process execution unit extracts a processor identification number from a communication identifier designated corresponding to a transmission destination terminal when a data transmission system call is executed by an application process, and a processor determined by the processor identification number On the other hand, the transmission data is forwarded so that the transmission processing to the terminal is executed by the communication processing server process in the processor. The communication identifier is uniquely defined in the multiprocessor system including a processor identification signal and an array element number in a terminal management table stored in the processor.

  The communication processing server process execution unit may include a communication identifier numbering unit, a communication identifier notification unit, and a processor identification number acquisition unit. The communication identifier numbering unit takes a unique number in the multiprocessor system obtained by combining the processor identification signal and the array element number in the terminal management table in the processor as a communication identifier when coupled with the processor of the terminal. Number. The communication identifier notifying unit notifies the application process of the communication identifier when notifying the processing result of terminal combination. The processor identification number acquisition unit extracts a processor identification number from a communication identifier designated corresponding to a transmission destination terminal when an application process executes a data transmission system call.

  The loosely coupled multiprocessor system 6 assigns a standby attribute to one processor and an active attribute to the remaining processors so that the standby processor operates as a backup when the active processor fails. Assigns a predetermined logical processor name, and when the standby processor backs up the failed current processor, it takes over the logical processor name of the current processor and assigns the array number of the logical processor name list as the processor identification number, so that the standby processor becomes active Even after taking over the terminal connected to the processor, the relationship between the terminal and the communication identifier may be made unchanged.

  Further, such a multiprocessor device may be used as a front-end processor system (hereinafter referred to as “FEP system”) in a network device described below. Specifically, an FEP system coupled to one or more host processors (hereinafter referred to as “hosts”) via a network and connected to a plurality of terminals via an exchange network is provided. Corresponding to a terminal specified by a communication processing application executed by the communication processing application executing means, managing a host processor for connecting the terminal to the dialing from the terminal and managing a network node in the host processor and executing the connection By referring to the network node registered as the connection destination name and changing and updating the connection destination name of the designated terminal by the communication processing application, the communication processing application and the communication access method are executed. When dialing from A network device that can flexibly and appropriately perform destination management, troubleshooting, and business processing between a switching network to which a terminal is connected and a network to which a host is coupled in accordance with the operation status of the system. .

  FIG. 1 shows the basic configuration of the network system as described above. The system shown in FIG. 1 includes terminals 1A and 1B, a switching network 2, an FEP system 3, a line / network 4 and a host 5.

  The terminals 1A and 1B are a plurality of terminals operated by a user, that is, an operator. In FIG. 1, two terminals 1A and 1B are shown. The switching network 2 connects a plurality of terminals 1A and 1B to the FEP system 3.

The FEP system 3 is coupled to a host 5 via a line / network 4. Although only one host 5 is shown in FIG. 1, a plurality of hosts 5 may be provided.
The host 5 includes a communication processing application execution unit 51 for executing a communication processing application, a communication processing server unit 52 for executing a communication processing server, and a communication access method processing unit 53 for executing processing of a communication access method. Have. Usually, the communication processing application execution unit 31 of the FEP system 3 may be smaller in scale than the communication processing application execution unit 51 of the host 5.

The FEP system 3 also includes a communication processing application execution unit 31 for executing a communication processing application, a communication processing server unit 32 for executing a communication processing server, and a communication access method processing unit 33 for executing processing of a communication access method. have. Usually FEP
The communication processing application execution unit 31 of the system 3 may be smaller in performance than the communication processing application execution unit 51 of the host 5.

  The FEP system 3 is specified by a communication processing application executed by the communication processing application execution unit 31, managing and connecting a node and a processor that connects the terminal to the dialing from the terminals 1 </ b> A and 1 </ b> B. The network node registered as the connection destination name corresponding to the terminal can be referred to, and the connection destination name of the designated terminal can be changed and updated by the communication processing application.

  The FEP system 3 may include a connection node management unit, a destination name reference unit, a destination name change unit, a connection process execution unit, and an operation information acquisition unit. In this case, typically, the destination name reference unit, the destination name change unit, and the operation information acquisition unit are provided in the communication processing application execution unit 31, and the connection node management unit and the connection processing execution unit include the communication access method. Provided in the processing unit 33. The communication processing application execution unit 31 and the communication access method processing unit 33 are coupled via the communication processing server unit 32.

  The connection node management unit holds a node management table indicating which processor and which network node the terminal is connected to for each terminal, and when the terminal is dialed from the terminal to the FEP system, It manages which network node of the processor is connected.

  The destination name reference unit is registered as the connection destination name of the designated terminal in the node management table when a communication processing application operating in the FEP system issues a system call for referring to the destination name by designating the terminal name. The network node name is notified to the communication processing application.

  The destination name changing unit is configured to specify a connection destination name of a designated terminal in the node management table when the communication processing application issues a system call for changing a destination name by designating a terminal name and a network node name. Update to node name.

  When dialing from the terminal to the FEP system, the connection processing execution unit refers to the connection node management table to determine a connection network node and transmits a connection request to the node.

