JP2006268596A - Redundancy system of service system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize minimizing the interruption of a service due to a failure, and collecting information on the failure reliably with a simple composition. <P>SOLUTION: The system comprises first and second service processing parts 21A and 21B achieving the same service function by implementing a program, first and second memory areas 41A and 41B provided in correspondence with the first and second service processing parts for storing the processing information related to the operation of the service, a save memory 60 for saving the storage information of the first or second memory areas, and a service control part 30 for operating a pair consisting of the first and second service processing parts and memory areas as an active system and a standby system. The service processing part 21A of the active system writes its processing information in the memories 41A and 41B of the active and standby systems. The service control part 30 switches the system immediately when a failure occurs in the active system, and, at the same time, saves the log information stored in the memory area 41A of the previous active system in the save memory 60. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a service system redundancy system, and more particularly to a service system redundancy system that provides a predetermined service function by executing a computer program. Today, various service functions such as various communication control devices are realized by computer program execution, and it is desired to improve the performance and reliability of such service systems.

  Conventionally, as a method of shortening the interruption time of this type of service system, not only the CPU of the computer system but also all other hardware such as memory are physically duplicated (redundant), and both systems are always synchronized. Even if the active computer system fails, the standby computer system can quickly take over the memory operation information, etc., making the standby system a new active system and maintaining the service status until the failure occurs. There is a way to resume. As a result, the service can be quickly recovered while collecting failure information in a short service interruption time.

  However, the duplication of hardware has a drawback that a lot of introduction cost and power consumption are required. Moreover, most of the failures (failures) that actually occur are failures due to software errors (bugs) or overloads, and it is unlikely that the hardware will actually fail and the system will switch, and it is not possible to duplicate hardware. In many cases, the investment was excessive.

  Also, conventional computer systems are provided with a function that forcibly terminates a program and outputs the contents of the previous memory or register to a failure information file when a fatal error that prevents processing from continuing occurs. It is an indispensable function for investigating the cause of failure.

  However, the failure file output processing time at that time depends on the load module (LM) size of the corresponding program and the size of the memory area captured in the program processing. It could take several minutes and had a significant impact on service interruption time.

  In particular, in a server system that requires real-time processing such as handling audio, it is necessary to restart the program as soon as possible when the program fails. However, if the program is restarted before the failure file information collection is completed, the registers and stack Since the memory is overwritten, there is a possibility that valid information cannot be obtained.

In this regard, conventionally, a system bus 6 used for normal access to a computer system and a diagnostic bus 7 dedicated to maintenance / diagnosis for serially transferring data are provided, and local memory information from a fault unit 3 is transferred to the system bus. In the case where data cannot be dumped to the main memory 1 via 6, it is known that the local memory information can be collected even by a failure related to the system bus 6 by collecting via the diagnostic bus 7 ( Patent Document 1).
JP-A-6-342387 (summary, figure)

  However, although the above configuration is relatively simple with a separate serial diagnostic bus, when the system diagnostic bus 7 cannot be used and the serial diagnostic bus 7 is used, sufficient fault information is always collected in a short time. There is a problem that cannot be done.

  The present invention has been made in view of the above-described problems of the prior art, and the object of the present invention is to minimize service interruption due to a failure and to reliably collect failure information at that time with a simple configuration. It is to provide a redundant system of service system.

  The above problem is solved by the configuration of FIG. That is, the service system redundancy system of the present invention (1) is a service system redundancy system that provides a predetermined service function by computer program execution, and is a first system that realizes the same service function by program execution. , Second service processing units 21A, 21B and first and second memory areas 41A, 41B provided corresponding to the first and second service processing units, respectively, for storing processing information relating to service operation. Service management that operates a pair of the save memory 60 that saves the storage information of the first or second memory area and the first and second service processing units and the memory area as an active system and a standby system The active service processing unit 21A writes its own processing information report into the active and standby memories 41A and 41B. Service management unit 30 is configured to retract switch quickly system by failure of the active system, and the log information stored in the memory area 41A of the old active system to save memory 60.

