JP2015082197A - Computer system and computer system processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a function to resume processing of a copy destination computer synchronously with a copy source computer in a computer system configured by a plurality of computers different in OS, CPU, and the like.SOLUTION: A computer system includes: a plurality of processing computers 11 and 12 performing processing on data in different forms, respectively; and a management computer. The processing computers 11 and 12 include: processing units 211 and 221 each generating data indicating a midway result of the processing and executing next processing; storage units 220 and 230 each storing therein the data on the midway result; conversion units 212 and 222 each converting the stored data to data in a preset form; transfer units 215 and 225 each outputting the data in the preset form to the other processing computer 11 or 12; and communication units each outputting a result of the processing to the management computer, respectively. The management computer compares the results of the processing input from the processing computers with each other, and outputs the result of the processing having most matching items to the computer system.

Description

  The present invention relates to a computer system.

  Ensuring the reliability of computers is an important issue because the use of computers has progressed to control social infrastructure, and computer failures have a significant impact. A computer system that covers a failure by a plurality of computers and continues processing by exchanging the computer in which the failure has occurred is also widespread.

  For example, in Patent Document 1, in an FTC system (Fault Tolerant Computer System), in order to match the memory data of the operating computer with another computer while the operating computer is not stopped and the processing is continued, Take a snapshot of all the data in the memory at a certain point in the active copy source computer, write the snapshot data to the memory of the copy destination computer that has been replaced due to a failure, and perform the processing of the copy destination computer. It is disclosed to resume in synchronization with a computer.

JP 2012-208964 A

  According to the technique disclosed in Patent Document 1, an FTC system can be configured by a plurality of computers having the same configuration. However, on multiple computers with different configurations, such as OS (Operating System) and CPU (Central Processing Unit), even if the snapshot data is written to the memory of the copy destination computer, the copy source computer has different OS and CPU. Unlike the data format of the copy destination computer, the data format specified by the endian used, the number of bits, the structure configuration, and the like may not be able to resume the processing of the copy destination computer.

  An object of the present invention is to copy data in a computer system configured by a plurality of computers using different OSs or CPUs, respectively, and resume the processing of the copy destination computer in synchronization with the copy source computer. .

A computer system according to the present invention is a computer system composed of a plurality of processing computers that process data of different formats and a management computer, and each of the processing computers generates data as an intermediate result of the processing. A processing unit that executes the next process based on the intermediate result data; a storage unit that stores the intermediate result data in a format that can be processed by the processing unit; and the stored data is predetermined. Converting the data into a format, converting the data in the predetermined format into data in a format stored in the storage unit, and outputting the data in the predetermined format to the other processing computer, A transfer unit that inputs from the other processing computer, a communication unit that outputs a result of the processing to the management computer,
The first processing computer added to the computer system inputs the data in the predetermined format from the second processing computer already existing in the computer system by the transfer unit, and the input The converted data is converted into data in a format that can be processed by the processing unit in the conversion unit, the converted data is stored in the storage unit, the next processing is executed based on the stored data, and the execution The result of the processing including the output is output to the management computer, and the management computer compares the result of the processing input from each of the plurality of processing computers including the first processing computer and the second processing computer. And outputting the result of the processing with the largest number of matches to the outside of the computer system.

  The present invention is also grasped as a processing method of a computer system.

  According to the present invention, even if a computer system is configured with a plurality of processing computers that process different types of data by using different OSs or CPUs, the data is converted between the plurality of processing computers. By copying, it is possible to resume in synchronization.

It is a figure which shows the example of a structure of the FTC system containing a some calculation apparatus. It is a figure which shows the example of the module structure of a calculation apparatus. It is a figure which shows the example of the processing sequence until it restarts an application. It is a figure which shows the example of the variable table registered with a memory allocation process flow. It is a figure which shows the example of a page table. It is a figure which shows the example of a processing sequence of memory update. It is a figure which shows the example of the process sequence of a data update detection and data format conversion. It is a figure which shows the example of a processing sequence of update data transfer and update data writing. It is a figure which shows the example of the processing sequence of the data access after application resumption. It is a figure which shows another example of the processing sequence until it restarts an application. It is a figure which shows another example of the process sequence of data update detection and data format conversion.

  Hereinafter, preferred embodiments of the FTC system will be described with reference to the drawings.