  The operation information acquisition unit acquires operation information used by the communication processing application to change a connection destination name of a terminal. Further, the operation information acquisition unit refers to a host operation schedule, and the connection node management unit is based on the operation schedule of the host, and when the host is operating, the connection node of the terminal is a network node in the host, When the host is stopped, the destination may be changed as a network node in another host that can act on behalf of the host.

  In addition, the operation information acquisition unit refers to the host operation schedule, and the connection node management unit based on the operation schedule of the host, the host connection node as a network node in the host during host operation, When the host is stopped, the communication processing application executed in the FEP system may be changed as a network node connected to the terminal.

  The operation information acquisition unit may determine an operation schedule by a communication processing application operating in the FEP system communicating with a communication processing application operating in the host.

  In the multiprocessor system 6 according to the present invention, a communication processing server process dynamically allocates a communication identifier structured as a combination of two pieces of information of a processor identification number and an array number in a terminal management table at the time of terminal combination, The communication identifier uniquely defined in the multiprocessor system is notified to the application process, and when the transmission system call to the coupling terminal by the application process, the processor is determined based on the communication identifier specified for each coupling terminal. Since the transmission process is executed by determining and forwarding the transmission data to the processor, even if an abnormality occurs in the processor and the terminal is handed over to the backup processor, the universality of the interface of the transmission system call is guaranteed. The

  In the network system as described above, the FEP system 3 incorporating the communication processing application is coupled to one set or a plurality of hosts 5 and, at the time of business execution, connections from the terminals 1A and 1B to the nodes in the network are made. In this case, by simply dialing the FEP system 3 at all times, the FEP system 3 distributes connection requests to the network nodes effective at that time, and connects the terminals 1A and 1B to the effective nodes. According to such a network system, when an FEP system incorporating a communication processing application is coupled to one or a plurality of hosts, and a plurality of terminals are connected to nodes in the network at the time of business execution, By always dialing to the FEP system, the terminal distributes a connection request from the FEP system to a valid network node at that time, and connects the terminal to the valid node. A network device capable of flexibly and appropriately performing destination management, troubleshooting and business processing between a switching network to which a host processor is connected and a network to which a host processor is coupled, according to the operation status of the system Can be provided.

  According to the present invention, the communication processing server process dynamically assigns a communication identifier structured as a combination of two pieces of information of the processor identification number and the array element number in the terminal management table at the time of terminal combination, and the multiprocessor Notifying the application process of this communication identifier uniquely defined in the system, and determining a processor based on the communication identifier designated corresponding to each combined terminal when the application process sends a call to the combined terminal. Data transmission / reception processing can be performed from an application process in each processor by forwarding transmission data to the processor, and a processor to which a terminal is coupled can be dynamically changed. Share terminal It can, as a possible constantly operational, the destination management in a network device, it is possible to provide a multiprocessor system which can be suitably used in the failure dealing and business processing.

Example 1
As a specific embodiment of the network device, the network device according to the first embodiment will be described with reference to the drawings.

  FIG. 3 shows the configuration of the network device according to the first embodiment. 3 includes terminals 1A, 1B..., Switching network 2, FEP system 3, line / network 4, host 5A.

  The terminals 1A, 1B... Are connected to the FEP system 3 via the switching network 2. The FEP system 3 is coupled to one or more hosts 5A through a line / network 4 constituted by a high-speed LAN (local area network) line or a network line.

  The hosts 5A... Each have a communication processing application execution unit 51 for executing a communication processing application, a communication processing server unit 52 for executing a communication processing server process, and a communication access method process for executing a communication access method process. A portion 53 is provided.

  The FEP system 3 also includes a communication processing application execution unit 31 for executing a communication processing application, a communication processing server unit 32 for executing communication processing server processing, and a communication access method processing unit for executing processing of a communication access method. 33. In general, the communication processing application execution unit 31 of the FEP system 3 may be smaller than the communication processing application execution unit 51 of the host 5 in terms of performance.

  The communication access method processing unit 33 holds a node management table 34. In the node management table 34, a terminal name, a terminal network address, a definition designation node name (and its valid flag), and an application designation node name (and its valid flag) are registered for each terminal.

  The FEP system 3 is designated by a communication processing application executed by the communication processing application execution unit 31 and management and connection of a processor and a network node that connect the terminal to the dialing from the terminals 1A, 1B. The network node registered as the connection destination name corresponding to the selected terminal can be referred to, and the connection destination name of the designated terminal can be changed and updated by the communication processing application.

  FIG. 4 shows the FEP system 3 and its peripheral parts in the embodiment of FIG. 3 in more detail. In FIG. 4, hosts 5A and 5B and terminals 1A, 1B and 1C are shown.

  As shown in FIG. 4, the communication processing application execution unit 31 of the FEP system 3 includes an operation information acquisition unit 35, a destination node name reference unit 36, and a destination node name change unit 37. The legal processing unit 33 holds a node management table 34 and includes a connection node management unit 38 and a connection processing execution unit 39.