  According to the present invention (1), the service processing units 21A and 21B realized by program execution and the duplicated memory areas 41A and 41B are provided, and the active system, while synchronizing (matching) the storage information of both areas, With a simple configuration that operates as a standby system, when a failure occurs in the active system, the system can be switched quickly, and log information related to the previous service processing can be reliably collected from the memory area 41A of the former active system. .

  According to the present invention (2), in the above-mentioned present invention (1), the first and second services that are bound to the first and second service processing units and perform the memory access to the first and second memory areas, respectively. An adapter unit is provided, and the active adapter unit normally writes the same data to the active and standby memory areas in accordance with a data write command from the active service processor, and the standby service processor. However, the data is written only in the active memory area because of failure or because the data in the standby memory area is being saved. Therefore, the operation data can be continuously updated, and the memory information of the old working system after the system switching can be surely protected.

  In the present invention (3), in the above-mentioned present invention (2), the service management unit performs data writing during a period when the standby service processing unit 21A is faulty or data in the standby memory area 41A is being saved. The stored area information of the active memory 41B is retained, and the storage data corresponding to the area information of the active memory 41B is copied to the standby memory area 41A by the restoration of the standby service processing unit 21A. Is.

  According to the present invention (3), with a simple configuration that only holds the area information of the active memory 41B where data has been written, when the standby system 21A is restored, the data in the area of the active memory 41B is waited. The contents of the system memory 41A can be easily synchronized (matched).

  In the present invention (4), in the present invention (1), for example, as shown in FIG. 9, the first and second service processing units 21A and 21B are configured to operate by their own CPUs 11A and 11B, respectively. It is what. Therefore, a required service function can be provided with high performance and safety by a relatively simple duplex configuration.

As described above, according to the present invention, service interruption due to a failure can be minimized, and failure information at that time can be reliably collected. Therefore, this type of service system using a computer can be provided inexpensively and safely. it can.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described in detail with reference to the accompanying drawings. Note that the same reference numerals denote the same or corresponding parts throughout the drawings. FIG. 1 is a block diagram of a computer system according to an embodiment, and shows a configuration example by a multi-programming method (a plurality of service processes are executed in parallel by one CPU).

  In the figure, 100 is a multi-programming computer system, 11 is a CPU, 10A and 10B are service processors A, B, 20A and 20B which can realize the same service function (voice communication function, etc.) by executing a service program. Main memories (RAM, cache memory, etc.) for storing programs, 21A and 21B are realized by execution of service programs, and service processing units A, B, 22A and 22B are virtual memories A and B of service processing units A and B. The adapter units A, B, and 30 that interface access respectively perform monitoring of failures of the service processors 10A and 10B, management of service operation states of the active and standby systems, switching control between the active and standby systems when a failure occurs, and the like. Service management unit to perform, 40 for service system operation A memory for storing various processing information (may be a part of the main memory), 41A and 41B are virtual memories A, B and 50 which are duplicated into an active system and a standby system, and a program related to an operating system (OS) is stored The operating system unit 60 is a disk device (DISK) for saving and storing information of the virtual memory A / B.

  In the following description, the service processing units 21A, 21B, etc. are simply referred to as service processing units A, B, etc. for the sake of simplicity. Further, the service processing units A and B have dual (redundant) functions, and the figure shows the case where A is the active system and B is the standby system (standby system).

  The service processing units A and B and the service management unit 30 are realized by program execution of the CPU 11 using the OS 50 as a platform.

  The virtual memory A is an operational memory used by the active service processing unit A for reading and writing processing information related to service operation, and the virtual memory B is prepared for system switching in the event of a failure by the active service processing unit A. When the system is switched when a failure occurs, the service processing unit B that has become a new active system continues to read and write processing information related to service operation. This is the operational memory used for On the other hand, the virtual memory A that has completed the collection of log information related to the failure becomes a standby memory in which the service processor B of the active system copies the same information as the virtual memory B in preparation for switching at the time of failure.