  FIG. 1 is a diagram showing an example of the configuration of an FTC system that is composed of a plurality of computing devices adopting different CPUs and OSs, and by executing the same processing in parallel in each of the constituting computing devices. In the following, the FTC system 10 including two computing devices A 11 and B 12 will be described as an example mainly to explain data transfer. However, the number of computing devices is increased in accordance with system requirements. If the majority of the processing results of each computing device is taken, three or more may be used. The main components of the FTC system 10 are a computing device A 11 and a computing device B 12, which are interconnected via the internal network 13, and a gateway server 14 which is connected to the external network 15 and relays communication with the external system 16. is there.

  The FTC system 10 provides a service by receiving a request from the external system 16 via the external network 15, performing processing for the request, and returning a processing result to the external system 16. Inside the FTC system 10, the gateway server 14 receives a request from the external system 16, and transfers the request to the computing device A 11 and the computing device B 12 via the internal network 13. Receiving the request, the computing device A 11 and the computing device B 12 each execute processing for the request, and transmit the processing result to the gateway server 14 via the internal network 13. The gateway server 14 creates a response to the requesting external system 16 from the processing results received from the computing devices A 11 and B 12 and transmits the response to the external system 16 via the external network 15. At this time, the gateway server 14 performs comparison and collation, correctness determination, and the like on the processing results received from the calculation device A11 and the calculation device B12. Here, when there are three or more computing devices, the most probable processing result may be transmitted to the external system 16 by majority vote or the like.

  Here, for example, the FTC system 10 is a control system that controls the external system 16, the request from the external system 16 is measured input data, and the processing result that is a response to the external system 16 is the control data for the external system 16. It may be. Further, since the gateway server 14 operates on a plurality of computing devices as described above, it is also a management computer for the plurality of computing devices.

  The main hardware configurations of the computing device A 11 and the computing device B 12 are a CPU, a storage device (memory, hard disk), a communication device, and the like. The storage device stores a program and data of an application that executes processing in response to a processing request, a library and an OS that provide basic functions necessary for the application to operate. The processing result by the application is transmitted to the gateway server 14 via the internal network 13 by the communication device.

  In the FTC system 10, different types of CPUs and OSs may be applied to each of the computing device A 11 and the computing device B 12. The application executes the same processing content in the same processing order in each of the computing device A 11 and the computing device B 12. By processing in parallel with multiple computing devices in response to the same request from the external system 16, even if a failure occurs in one or more computing devices, the remaining computing devices continue to operate as a system Normal service provision can be continued without interruption, improving the fault tolerance and reliability of the system.

  Further, in the FTC system 10, after one or more computing devices are removed from the parallel processing between the computing devices, the other computing devices that are continuing the parallel processing to the computing device that has undergone repair, replacement, etc. By copying and matching application data, it has a function of re-computing a computing device to recover the parallel processing state of all original computing devices.

  FIG. 2 is a diagram showing an example of a functional module configuration of a computing device constituting the FTC system. Here, an example of the calculation device A 11 and the calculation device B 12 is shown for the purpose of explaining data transfer later, but the configuration is the same in three or more calculation devices. The components of the computing device A 11 are stored in an application execution unit 211 that executes an application that performs processing in response to a request from an external system, a memory 213 that stores data used by the application during processing, and a memory 213. Data is in a format that depends on the type of CPU, OS, etc. for each computing device, but is converted to an environment-independent format that can be interpreted by any computing device without depending on the type of CPU, OS, etc. Environment-independent format data store 214 for storing the read data, read the data written by the application into the memory 213, convert the data to the environment-independent format, store it in the environment-independent format data store 214, and re-install the computing device In processing, the data management unit 212 includes a data management unit 212 that executes processing such as transmitting the environment-independent format data from the computing device A 11 to the computing device B 12.

  Further, when the communication unit 215 and the application execution unit 211 that execute communication between the calculation devices or between the calculation device A 11 and the gateway server 14 via the internal network 13 access the data on the memory 213, address resolution and dirty are performed. The memory management unit 216 that executes flag management, the page table 217 that stores data necessary for address resolution and dirty flag management of the memory management unit 216, and the data management unit 212 correspond to the data on the memory 213 and the data For associating environment-independent format data on the environment-independent format data store 214 with each other and further determining whether or not the environment-independent format data has been transferred from the computer A 11 to the computer B 12 after the last update of the environment-independent format data A variable table 218 that stores flags is included.