  The operation information acquisition unit 35 acquires operation information (operation schedule of each host) used by the communication processing application to change the connection destination name of the terminal 1A. At this time, the operation information acquisition unit 35 may acquire the operation information by referring to the operation schedule of the hosts 5A... And the communication processing application that operates in the FEP system mutually communicates with the communication processing application that operates in the host. The operation schedule may be determined and acquired by performing communication. When the communication processing application operating in the FEP system 3 generates a destination name reference system call when the communication processing application operating in the FEP system 3 generates a destination name reference system connection destination name of the specified terminal in the node management table 34. The registered network node name is notified to the communication processing application.

  The destination node name changing unit 37 specifies the connection destination name of the specified terminal in the node management table 34 when the communication processing application issues a system call for changing the destination name by specifying the terminal name and the network node name. Update to the network node name.

  The connection node management unit 38 uses the node management table 34 indicating which processor and which network node the terminal is connected to for each terminal, and when the terminal 1A... It manages which processor and which network node the terminal is connected to.

  When dialing from the terminal 1A to the FEP system 3, the connection processing execution unit 39 refers to the node management table 34, determines a connection network node, and transmits a connection request to the node.

  The operation information acquisition unit 35 refers to the operation schedule of the hosts 5A, and the connection node management unit 38 determines the connection node of the terminal when the host is operating based on the operation schedule of the hosts 5A. The network node in 5A... May be changed as a network node in another host that can substitute for the host 5A.

  Further, the operation information acquisition unit 35 refers to the operation schedule of the hosts 5A ..., and the connection node management unit 38 determines the connection node of the terminal when the host is operating based on the operation schedule of the host 5A ... 5A... May be changed so that the communication processing application executed in the FEP system 3 when the host is stopped is the network node connected to the terminal.

  Furthermore, the operation information acquisition unit 35 may determine an operation schedule by a communication processing application operating in the FEP system communicating with a communication processing application operating in the host.

  3 and FIG. 4, as described above, the hosts 5A... And the FEP system 3 are connected by a line / network 4 composed of, for example, a high-speed LAN line. The terminals 1A are connected to the FEP system 3 via the exchange network 2 and perform file transfer, for example, to the hosts 5A.

  Further, the communication processing application executed by the communication processing application execution unit 31 of the FEP system 3 communicates with the communication processing application executed by the hosts 5A, and acquires the operating status of the host machine.

  The communication access method processing unit 33 operating in the FEP system 3 expands the network definition in which the connection node name is described for each terminal, and holds it as a node management table 34 in the main storage area of the FEP system 3. This node management table 34 manages two destinations, that is, a node name specified by the network definition and a node name specified by the communication processing application. When the communication processing application has not issued a request for changing the node name, when dialing from the terminal 1A..., It is distributed to the node specified by the network definition stored in the node management table 34.

  On the other hand, when the communication processing application issues a node name change request, this name is reflected in the communication processing application designation field in the node management table 34. Then, when dialing from the terminal 1A, the name specified by the communication processing application is stored in the node management table 34. Therefore, this is preferentially adopted and distributed to the node.

  The communication processing application in the FEP system 3 changes the destination node name to the FEP system 3 itself before the host 5A stops. Then, the host node name is changed to a node in the host 5A... When the host 5A.

  When the destination node is the FEP system 3 itself, the communication processing application in the FEP system 3 temporarily stores files transferred from the terminals 1A. Then, when the host machine is in an operating state and receives a transfer instruction, it is transferred to the hosts 5A.

  If the destination node is the FEP system 3 itself and the communication processing application in the FEP system 3 can process the data transferred from the terminal 1A, instead of the host, A file transferred from the terminal 1A... May be processed instead of the host machine.

  The route P1 shown in FIG. 3 shows the mutual communication for determining the operating state, the route P2 shows that the destination switching instruction issued according to the change in the operating status is reflected in the terminal management table, and the route P3 and P4 refer to the terminal management table at the time of connection, and show how they are actually connected to valid destinations.

  5, 6, 7, 8, and 9 show the connection node management unit 38, the destination node name reference unit 36, the destination node name change unit 37, the connection process execution unit 39, and the operation information acquisition unit 35. It is a flowchart which shows operation | movement in detail.

  The processing of the connection node management unit 38 shown in FIG. 5 is performed as follows. When the process is started, it is determined whether or not to perform the IPL (Initial Program Loader) process (step S11). If “true”, the exchange terminal definition information is read from the system file (step S12). Next, definition check processing such as syntax check and semantics of exchange terminal definition information is performed (step S13), and the result of this check processing is determined (step S14). Node management table areas corresponding to the number of exchange terminals are acquired in the main storage area 3, and the definition information read out earlier is reflected (step S15), and the process ends.