  Thus, the working memory for the service processing unit A is the virtual memory A and the standby memory is the virtual memory B, and the working memory for the service processing unit B is the virtual memory B and the standby memory is the virtual memory A. Since the service processing units A and B are generated (compiled) from the same source file, it is necessary for each service processing unit a and B to translate (convert) memory access destinations (addresses) for the operation memory and the standby memory. is there.

  Therefore, the adapter unit A translates the access to the working and standby memory from the service processing unit A into access to the virtual memories A and B, while the adapter unit B virtualizes the access to the working and standby memory from the service processing unit B. Translate to access to memory B, A.

With this configuration, the stored information in the virtual memories A and B is matched in advance. Now, assuming that the service processing unit A is an active system, the adapter unit A performs data reading R only from the virtual memory A in accordance with a data reading command from the service processing unit A. The adapter unit A performs data writing W1 and W2 in the same (corresponding) areas of the virtual memories A and B in accordance with a data write command from the service processing unit A. Since this process is only executed by the adapter unit A, the programmer who writes the service program does not need to be aware of such a memory access mechanism. On the other hand, the service processing unit B during this period does not perform memory access to the virtual memories A and B. Therefore, the stored contents of the virtual memories A and B always match after the system operation is started. As a result, the service management unit 30 can quickly switch the system at any time.

  FIG. 2 is a diagram for explaining the construction procedure of the service processor according to the embodiment, and shows the procedure of generating an object program for accessing the virtual memories A and B via the adapter unit from the source program of the service processor. In order for the service processing units A and B to properly access the virtual memories A and B regardless of the active / standby system, the memory access part of the service processing units A and B is changed to the adapter units A and B. There is a need to.

  In order to realize this, when the source file of the service program is first input to the precompiler and precompiled, an intermediate file in which the memory access portion is converted via the adapter is generated. Furthermore, when this intermediate file is linked with the adapter component and the OS-dependent component, and compiled by the compiler, an adapter-compatible service program (object file) that considers access to the virtual memories A and B via the adapter section as the final output ) Is generated.

  FIG. 3 specifically shows the operation of the precompiler according to the embodiment. Here, instructions related to memory access are converted into those via the adapter unit. For example, since the variable definition “static int X” in the source file 1 is an internal reference, it does not relate to memory access. On the other hand, “shmget (aa, bb)” is an instruction for securing an area in the virtual memory, and is precompiled into an intermediate file “adp_shmget (aa, bb)” via the adapter unit. Others can be understood similarly. Further, since the variable X of the source file 2 is the external reference “extern int X”, it is converted to the adapter unit.

  Next, the recovery operation when an active system failure occurs with such a configuration will be described. 4 and 5 are diagrams (1) and (2) for explaining the operation of the computer system according to the embodiment. FIG. 4 is a diagram when a failure occurs in the service processor A of the active system and the operating system is switched. Indicates the state.

  The service management unit 30 and the OS 50 monitor the operating state of the service processing unit A. If any failure occurs in the service processing unit A in this state, this failure is detected by either the OS 50 or the service management unit 30. Is done. When the OS 50 detects a failure in system operation, the service management unit 30 is notified of the occurrence of the failure. When the service management unit 30 detects a failure in service (application) operation, the OS 50 performs service processing. Notification of occurrence of failure in part A.

  The service management unit 30 collects execution information (register information, stack information, etc.) of the service processing unit A operating under OS management as failure information and saves it in a log file on the DISK 60. Further, the service processing unit B is notified of a service takeover request, and information on the virtual memory area A used by the failed service processing unit A is also saved in the DISK 60. Since these saving processes are performed by the old operational system (service processing unit A, virtual memory A, service management unit 30), when viewed from the new operational system service processing unit B, they are executed in the background. The service is resumed without shortening the system interruption time due to the collection and without causing the service performance degradation due to the load accompanying the log collection.