  As an example of the environment-independent format of the data stored in the environment-independent format data store 214, an XML (Extensible Markup Language) format, a JSON (JavaScript (registered trademark) Object Notation) format, or the like is applicable. Data in these formats can be handled as the same data regardless of the type of CPU, OS, etc., between computing devices composed of different CPUs, OSs, etc.

  The memory management unit 216 and the page table 217 may be an MMU (Memory Management Unit) provided in a general CPU and a page table having a format corresponding to the MMU. The application execution unit 211 and the data management unit 212 may be functional modules realized by executing a program of the CPU A 219. The page table 217, the memory 213, the environment-independent format data store 214, and the variable table 218 may be different areas of the storage device 220, which is one memory, for example, or may be physically different in whole or in part. It may be a memory or the like. Furthermore, the storage device 220 may be a hard disk. In the second embodiment, an example will be described in which the memory 213 is replaced with a hard disk and stored as a file.

  Although the components of the calculation device B 12 are different from each other, they are the same as those of the calculation device A 11 and will not be described.

  FIG. 3 shows another calculation in which parallel processing is continued for one or more of the calculation devices constituting the FTC system 10 after being removed from the parallel processing between the calculation devices. It is a figure which shows the example of the sequence of the process of recompiling a computer apparatus which recovers the parallel processing state by all the original computer apparatuses by copying the data of an application from an apparatus, and making it correspond. In FIG. 3, the memory 213, the environment-independent format data store 214, and the memory 223 are described to mainly describe data written by execution of the application. However, in the operations of the application execution unit 211 and the data management unit 212, etc. A page table 217 and a variable table 218 whose description is omitted are used. The operation using these will be described later.

  First, the application execution unit 211 of the computing device A 11 performs initial settings and the like necessary for executing the application by the application initial activation 300. Then, the memory allocation request 301 requests the data management unit 212 for memory allocation. In memory allocation 310, data management unit 212 allocates a new area in memory 213 based on the memory allocation request received from application execution unit 211, and passes the address of the allocated area to application execution unit 211.

  The application execution unit 211 executes the memory update process on the area at the address received from the data management unit 212 on the memory 213 in the memory update 302 in order to save the data in the middle of the process in the execution of the application process. . This is the assignment or saving of a value to a variable in an application program. Using the value of that variable, the application process 303 further executes the next application process, and further saves the result data during the process. To do. For this reason, normally, the memory update 302 is repeatedly executed based on the processing content of the application. In response to this memory update 302, the data management unit 212 uses the data update detection and data format conversion 311 to perform processing immediately after the previous data update detection and data format conversion 311 in the area allocated on the memory 213. The updated difference area is identified, data is read from the updated area on the memory 213, the data is converted into the environment-independent format, and the converted data is recorded in the environment-independent format data store 214. Data update detection and data format conversion 311 are repeatedly executed by the data management unit 212. The timing of the repetitive execution may be execution at a fixed period (execution every 1 second), or may be executed based on the update amount by counting the memory update 302 or the like.

  When the data management unit 212 receives the start request for re-installation of the computing device from the computing device B 12, the entire data transfer 312 reads all the data on the environment-independent format data store 214, and all the data is calculated by the computing device B. Send to 12. In the update data transfer 313, the data management unit 212 performs the update on the environment-independent format data store 214 updated after all the data transfer 312 executed immediately before or the process executed later in the update data transfer 313. Data (update difference data) is read, and the update difference data is transmitted to the computer B 12. The update data transfer 313 is repeatedly executed until there is no update difference data that has not been transmitted from the environment-independent format data store 214 to the computer B12. Note that the update data transfer 313 may be executed each time an update of the memory 213 by the memory update 302 is detected, or updates within a fixed period are collectively detected, and every fixed period is detected. It is also possible to execute it.

  When the update data transfer 313 catches up with the memory update 302 and it is detected that there is no update difference data that has not been sent from the environment-independent data store 214 to the computing device B 12, data management is performed by a re-installation completion notification 314. The unit 212 transmits a computer device reassembly completion notification to the computer B12 to complete the computer device reassembly processing on the computer A11 side. If the frequency of memory update 302 is high and unsent update difference data does not disappear, memory update 302 is suppressed by temporarily stopping the operation of the application execution unit 211, and unsent update difference data disappears At this time, a completion notice may be transmitted by a re-installation completion notice 314, and the operation stop of the application execution unit 211 may be released.