  If it is determined in step S11 that it is “false”, it is determined whether it is a destination node name reference or change request (step S16). If it is determined that it is “true”, a request is made. The node management table 34 is searched using the terminal name as a key (step S17). Next, it is determined whether or not the search is successful (step S18). If “true”, it is determined whether or not it is a reference request (step S19). If it is determined in step S19 that it is “true”, it is determined whether or not the application specified node is valid (step S20). If it is determined that it is “true”, the application specified node is determined. The name is stored in the output information area (step S21), and the process is terminated.

  If it is determined in step S20 that it is “false”, the definition designation node name is stored in the output information area (step S22), and the process is terminated. If it is determined in step S19 that it is “false”, that is, it is a change request, it is determined whether or not the change node name is binary zero (step S23). It is determined whether the node name is valid (step S24). If it is determined as “true” in step S24, the application designation valid flag in the node management table is turned on (step S25), and the application designation node name is stored in the node management table 34 (step S26). ), The process is terminated.

If it is determined as “false” in step S23, the application designation valid flag in the node management table 34 is turned off (step S27), and the process ends.
If it is determined as “false” in steps S14, S16, S18, and S24, it is an error, so that the return value is abnormal (step S28), and the process is terminated. The processing of the destination node name reference unit 36 shown in FIG. 6 is performed as follows.

When the process is started, the connection node management unit 38 is requested to refer to the destination node name (step S31), and it is determined whether or not the normal state is restored (step S32). The communication processing application is notified (step S33), and the processing is terminated. If it is determined in step S32 that it is “false”, a detailed code for abnormal recovery is added to notify the application of abnormal termination (step S34), and the process is terminated.

  The processing of the destination node name changing unit 37 shown in FIG. 7 is performed as follows. When the process is started, the connection node management unit 38 is requested to change the destination node name (step S41), and it is determined whether or not the normal state is restored (step S42). The normal end is notified (step S43), and the process ends. If it is determined in step S42 that it is “false”, a detailed code for abnormal recovery is added to notify the application of abnormal termination (step S44), and the process is terminated.

  The process of the connection process execution part 39 shown in FIG. 8 is performed as follows. When the process is started, the node management table is searched using the address of the terminal at the time of dialing reception as a key (step S51), and it is determined whether or not the application designated node is valid (step S52). If there is, the address is obtained from the designated node name, a connection request is notified to the node (step S53), and the process is terminated. If it is determined in step S52 that it is “false”, an address is obtained from the definition node name, a connection request is notified to the node (step S54), and the process ends.

  The processing of the operation information acquisition unit 35 shown in FIG. 9 is performed as follows. When the process is started, it is determined whether or not it is a connection node change opportunity (step S61). If “true”, the connection node name is acquired from the operation information (step S62). Further, the destination node name changing unit 37 obtains the connection destination node name (step S63), and it is determined whether or not the changing process is normally completed (step S64). If “true”, the process is terminated. If it is determined as “false” in step S64, an abnormality process is performed (step S65), and the process ends. If it is determined in step S61 that it is “false”, the processing is ended as it is.

  In a conventional network device, when dialing from a terminal via a switching network and connecting to a node in the network, the connection destination of the terminal is fixedly determined by definition, or one of the following How to switch the combined destination of the terminal using the method. (A) The terminal operator dials a predetermined connection node in the network in consideration of the operation schedule. (B) The destination is dynamically switched by incorporating a program for switching and dialing the destination to be connected based on a predetermined operation schedule in the terminal.

  In the operation of the task of dialing from a terminal connected via the exchange network and switching the connection node, in the method (a), it is necessary for the terminal operator to dial in consideration of the operation schedule. Such a method has a drawback that the burden on the terminal operator becomes heavy when the operation schedule becomes complicated.

  In the method (b), it is necessary to incorporate a destination switching program in the terminal. In this case, since a program for switching destinations must be installed in all terminals via the switching network used for business operations, there is a problem that costs increase when dealing with changes or changes in the operation schedule. .

Example 2
Next, as a specific embodiment of the multiprocessor device, a multiprocessor device according to a second embodiment will be described with reference to the drawings.

<System configuration>
FIG. 10 shows the configuration of the multiprocessor device according to the second embodiment.
The multiprocessor device shown in FIG. 10 includes terminals 1A, 1B, 1C, and 1D and a loosely coupled multiprocessor system 6. The loosely coupled multiprocessor system 6 includes line adapters 7A and 7B, an IO (input / output) bus 8, and processors 9A, 9B, and 9C.

  The line adapter 7A is connected to the terminals 1A and 1B, and the line adapter 7B is connected to the terminals 1C and 1D. These line adapters 7A and 7B are connected to an IO bus 8, and processors 10A, 9B and 9C are connected to the IO bus 8.

  Each processor 9A, 9B, 9C includes application process execution units 91A, 91B, 91C, communication processing server process execution units 92A, 92B, 92C, and communication access method process execution units 93A, 93B, 93C, respectively. The communication processing server process execution units 92A, 92B, and 92C extract a processor identification number from a communication identifier designated corresponding to a destination terminal when executing a system call for data transmission by an application process. The transmission data is forwarded to the processor determined by the above in order to cause the communication processing server process in the processor to execute transmission processing to the terminal. The communication identifier includes a processor identification signal and an array element number in a terminal management table stored in the processor, and is uniquely defined in the multiprocessor system.