On the other hand, the service processing unit (adapter unit) B, which has become a new operation system, reads data from the virtual memory B and tries to write the same data to the virtual memories B and A, but the virtual memory A protects the failure information. Or data cannot be written because it is being collected. Therefore, the adapter unit B in this case notifies the service management unit 30 of at least the area information of the virtual memory B in which the data is to be written, and the service management unit 30 holds this.

  FIG. 5 shows an operation when the log collection / save operation of the virtual memory A is completed. When the service processing unit A outputs a recovery notification upon completion of log collection, the service management unit 30 receives the request and makes a copy request to the active adapter unit 22 and notifies the stored area information. Receiving this, the adapter unit B reads the information of the area of the virtual memory B and writes it in the corresponding (same) area of the virtual memory A. When there are a plurality of areas stored in the service management unit 30, the same copying process is repeated, and when the data copying of all the storage areas is completed, the data in the virtual memories B and A is synchronized (matched). According to this method, since it is not necessary to copy all the data between the virtual memories A and B, the efficiency is high.

  Further, the service management unit 30 restarts the service processing unit A as a standby system after the completion of all copying, and notifies the adapter unit B of the fact, whereby the data to the virtual memory A by the adapter unit B is transmitted. Writing W2 begins. In this way, the active system and the standby system are completely switched.

  FIGS. 6 to 8 are operation sequence diagrams (1) to (3) of the adapter unit according to the embodiment, and show some typical operation sequences for realizing the above-described operation of the adapter unit. The upper half of FIG. 6 shows a process in which the current service processing unit A acquires a memory area from the adapter unit A. This process is performed as a pre-process for securing the area when the service processing unit A wants to access an area with the virtual memories A and B.

  When the service processing unit A makes a memory area acquisition request to the adapter unit A, the adapter unit A checks the operating state of the service processing unit B in step S11, and the service processing unit B is operating as a standby system (that is, normal) ), The memory areas of the virtual memories A and B are acquired in steps S12 and S13, and an area acquisition response is sent to the service processing unit A. Further, even when the service processing unit B has failed or saved log information in the determination in step S11, the memory areas of the virtual memories A and B are acquired in steps S14 and S15, and an area acquisition response is sent to the service processing unit A. Do. In any of the above cases, the service management unit 30 is notified of the memory area acquisition information, and the service management unit 30 that has received the information holds the acquisition information in step S16. Accordingly, the service management unit 30 can manage which memory area of the virtual memory A or B is accessed. In this example, it is not necessary to branch the process in step S11, but the process is branched for the sake of clarity.

  The lower half of FIG. 6 shows a process in which the active service processing unit A reads data from the adapter unit A. When the service processing unit A makes a memory read request to the adapter unit A, the adapter unit A reads data from the acquisition area of the virtual memory A and sends a memory read response to the service processing unit A in step S17.

  The upper half of FIG. 7 shows a process in which the current service processing unit A writes the same data in the virtual memory A and, if possible, the virtual memory B. When the service processing unit A makes a memory write request to the adapter unit A, the adapter unit A checks the operating state of the service processing unit B in step S21, and if the service processing unit B is operating as a standby system, In steps S22 and S23, data is written to the virtual memories A and B, and a memory write response is sent to the service processing unit A. If it is determined in step S21 that the service processing unit B is in failure or saving, data is written only in the virtual memory A in step S24, and a memory write response is sent to the service processing unit A. Therefore, destruction of the contents of the virtual memory B can be effectively prevented.

  In the lower half of FIG. The case where the memory area B is released is shown. When the service processing unit A makes a memory release request to the adapter unit A, the adapter unit A checks the operating state of the service processing unit B in step S25, and if the service processing unit B is operating as a standby system, the step In S26 and S27, the memory areas of the virtual memories A and B are released, and a memory release response is sent to the service processing unit A. If it is determined in step S25 that the service processing unit B is in failure or saving, the virtual memories A and B are released in steps S28 and S9, and a memory release response is sent to the service processing unit A.

  In either case, the service management unit 30 is notified of the memory release information, and the service management unit 30 that has received the information holds the acquired information in step S30. Thereby, the service management unit 30 can manage that the memory access to the virtual memories A and B is released.