  In response to the above operation of the computing device A 11, the application execution unit 221 of the computing device B 12 is activated by the application restart 330 to resume application processing, and is placed in a standby state for application execution. Then, after an area necessary for the application to resume processing is created on the memory 223, the processing of the application is resumed. In response to the memory allocation request 331, the application execution unit 221 requests the data management unit 222 to allocate memory to the data management unit 222. In memory allocation 320, data management unit 222 allocates a new area in memory 223 based on the memory allocation request received from application execution unit 221, and passes the address of the allocated area to application execution unit 221. Since the computing device B 12 is different from the computing device A 11, an area corresponding to the memory allocation 310 can be secured, although the address and capacity allocated to the memory 223 by the memory allocation 320 may be different.

  Next, in application preparation completion notification 332, application execution unit 221 notifies data management unit 222 that application preparation has been completed. In response to the re-assembly start request 321, the data management unit 222 transmits a calculation device re-assembly start request to the computer A 11. On the other hand, since the data is transmitted by the all data transfer 312, in the all data write 322, the data management unit 222 applies the environment-independent format data received from the calculation device A 11 to the application on the calculation device B 12. Is re-converted into a usable format, and is written into the memory 223 based on the address allocated by the memory allocation 320.

  Further, in the update data writing 323, the data management unit 222 reconverts the update difference data in the environment-independent format received from the computing device A 11 into a format that can be used by the application on the computing device B 12, and the memory Write to the memory 223 based on the address assigned in the assignment 320. The update data writing 323 is repeatedly executed every time update difference data is received from the computing device A11. When the data management unit 222 receives the completion notification of the reintegration of the computing device from the computing device A 11 through the application resume notification 324, the data management unit 222 sends an application resume notification to the application execution unit 221. When the application execution unit 221 receives the application restart notification from the data management unit 222 in the application restart 333, the application execution unit 221 uses the address received in the memory allocation request 331, and converts it into the reconverted data written in the address. Resume application processing based on

  FIG. 4 shows a processing flow of the memory allocation 310 executed when the data management unit 212 receives a memory allocation request from the application execution unit 211, and an example of a schema of the variable table 218 registered in the processing. FIG. The variable table 218 will be described, but the description of the schema of the variable table 228 is omitted because it is the same. The variable table 218 may be recorded in an arbitrary recording form on an arbitrary storage device. Each row of the variable table 218 corresponds to variable data used by the application. The data type 401 stores the data type of variable data. The data type includes an integer type, a floating point type, a character string type, and the like, and defines the size, encoding format, and the like of the data on the memory 213.

  The virtual address / offset 402 stores the value of the virtual address that specifies the storage position of the variable data on the memory 213. The virtual address will be described later with reference to FIG. The offset is an offset in the file and will be described in the second embodiment. The variable name 403 stores a variable name that is the name of variable data. The variable name is also used as a key value that defines the storage location on the environment-independent format data store 214. The transfer status 404 stores a flag that defines whether or not the variable data on the environment-independent format data store 214 has been transmitted from the computing device A 11 to the computing device B 12 after the last update. There are two possible values for the transfer status 404: “not transferred” indicating no transmission and “transfer completed” indicating a transmission.

  As a processing flow, in the memory allocation request reception 411, the data management unit 212 receives a memory allocation request from the application execution unit 211. In the memory allocation request, the application execution unit 211 specifies the data type to be stored in the allocated memory and the variable name of the data as arguments. In the necessary memory size acquisition 412, the data management unit 212 acquires a memory size necessary for storing the data of the specified data type on the memory 213. In general, a memory size necessary for storing data of the data type in the memory 213 in an environment for each computing device is uniquely defined in advance.

  In memory area allocation 413, the data management unit 212 allocates a memory area on the memory 213 that can store data of a necessary memory size. For the memory area allocation, a memory allocation routine (POSIX malloc function or the like) provided in a general computer environment may be used, or a unique memory allocation implementation may be used. In the variable table registration 414, the data management unit 212 registers the allocated memory area in the variable table 218 as a new row. At that time, the data type 401 is the data type of the argument, the virtual address / offset 402 is the start address of the allocated memory area, the variable name 403 is the variable name of the argument, and the transfer status 404 is “untransferred” Are registered respectively. Using the memory allocation result return value 415, the data management unit 212 returns the start address of the allocated memory area to the application execution unit 211.