  The communication processing server process execution units 92A, 92B, and 92C include communication identifier numbering units 94A, 94B, and 94C, communication identifier notification units 95A, 95B, and 95C, and processor identification number acquisition units 96A, 96B, and 96C, respectively. Yes.

  The communication identifier numbering unit 94A, 94B, 94C is a unique number in the multiprocessor system obtained by combining the processor identification signal and the array element number in the terminal management table in the processor at the time of connection with the processor of the terminal. Numbered as a communication identifier.

  The communication identifier notification units 95A, 95B, and 95C notify the application process of the communication identifier when notifying the processing result of the terminal connection to the application process. The processor identification number acquisition units 96A, 96B, and 96C take out the processor identification number from the communication identifier designated corresponding to the transmission destination terminal when the system call for data transmission by the application process is executed.

  The IO bus 8 is used for data input / output between the processors 9A to 9C and the line adapters 7A and 7B. That is, FIG. 10 shows a multiprocessor system 6 in which a plurality of processors 9A, 9B, 9C and (LAN) line adapters 7A, 7B are loosely coupled by an IO bus 8. The line adapters 7A and 7B can be accessed from any processors 9A, 9B and 9C.

  Furthermore, on each processor 9A, 9B, 9C, application process execution units 91A, 91B, 91C, communication server process execution units 92A, 92B, 92C, and communication access method process execution units 93A, 93B, 93C respectively correspond. Provided. A transmission / reception system call generated by each application process is received by the communication processing server process on the processor. Then, depending on the position of the processor to which the terminal is coupled, the processor itself decides whether to process using the communication access method process or to request the communication processing server process on another processor. Perform branch processing.

Connections PA, PC, and PD indicated by broken lines in FIG. 10 indicate examples of logical connection relationships, and such connection relationships are referred to as connections, sessions, communication paths, or the like. The logical connection relationship between the terminals 1A to 1D and the processors 9A to 9C through the connections PA, PC, and PD is as follows. The terminal 1A is connected to the processor 9A, the terminal 1B is not connected to the system, and the terminal 1C Coupled to processor 9C, terminal 1D is coupled to processor 9A.

  A data flow F1 shows a flow of data when the terminal 1A is combined with the processor 9A. A data flow F2 shows a case where an application process in the processor 9B executes a data transmission system call to the terminal 1C. Shows the data flow. The logical processor names of the processors 9A, 9B, and 9C are PM01, PM02, and PM03, respectively.

<Numbering of communication identifier>
The communication identifier numbering is based on the call request from the application process or the application from the terminal, as shown in FIG. 11 and FIG. It is executed in a terminal connection process triggered by an incoming call notification to the process.

  FIG. 13 shows a flowchart of the communication identifier numbering process. The communication processing server process searches for a free entry from the terminal management table list in its own processor before requesting the connection processing to the communication access method process (step S71). Then, it is determined whether or not the search is successful (step S72). If “true”, the array number of the searched entry is acquired (step S73), and the array number of the entry and the logical processor name list are obtained. After assigning a number combined with the array number of its own processor as a communication identifier (step S74), it is stored in the terminal management table (step S75), and the process is terminated.

  If it is determined as “false” in step S72, that is, if there is no empty entry in the terminal management table list, the terminal management table list is expanded to acquire the first entry of the expanded portion (step S76), and step S74. Perform the process.

The communication identifiers thus numbered are designated as user request identifiers when making a connection request to the communication access method, as shown in FIGS.
<Notification of communication identifier>
As shown in FIG. 11 and FIG. 12, the result of the joining process is notified from the communication access method process to the communication processing server process asynchronously with the joining request. Further, the communication processing server process notifies the application process of the combined processing result together with the communication identifier.

  FIG. 14 shows a flowchart of the communication identifier notification process. At this time, the communication processing server process searches the terminal management table using the communication identifier added and notified to the processing result as a key to determine whether or not the connection is successful (step S81). If the determination result in step S81 is “true”, that is, the combined result is normal, the communication identifier in the terminal management table is stored in the communication identifier in the processing result data (step S82), and the communication identifier is stored in the application process. At the same time, the processing result data is notified (step S83), the processing is normally completed, and the processing is terminated. If it is determined as “false” in step S81, that is, if there is an abnormality, the communication identifier in the terminal management table is cleared to zero in order to invalidate the communication identifier in the processing result data (step S84), and the terminal management table is listed. (Step S85), the cause of the error is added to the application process and the processing result data is notified (step S83), the processing is terminated abnormally and the processing is terminated.

<Acquisition of processor identification number>
As shown in the data flow at the time of data transmission in FIG. 15, the application process designates a communication identifier corresponding to the combined terminal and acquires a logical processor identification number when executing a transmission system call.