  FIG. 8 shows a process for matching the data between the virtual memories A and B accompanying the failure recovery of the service processing unit A (that is, the log information collection of the virtual memory A is completed). Upon receiving the recovery notification from the service processing unit A, the service management unit 30 determines whether or not there is a memory area saved regarding the access to the virtual memory B in step S41. If there is, a copy request between the virtual memories B and A is made to the working adapter unit B, and the area information is notified.

  Receiving this, the adapter unit B copies the data in the area of the virtual memory B to the corresponding area of the virtual memory A in step S50, and sends a copy completion response to the service management unit 30.

  Receiving this, the service management unit 30 resets the information of the acquisition area in step S42, and returns to step S41. Thus, when all of the copy processing of one or more storage areas is completed, the service processing unit A is notified of restart of the standby system.

  FIG. 9 is a block diagram of a computer system according to another embodiment, and shows a configuration example of a multi-CPU system in which two service processes can be executed in parallel by two CPUs. In this example, the service processing unit A is realized by executing the service program of the CPU 11A, and the service processing unit B is realized by executing the service program of the CPU 11B. CPU-A operates under OS-A, and CPU-B operates under OS-B. Thus, the processing performance of the service processing units A and B is greatly improved.

  On the other hand, the system management unit 30 in this example is basically realized by program execution of the standby CPU 11B, and the active CPU 11A in this case can concentrate exclusively on providing the original service function.

  In the above embodiment, an access method for acquiring a memory area in advance has been described. However, the present invention is not limited to this. Various other memory access methods can be employed.

  Moreover, although several embodiment suitable for the said invention was described, it cannot be overemphasized that the structure of each part, control, a process, and these combination can be variously changed within the range which does not deviate from this invention. .

1 is a block diagram of a computer system according to an embodiment. It is a figure explaining the construction procedure of the service processor by embodiment. It is a figure explaining operation | movement of the precompiler by embodiment. FIG. 5 is a diagram (1) illustrating an operation of the computer system according to the embodiment. FIG. 6 is a diagram (2) illustrating the operation of the computer system according to the embodiment. It is an operation | movement sequence diagram (1) of the adapter part by embodiment. It is an operation | movement sequence diagram (2) of the adapter part by embodiment. It is an operation | movement sequence diagram (3) of the adapter part by embodiment. And FIG. 12 is a block diagram of a computer system according to another embodiment.

Explanation of symbols

11 CPU
10A, 10B Service processor A, B
20A, 20B Main memory A, B
21A, 21B Service processing part A, B
22A, 22B Adapter part A, B
30 Service management unit 41A, 41B Virtual memory A, B
60 Disk unit (DISK)
100 computer system

Claims (4)

A redundancy system for a service system that provides a predetermined service function by executing a program on a computer,
First and second service processing units for realizing the same service function by program execution;
First and second memory areas respectively provided corresponding to the first and second service processing units and storing processing information related to service operation;
A save memory for saving storage information in the first or second memory area;
A service management unit that operates the pair of the first and second service processing units and the memory area as an active system and a standby system;
The service processor of the active system writes its processing information report in the memory of the active system and the standby system, and the service management section switches the system promptly when a failure occurs in the active system, and the old service system memory area. A service system redundancy system characterized in that stored log information is saved in a save memory. The first and second adapter units are respectively bound to the first and second service processing units and perform memory access to the first and second memory areas.
The active adapter unit normally writes the same data to the active and standby memory areas in accordance with a data write command from the active service processor, and the standby service processor is in trouble or 2. The service system redundancy system according to claim 1, wherein the data is written only in the active memory area because the data in the standby memory area is being saved. The service management unit holds the area information of the active memory in which data has been written while the standby service processing unit is faulty or data in the standby memory area is being saved. 3. The service system redundancy system according to claim 2, wherein the stored data corresponding to the area information of the active memory is copied to the standby memory area by restoration of the service processing unit. 2. The service system redundancy system according to claim 1, wherein each of the first and second service processing units is configured to operate by a unique CPU.

Images