  FIG. 5 is a diagram showing an example of the schema of the page table 217. Since the schema of the page table 227 is the same, the description thereof is omitted. The page table 217 may be recorded in an arbitrary recording form on an arbitrary storage device. Each row in the page table 217 corresponds to a different memory page. Here, the memory page is a data area of a certain size defined in a general processor. The virtual address 501 stores the value of the virtual address. A virtual address is an offset that specifies the position of a memory page in a memory space used by an application, and can be set independently of an offset on a physical memory. The application, that is, the application execution unit 211 accesses the memory 213 using the value of the virtual address. The physical address 502 stores a physical address value. The physical address is an offset of the memory page on the physical memory. The value of the virtual address and the value of the physical address corresponding to the value of the virtual address are converted by the memory management unit 216 using the page table 217.

  The dirty flag 503 stores the value of the dirty flag. The dirty flag 503 can take two values, “set” and “cleared”. The dirty flag 503 is a flag indicating whether or not the memory page has been updated between the last time the dirty flag is set to “cleared” and the next time the dirty flag is checked. When updated, the dirty flag is changed to “set”. Note that even if the memory page is updated in the "set" state and "set" is further set, "set" is still set, so when "set" is set, "cleared" "There is no need to make sure that. If it is “set” when the dirty flag 503 is confirmed, it is set to “cleared”. These settings will be further described with reference to FIGS.

  FIG. 6 is a diagram illustrating an example of a processing sequence for the application execution unit 211 to update data on the memory 213. In response to the memory update request 601, the application execution unit 211 requests the memory management unit 216 to update data on the memory 213. At this time, the application execution unit 211 specifies the address, size, and new data to be used for update of the update target data. In the page table entry search 602, the memory management unit 216 searches the page table 217 for an entry of the page table 217 in which the specified address and the value of the virtual address 501 match, and acquires them. In the memory address conversion 603, the memory management unit 216 refers to the physical address 502 of the matched entry, and identifies the physical position of the update target data on the memory management unit 216.

  Then, in the dirty flag setting 604, the memory management unit 216 sets “set” to the dirty flag 503 of the matched entry. In the memory content update 605, the memory management unit 216 updates the data content to be updated on the memory 213 with the designated new data. In memory update completion 606, the memory management unit 216 notifies the application execution unit 211 that the memory update has been completed.

  Through the above processing, when the application execution unit 211 updates the data on the memory 213, the fact that the data has been updated is recorded in the page table 217. By referring to the page table 217, whether the data has been updated or not is recorded. Can be determined. As another processing sequence for determining whether or not the data is updated, the data on the memory 213 is temporarily set to a read-only state under the same condition as that set to “cleared”, and the application execution unit 211 It is also possible to record the occurrence of an update during the handling of exceptions that occur when updating. Here, the read-only state setting and exception handling processing may be performed by a program that manages the memory 213 executed by the CPU A 219 independent of the application execution unit 211. In general, there is a memory management program that sets the MMU and page table. In addition, as another processing sequence for determining whether or not the data is updated, a copy of the data on the memory 213 is recorded in advance on an arbitrary recording medium, and the occurrence of the update is detected by comparing with the updated data. And so on. Here, copying and comparison of data may be performed by the memory management unit 216, or all assigned areas in the memory 213 may be copied, or may be copied periodically and compared periodically. Further, any other determination processing sequence that functions similarly may be used.

  In FIG. 7, the data management unit 212 detects updated data on the memory 213, converts the updated data into an environment-independent format, and converts the environment-independent format data into an environment-independent format data store. 22 is a diagram showing an example of a sequence of processing stored in 214. FIG. First, in entry acquisition 701 from the variable table, the data management unit 212 acquires one entry from the variable table 218. In the variable update presence / absence confirmation 702, the data management unit 212 acquires an entry in the page table 217 corresponding to the value of the virtual address / offset 402 of the entry in the variable table 218 and the value of the virtual address 501, and the dirty flag 503 is set. Whether or not the data on the memory 213 corresponding to the virtual address / offset 402 of the entry in the variable table 218 has been updated is determined based on whether or not it is “set”. If the update present 703 is NO, the processing from the variable data acquisition 704 to the setting 707 of the variable table entry to the untransferred state is omitted, and the process proceeds to the processing for the next entry in the variable table 218.