  FIG. 16 shows a flowchart of logical processor identification number acquisition processing. The communication processing server process that has received the transmission request checks the validity of the processor identification number of the designated communication identifier (step S91). When step S91 is “true”, that is, when the logical processor identification number of the communication identifier is normal, the position of the processor is determined from the processor identification number in the communication identifier. At this time, it is checked whether or not the processor identification number is equal to the value of its own processor (step S92), and if it is “true”, that is, if the processor is the processor itself, a transmission request is sent to the communication access method process (step S93), the process is terminated. If step S92 is “false”, that is, if the processor is another processor, the communication processing server process in the other processor is requested to perform transmission processing by forwarding transmission data (step S94). finish. If “false” in step S91, that is, if the processor identification number is invalid, this is notified to the application process (step S95), and the process is terminated.

FIG. 17 schematically illustrates the relationship between the logical processor name list, the terminal management table list of the processors 9A and 9C, and the management table of the terminals 1A, 1C, and 1D.
The logical processor name list includes a plurality of arrays each storing a logical processor name, physical processor identification information, and other management information. In FIG. 17, array 0 is an unused state, and array 1 includes PM01 is stored as the logical processor name, and information indicating the processor 9A is stored as the physical processor identification information. The array 2 stores information indicating PM02 as the logical processor name and information indicating the processor 9B as the physical processor identification information. , Array 3 stores information indicating the logical processor name PM03 and physical processor identification information indicating the processor 9C.

In the terminal management table list of the processor 9A, the arrays 0, 3 and 4 are unused, the array 1 stores the management table of the terminal 1A, and the array 2 stores the management table of the terminal 1D. The management table of the terminal 1A stores a communication identifier “01000001” and other management information. “01” of the communication identifier “01 000001” is the array number of the logical processor name list, and “000001” is the array number of the terminal management table list. In the management table of the terminal 1D, a communication identifier “01 000002” and other management information are stored. “01” of the communication identifier “01 000002” is the array number of the logical processor name list, and “000002” is the array number of the terminal management table list.

  In the terminal management table list of the processor 9C, the arrays 0 and 2 to 4 are unused, and the array 1 stores the management table of the terminal 1C. In the management table of the terminal 1C, a communication identifier “03 000001” and other management information are stored. “03” of the communication identifier “03 000001” is the array number of the logical processor name list, and “000001” is the array number of the terminal management table list.

<Taking over the terminal when an error occurs in the processor>
In such a system, backup can be effectively performed when an abnormality occurs in a certain processor.

In other words, the above-described loosely coupled multiprocessor system 6 assigns a standby attribute to one processor and an active attribute to the remaining processors, and the standby processor operates as a backup when the active processor fails. Assigns a predetermined logical processor name, and when the standby processor backs up the failed current processor, the standby processor takes over the logical processor name of the current processor and assigns the array number of the logical processor name list as the processor identification number. Even after taking over the terminal connected to the current processor, the relationship between the terminal and the communication identifier can be made unchanged.

  When an abnormality occurs in a certain processor, the logical relationship between the process in the adapter and the communication access method process in the backup processor is taken over. Furthermore, the logical relationship between the communication access method process and the communication processing server process is also taken over. As a result, the terminal is taken over from the processor in which the abnormality has occurred to the backup processor.

  Here, the communication processing server process that has recognized that an abnormality has occurred in a certain processor associates an entry corresponding to the logical processor name of the processor that has ended abnormally with a backup. The entry corresponding to the logical processor name that has been assigned to the backup processor until then manages the processor as a blocked state. As a result, the array number corresponding to the logical processor name does not change, so the relationship between the logical processor name and the terminal coupled to the logical processor does not change, and the application processor assembles the process in consideration of the logical processor, Even if an abnormality occurs in a certain processor, the processing can be continued.

  An example of specific processing when a failure occurs in the processor 9A in this case will be described with reference to FIGS. In a state before the failure of the processor 9A, as shown in FIG. 18, the terminal 1A is connected to the processor 9A via the connection PA, and the terminal 1D is similarly connected to the processor 9A via the connection PD. .

  As shown in FIG. 20, the logical processor name list at this time is composed of a plurality of arrays each storing a logical processor name, physical processor identification information, processor status, active standby attribute, and other management information. . That is, here, as management information, the state of the processor and the active standby attribute are added to the example of FIG. In FIG. 20, array 0 is an unused state, and array 1 shows PM01 as the logical processor name, processor 9A as the physical processor identification information, active state as the processor state, and active as the active standby attribute. Each piece of information is stored. Similarly, the array 2 stores PM02 as the logical processor name, the processor 9B as the physical processor identification information, the active state as the processor state, and the information indicating the current state as the active standby attribute. The array 3 stores the logical processor name. PM03 as the name, processor 9C as the physical processor identification information, active state as the processor state, and information indicating standby as the active standby attribute are stored. That is, in this case, the processor 9C is a standby processor used for backup.