  If the update presence 703 is YES, the data management unit 212 sets the dirty flag 503 to “cleared”, and reads the content of the updated data on the memory 213 in the variable data acquisition 704. In the conversion 705 of the variable data to the environment-independent format, the data management unit 212 uses the information such as the data type 401 of the entry acquired from the variable table 218 and the endian of the computer A 11 to read the memory 213. The above data is converted into an environment-independent format. In storage 706 of environment-independent format data, the data management unit 212 stores the environment-independent format data in the environment-independent format data store 214 in association with the variable name 403 of the entry in the variable table 218. In the setting 707 of the variable table entry to the untransferred state, the data management unit 212 sets the transfer state 404 of the entry in the variable table 218 to “untransferred”.

  Upon completion of the processing of all variables 708, the data management unit 212 sequentially executes the processing from the entry acquisition 701 from the variable table to the setting 707 of the variable table entry to the untransferred state for all entries of the variable table 218. The data update detection and data format conversion process is completed.

  FIG. 8 shows an environment in which the data management unit 212 on the computing device A 11 has been updated since the last update difference data transfer and data write processing among the environment-independent format data on the environment-independent format data store 214. Independent format data is detected, transmitted to the computing device B 12, and the environment-independent format data received by the computing device B 12 is converted from the environment-independent format into a format that can be used by the application, and the converted data FIG. 6 is a diagram illustrating an example of a sequence of processing for writing the data to the memory 223; First, the data management unit 212 of the computing device A 11 acquires one entry from the variable table 218 in the entry acquisition 801 from the variable table, reads the transfer status 404 of the entry in the transfer status 802, and transfers the “transfer” If it is “completed” state, that is, NO, processing from variable data acquisition 803 to setting 805 to the transferred state of the variable table entry is omitted, and the process proceeds to the next entry in the variable table 218.

  If the transfer state 802 is “non-transfer” state, that is, YES, in the variable data acquisition 803, the data management unit 212 stores the data in the environment-independent format data store 214 corresponding to the variable name 403 of the entry in the variable table 218. Get environment-independent format data. In variable data transfer 804, the data management unit 212 transmits the data type, variable name, and environment-independent format data obtained from the variable table 218 to the computing device B12. Then, in the setting 805 of the variable table entry to the transferred state, the data management unit 212 sets the transfer state 404 of the entry in the variable table 218 to “transferred”. Upon completion of the processing of all variables 806, the data management unit 212 sequentially executes the processing from the entry acquisition 801 from the variable table to the setting 804 of the variable table entry to the transferred state for all the entries in the variable table 218. .

  On the other hand, the data management unit 222 of the computing device B 12 converts the environment-independent format data received from the computing device A 11 into a format that can be used by the application on the computing device B 12 in the data format reconversion 811. Reconvert. The reconversion converts environment-independent format data to a data format such as endian of the data type of the computer B 12 corresponding to the data type received from the computer A 11. Then, in the variable data writing 812, the data management unit 222 stores the reconverted data in the memory 223. As the address of the memory 223 to be saved, the variable name received from the computer A 11 is searched from the variable name 403 of the variable table 228, and the value of the virtual address / offset 402 of the entry of the matched variable name is used. Here, since the memory allocation request 331 requests the same variable name as the memory allocation request 301 as an argument, there is a matching variable name. Note that since the variable table 228 also has a data type 401, the data type 401 of the entry of the variable name that matches the received variable name may be used instead of the received data type.

  As described above, the calculation device B 12 having a different data format can also have the same memory contents as the calculation device A 11, so that the application process can be resumed from the time of the contents of the memory. Then, by executing the same processing contents in a plurality of computing devices in the same processing order, it is possible to improve the fault tolerance and reliability of the system.

  In the above description, the memory allocation 320 is executed in advance before receiving data from the calculation device A 11 in the calculation device B 12, and the address of each variable to be used is determined. For example, the memory allocation 320 is executed in advance. It does not have to be. In FIG. 3, the memory allocation request 331 and the memory allocation 320 are omitted, and the variable table 228 is registered at the update data write 323. That is, in FIG. 8, instead of reading the value such as the virtual address / offset 402 from the variable table in the variable data write 812, the value of the virtual address / offset 402 is generated by the same process as the memory area allocation 413 in FIG. The data type and variable name received from the computer A 11 are written in the variable table 228. Then, the conversion data is stored at the position of the generated virtual address / offset 402 value.