  When a failure occurs in the processor 9A in this state, as shown in FIG. 21, the information stored in the array 1 includes the PM01 as the logical processor name, the processor 9C as the physical processor identification information, and the active state as the processor state. The active standby attribute is changed to information indicating the current service. The array 2 still stores PM02 as the logical processor name, the processor 9B as the physical processor identification information, the active state as the processor state, and the information indicating the current use as the active standby attribute. The array 3 is the logical processor name. PM03, there is no physical processor identification information, the processor status is changed to the blocked status, and the active standby attribute is not set.

As a result, the logical processor number of the processor 9A in which the failure has occurred is the processor 9
As shown in FIG. 19, the processor 9A takes over the function of the processor 9A in which the failure has occurred, the terminal 1A is connected to the processor 9C via the connection PA, and the terminal 1D is connected to the connection PD. To the processor 9C. At this time, the processor 9A is in a failure occurrence state, but in the case of a software failure, re-IPL is performed, the processor is restarted, and the processor is restored. When a hardware failure occurs, the operator is prompted to replace the processor module, and the failure is recovered.

  When the failure of the processor 9A is recovered in this way, as shown in FIG. 22, the array 3 has a logical processor name of PM03, a physical processor identification information of the processor 9A, and a processor state of an active state. Information indicating standby is stored as the active standby attribute, and the processor 9A enters the standby state as the next backup standby processor.

  When an application process executes a system call for data transmission to a terminal coupled to the system, it is necessary to specify a communication identifier corresponding to the coupled terminal. By the way, the communication identifier is notified only to the application process that executed the combining process. Therefore, other application processes, particularly application processes in the PM to which the terminal is not coupled, need to acquire a communication identifier in some way. There are various methods for acquiring the communication identifier in this case. For example, the application process that has executed the combining process may broadcast the communication identifier to all processes. In this case, broadcasting is necessary to invalidate the communication identifier even when the terminal is disconnected.

  In a conventional loosely coupled multiprocessor system, a table defining the relationship between a terminal and a processor coupled to the system is created during system generation. Then, the table is loaded into the main storage device of each processor when the system is started, and the processor is determined by referring to the load information when the transmission system call issued by the application process is processed. For this reason, the processor to which the terminal is coupled cannot be dynamically changed, for example, for the purpose of load distribution of the processor.

  There has also conventionally been a loosely coupled multiprocessor system configured to dynamically change a processor to which a terminal is coupled. However, in this case, application processes that can perform transmission / reception processing with a terminal coupled to the processor are limited to those that operate on the processor.

  When realizing terminal use in a loosely coupled multiprocessor system, the performance of terminal coupling processing and the performance of data transmission / reception processing cannot be improved at the same time. Therefore, a trade-off according to the business operation is required. The problem is whether this choice is done by the system or by the user.

  First, as a method for the system to select one of two methods, there are a method based on terminal operation definition and a method in which an application process designates a parameter at the time of combination. However, it is not realistic in that the processing logic of the system program is complicated and that the system program is complicated because it consists of two systems.

  On the other hand, when the user chooses between the two, there is little problem because only one is selected in accordance with the operation of his / her business. As described above, it is difficult for the system to realize terminal sharing while satisfying the operational requirements of the user's business.

Furthermore, when a failure occurs in a processor to which a terminal is coupled, there is no function that can take over the terminal to a backup processor and share the terminal taken over from each processor.

  However, in a loosely coupled multiprocessor system, when a terminal is coupled to the system, the decision logic of the processor to be coupled is determined by a plurality of methods such as a method defined during generation and a method selected according to the load status of each processor during operation. It is desirable to be able to use a combination of methods to improve the flexibility of business design.

  Similarly, from the viewpoint of improving the flexibility of business design, it is desirable that a terminal coupled to the system can be shared by application processes in each processor constituting the system.

  Further, when a loosely coupled multiprocessor system is used as a fault tolerant computer, it is desirable that the interface of the transmission / reception system call executed by the application process is unchanged before and after the coupled terminal is taken over from the failed processor to the backup processor.

  None of the conventional multiprocessor systems satisfy all of the above-mentioned requirements, but the multiprocessor device of the second embodiment described above can satisfy all of these requirements.

  Therefore, in the multiprocessor device of this embodiment, the communication identifier is structured as described above, and the application process extracts the processor identification number from the communication identifier specified for the combined terminal at the time of the system call, so that the terminal is combined. It is possible to identify the processor that is present. Furthermore, by assigning an array number in the logical processor name list as the processor number, even if an abnormality occurs in the processor and the terminal is handed over to the backup processor, the universality of the interface of the transmission system call is guaranteed. The

Furthermore, the multiprocessor system configured as described above can be used as the FEP system in the network apparatus of the first embodiment described above.
In this case, specifically, for example, the FEP system 3 shown in FIG. 3 is configured by the loosely coupled multiprocessor system 6 shown in FIG. 10, and the network device of the first embodiment can be configured more effectively. it can.