  In this case, since the application execution unit 221 has not acquired the address of each variable at the time of the application restart 333, the process sequence as shown in FIG. 9 is executed. FIG. 9 is a diagram illustrating an example of a processing sequence in which the application execution unit 221 accesses data on the memory 223 in the computing device B 12 after the re-installation is completed. In the address inquiry 901, the application execution unit 221 of the computer B 12 notifies the data management unit 222 of the variable name of the data to be accessed and inquires about the address of the data. In the variable search 911, the data management unit 222 searches the variable table 228 for an entry corresponding to the variable name and the variable name 403. In the address notification 912, the data management unit 222 notifies the application execution unit 221 of the value of the virtual address / offset 402 of the entry as an address. In the data access 902, the application execution unit 221 accesses the data stored at the address on the memory 223. In this way, the application execution unit 221 needs to inquire about the address for each data access 902, but it is not necessary to execute the memory allocation 320 or the like in advance.

  Further, the application execution unit 221 may acquire addresses collectively before the application restart 333 instead of every data access 902. FIG. 10 shows that the computing device B 12 temporarily saves the environment-independent format data in the environment-independent format data store 224, and collectively reconverts the data format according to the memory allocation request 331 of the application execution unit 221. FIG. 11 is a diagram showing an example of a sequence of processing for saving to a memory 223 and passing an address to an application execution unit 221.

  As shown in FIG. 8, all data write 322 and update data write 323 in FIG. 3 reconvert data and store it in the memory 223, whereas all data write 325 and update data write 326 in FIG. The environment-independent format data is stored in the environment-independent format data store 224 without re-converting the data. Then, in response to the memory allocation request 331 immediately before the application restart 333, the data management unit 222 executes memory allocation data format conversion 327 instead of simply performing memory allocation. In the memory allocation data format conversion 327, the data of the variable name specified by the argument of the memory allocation request 331 is read from the environment-independent format data store 224, and the re-converted data format is stored in the memory 223. The address is passed to the memory allocation request 331. In this way, even if the application execution unit 221 only requests memory allocation, the reconverted data is stored at the address of the allocated area. You can resume.

  As described above, the timing of data format re-conversion and the timing of sending an address to the application execution unit 221 can be variously combined. For example, the CPU B 229 of the computing device B 12 and the type of the memory management unit 226 Accordingly, a timing such as the shortest processing time may be selected.

  As a second embodiment, an example in which a file on a hard disk or the like is applied instead of a memory to a destination where the application execution unit 211 stores data will be described. The schema of the variable table 218 when the file is applied instead of the memory to the destination where the application execution unit 211 stores data is as shown in FIG. 4, and the virtual address / offset 402 is not a virtual address but a variable. Stores the offset value that specifies the storage location of the data file. Further, in the memory area allocation 413 in FIG. 4, a memory area is allocated, but in this example, a file area is allocated.

  FIG. 11 is a diagram illustrating an example of a sequence of data update detection and data format conversion processing when the application execution unit 211 applies a file instead of a memory as a data storage destination. This corresponds to FIG. The main components are almost the same as in the case of using a memory, but the file update detection function of the OS 231 is used instead of the page table 217 for determining whether or not to update data. When the application execution unit 211 updates a file, it always requests the OS 231 and the OS 231 executes the file update process. Since the record of the presence / absence of update may be implemented in any configuration in the OS 231, description of the implementation is omitted.

  After acquiring one entry from the variable table 218 in the entry acquisition 701 from the variable table, the data management unit 212 corresponds to the offset 402 of the entry in the variable table 218 in the variable update presence / absence confirmation 712 by the OS query. The OS 231 is inquired to determine whether the file data to be updated has been updated. If updated, the data management unit 212 reads the content of the data on the file in the variable data acquisition 714. The sequence of the other processes is as described with reference to FIG. 7, and the same processes are denoted by the same reference numerals.

  Also, the data update is executed by requesting the OS 231 rather than via the memory management unit 216. For this reason, the storage device 220 as a file and the OS 231 that controls the storage device 220 constitute a storage unit. Other processing sequences are the same as those already described with reference to the drawings.

  As described above, the calculation device B 12 having a different data format on the file can also have the same file contents as the calculation device A 11, so that the application processing is performed using the file contents. You can resume from the point of time.