It is a principle figure which shows the structure of the network device based on this invention. It is a principle figure which shows the structure of the multiprocessor apparatus which concerns on this invention. It is a block diagram which shows the structure of the network apparatus which concerns on Example 1 of this invention. FIG. 4 is a block diagram illustrating in detail a configuration of a main part of the network device in FIG. 3. 5 is a flowchart for explaining an operation of a connection node management unit of the network device of FIG. 6 is a flowchart for explaining an operation of a destination node name reference unit of the network device of FIG. 6 is a flowchart for explaining the operation of a destination node name changing unit of the network device of FIG. 5 is a flowchart for explaining an operation of a connection processing execution unit of the network device of FIG. 5 is a flowchart for explaining an operation of an operation information acquisition unit of the network device of FIG. It is a block diagram which shows the structure of the multiprocessor apparatus which concerns on Example 2 of this invention. It is a data flow figure at the time of the terminal coupling | bonding by the calling for demonstrating operation | movement of the multiprocessor apparatus of FIG. 10 typically. It is a data flow figure at the time of the terminal coupling | bonding by the incoming call for demonstrating operation | movement of the multiprocessor apparatus of FIG. 10 typically. 11 is a flowchart for explaining an operation of a communication identifier numbering process of the multiprocessor device of FIG. 10. 11 is a flowchart for explaining an operation of communication identifier notification processing of the multiprocessor device of FIG. 10. It is a data flow figure at the time of the data transmission for demonstrating typically operation | movement of the multiprocessor apparatus of FIG. 11 is a flowchart for explaining an operation of a logical processor identification number acquisition process of the multiprocessor device of FIG. 10. It is a figure for demonstrating typically the table used for the multiprocessor apparatus of FIG. FIG. 11 is a schematic diagram for explaining an operation when a processor failure occurs in the multiprocessor device of FIG. 10. FIG. 11 is a schematic diagram for explaining an operation at the time of processor failure backup of the multiprocessor device of FIG. 10. FIG. 11 is a schematic diagram of a logical processor name list before the occurrence of a failure for explaining an operation when a processor failure occurs in the multiprocessor device of FIG. 10. FIG. 11 is a schematic diagram of a logical processor name list when a failure occurs for explaining an operation when a processor failure occurs in the multiprocessor device of FIG. 10. FIG. 11 is a schematic diagram of a logical processor name list at the time of failure recovery for explaining an operation when a processor failure occurs in the multiprocessor device of FIG. 10.

Explanation of symbols

1A, 1B, 1C, 1D Terminal 2 Switching network 3 Front-end processor (FEP) system 4 Network 5, 5A, 5B Host (processor)
6 Loosely coupled multiprocessor system 7A, 7B Line adapter 8 Input / output (IO) bus 9A, 9B, 9C Processor 31, 51 Communication processing application execution unit 32, 52 Communication processing server unit 33, 53 Communication access method processing unit 34 Node management Table 35 Operation information acquisition unit 36 Destination node name reference unit 37 Destination node name change unit 38 Connection node management unit 39 Connection processing execution unit 91A, 91B, 91C Application process execution unit 92A, 92B, 92C Communication processing server process execution unit 93A, 93B, 93C Communication access method process execution part

Claims (3)

A multiprocessor having an application process means for executing an application process, a communication processing server process means for executing a communication processing server process, and a communication access method process means for executing a communication access method process is provided on an input / output bus. In a loosely coupled multiprocessor system coupled in parallel,
The communication processing server process means executes a system call for data transmission uniquely defined in a multiprocessor system including a processor identification signal and an array element number in a terminal management table stored in the processor and the application process The processor identification number is extracted from the communication identifier sometimes designated in correspondence with the destination terminal, and the transmission processing to the terminal is performed by the communication processing server process in the processor for the processor determined by the processor identification number. A multiprocessor device comprising means for forwarding transmission data to be executed. The communication processing server process means includes
A communication identifier numbering means for assigning, as a communication identifier, a unique number in the multiprocessor system obtained by combining the processor identification signal and the array element number in the terminal management table in the processor when combined with the processor of the terminal; ,
When the processing result of the terminal binding is normal to the application process, the communication identifier numbered by the communication identifier numbering means is notified together with the processing result of the terminal binding, and the processing result of the terminal binding is abnormal A communication identifier notification means for invalidating the communication identifier numbered by the communication identifier numbering means and notifying the processing result of the terminal binding ;
2. The processor identification number acquisition means for extracting a processor identification number from a communication identifier designated corresponding to a destination terminal when executing a system call for data transmission by an application process. Multiprocessor device. The loosely coupled multiprocessor system includes:
A multi-processor system that assigns a standby attribute to one processor and an active attribute to the remaining processors, and the standby processor operates as a backup in the event of a failure of the active processor,
Each processor then applies a predetermined logical processor name, standby processor by the take over Guko logical processor name and the processor identification numbers of the working processor when backing up the fault current processor, the standby processor is coupled to the working processor The multiprocessor device according to claim 1, wherein the relationship between the terminal and the communication identifier remains unchanged even after the terminal that has been used is taken over.

Images