  As a third embodiment, an example will be described in which a database is used instead of a file while the application execution unit 211 applies a file as a data storage destination in the second embodiment. The database includes a data storage unit that stores data and a data update interface that updates data in response to a request from the application execution unit 211. For this reason, the data storage unit and the data update interface constitute a storage unit. Other components are the same as when using a file, but because it is possible to record the presence or absence of data update with the data update interface, the OS file update detection function is not required to determine the presence or absence of data update. There is an advantage that can be applied even when the OS does not have the file update detection function. Further, when the data storage unit stores data, the data may be converted into an environment-independent format. In this case, there is an advantage that the variable data as the processing of the data management unit 212 shown in FIG.

211, 221: Application execution unit, 212, 222: Data management unit, 213, 223: Memory, 214, 224: Environment independent format data store, 215, 225: Communication unit, 216, 226: Memory management unit, 217, 227: Page table, 218, 228: Variable table

Claims (11)

A computer system composed of a plurality of processing computers that process data of different formats and a management computer,
Each of the processing computers is
A processing unit that generates data of an intermediate result of the process, and executes a next process based on the data of the intermediate result;
A storage unit that stores the intermediate result data in a format that can be processed by the processing unit;
Converting the stored data into data in a predetermined format, and converting the data in the predetermined format into data in a format stored in the storage unit;
A transfer unit that outputs the data in the predetermined format to the other processing computer and inputs from the other processing computer;
A communication unit that outputs a result of the processing to the management computer;
With
The first processing computer added to the computer system inputs the data in the predetermined format from the second processing computer already existing in the computer system by the transfer unit, and inputs the input data. The conversion unit converts the data into a format that can be processed by the processing unit, stores the converted data in the storage unit, executes the next processing based on the stored data, and includes the execution Output the processing result to the management computer,
The management computer compares the processing results input from each of the plurality of processing computers including the first processing computer and the second processing computer, and the result of the processing with the most number of matches is the computer. Output to outside the system,
A computer system characterized by The computer system according to claim 1,
The second processing computer is
The conversion unit detects storage of the generated intermediate result data, and converts the stored data into data of a predetermined format based on the detection;
A computer system characterized by The computer system according to claim 1,
The second processing computer is
The storage unit manages the storage area in units of pages by a page table, and stores the generated intermediate result data and sets the flag of the stored page of the page table,
The converter detects the set of flags, and converts the data stored in the page into data of a predetermined format based on the detection;
A computer system characterized by The computer system according to claim 1,
The second processing computer is
The storage unit can be further set to read-only, and the occurrence is recorded by handling an exception that occurs when storing data in a state set to read-only,
The converting unit converts the stored data into data in a predetermined format based on the occurrence record;
A computer system characterized by The computer system according to claim 1,
The second processing computer is
The storage unit records a copy of the stored data on a recording medium, compares the original data of the copy with the recorded data,
The converter converts the stored data into data of a predetermined format based on the result of the comparison;
A computer system characterized by A computer system according to claim 1, wherein
Each of the processing computers is
The storage unit stores the data in a memory;
A computer system characterized by A computer system according to claim 1, wherein
Each of the processing computers is
The storage unit stores the data in a file;
A computer system characterized by A computer system according to claim 1, wherein
Each of the processing computers is
The storage unit stores the data in a database;
A computer system characterized by A computer system according to claim 1, wherein
Each of the processing computers is
The conversion unit attaches the same variable name as the stored data variable name and converts it into data of a predetermined format, and the same variable name and storage position as the variable name of the predetermined data format And converting the data into a format stored in the storage unit, and storing the converted data in the storage location,
A computer system characterized by A computer system according to claim 1, wherein
The predetermined format is an XML format or JSON format,
Each of the processing devices
The conversion unit converts the stored data into the XML format or JSON format data, and converts the XML format or JSON format data into data stored in the storage unit;
A computer system characterized by A processing method of a computer system composed of a plurality of processing computers and management computers that process data of different formats,
A step of generating and storing data of an intermediate result of the processing by the first processing computer;
Converting the data generated and stored by the first processing computer into data of a predetermined format and storing the data;
Outputting the data converted and stored by the first processing computer to the second processing computer;
The second processing computer inputs the output data, converts the data into a format that can be processed by the second processing computer, and stores the data;
Executing the following processing based on the data generated and stored by the first processing computer and outputting a first processing result to the management computer;
A step of executing a next process based on the data stored by the second processing computer after the conversion and outputting a second processing result to the management computer;
The management computer compares the first processing result input from each of the plurality of processing computers including the first processing computer and the second processing computer with the processing result including the second processing result, Outputting the result of the processing with a large number of matches to the outside of the computer system;
A processing method for a computer system, comprising:

Images