JP2012256240A - Duplex system and memory synchronization method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate development of an application program and enhance program development efficiency in a duplex system which synchronizes memories of an active server and a stand-by server.SOLUTION: A duplex system comprises an active server 1 and a stand-by server 2. The active server 1 comprises: application processing means 101 and 103 which perform application processing; and first synchronization means 102 and 104 which, in response to a first update request generated in the application processing for updating a main memory 12 of the active server 1, transmit a second update request to the stand-by server 2 for updating a main memory 22 of the stand-by server 2. The stand-by server 2 comprises second synchronization means 202 and 204 which update the main memory 22 of the stand-by server 2 on receiving the second update request from the active server 1.

Description

  The present invention relates to a technique for synchronizing a memory of an active server and a memory of a standby server in a duplex system including an active server and a standby server.

  A server requiring reliability, such as a SIP server in an IP telephone system, has a duplex configuration. In a redundant configuration, a pair of servers with the same functions of the active system and the standby system are used as one cluster, and the memory contents are transferred from the active system to the standby system via the network connecting the systems, and the memory is synchronized. It is possible to switch the system by operating it.

  Patent Document 1 describes a method of synchronizing data of each database server when a new database server is incorporated into the system in a multiplexed database system.

JP 2007-241323 A

  However, in the conventional technology as described above, when developing an application program such as call processing, in addition to the original application processing, it is necessary to perform implementation in consideration of memory data transfer processing between systems. There is a problem that becomes complicated. That is, the application program becomes complicated, and it takes time and load to develop the program.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to facilitate development of an application program and improve program development efficiency in a duplex system in which memory synchronization is performed between an active server and a standby server. There is to make it.

  In order to achieve the above object, the present invention is a duplex system comprising an active server and a standby server, wherein the active server generates application processing means for performing application processing and the application processing, First synchronization means for transmitting a second update request for updating the main memory of the standby server to the standby server in response to a first update request for updating the main memory of the active server; The standby server includes second synchronization means for updating the main memory of the standby server when receiving the second update request from the active server.

  The present invention is a memory synchronization method for synchronizing the main memory of the active server and the main memory of the standby server performed by the duplex system.

  According to the present invention, development of an application program can be facilitated and program development efficiency can be improved in a duplex system in which memory synchronization is performed between an active server and a standby server.

It is a block diagram of the duplex system which concerns on embodiment of this invention. It is a figure which shows the program structure on each server. It is a sequence which shows the memory synchronization process of 1st Example. It is a sequence which shows the memory synchronization process of 2nd Example. It is a sequence which shows the memory synchronization process of a 3rd Example. It is a sequence which shows the memory synchronization process of a 4th Example.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a configuration diagram of a duplex system according to an embodiment of the present invention. The duplex system shown in the figure includes a pair of active server 1 (ACT server) having the same function and a standby server 2 (SBY server). The active server 1 and the standby server 2 transfer the contents of the memory from the active server 1 to the standby server 2 via the network 9 connecting the servers as one cluster, and operate while synchronizing the memories. This makes it possible to switch systems.

  The active server 1 of the present embodiment includes a plurality of CPUs 11, a main memory 12 (memory), a disk controller 13, a disk 14, and a NIC (Network Interface Card) 15, which are connected by a bus 16. ing. Similarly, the standby server 2 of the present embodiment includes a plurality of CPUs 21, a main memory 22, a disk controller 23, a disk 24, and a NIC 25, which are connected by a bus 26.

  As described above, the active server 1 and the standby server 2 of the present embodiment are multi-core servers including a plurality of CPUs in a housing (inside the server).

  FIG. 2 is a diagram illustrating an example of a program configuration of the active server 1 and the standby server 2. In the illustrated example, each of the active server 1 and the standby server 2 includes four CPUs 111 to 114 and 211 to 214.

  In the illustrated example, each of the active server 1 and the standby server 2 includes two server program processes (server process 1, server process 2: application processing means) 101, 103, 201, 203 in parallel. It is assumed that it is operating. The server program is an application program that performs predetermined application processing. A process is an execution unit of a program that receives a memory area allocation from the OS and executes processing.

  Each server process 101, 103, 201, 203 uses resources such as the CPU and main memories 12, 22 via the OSs 105, 205. In the illustrated example, it is assumed that the server process 1: 101 of the active server 1 operates on the CPU 1: 111, and the server process 2: 103 operates on the CPU 3: 113. Further, it is assumed that the server process 1: 201 of the standby server 2 operates on the CPU 1: 211 and the server process 2: 203 operates on the CPU 3: 213.

  In the present embodiment, the processes (memory synchronization process 1, memory synchronization process 2: synchronization means) 102, 104, 202 of the memory synchronization program for executing the memory synchronization processing between the active server 1 and the standby server 2 , 204 are provided for each server process. Note that the memory synchronization program is used as middleware that operates on the OSs 105 and 205.

  In the illustrated example, in the active server 1, a memory synchronization process 1: 102 that executes a memory synchronization process of the server process 1: 101, and a memory synchronization process 2: 104 that executes a memory synchronization process of the server process 2: 103, Is provided. The memory synchronization process 1: 102 operates on the CPU 2: 112, and the memory synchronization process 2: 104 operates on the CPU 4: 114.

  Similarly, the standby server 2 includes a memory synchronization process 1: 202 that executes a memory synchronization process of the server process 1: 201, and a memory synchronization process 2: 204 that executes a memory synchronization process of the server process 2: 203. . The memory synchronization process 1: 202 operates on the CPU 2: 212, and the memory synchronization process 2: 204 operates on the CPU 4: 214.

  As described above, the active server 1 and the standby server 2 of the present embodiment are multi-core servers having a plurality of CPUs in each case, and perform a memory synchronization process for transferring data between systems, and an application process. By dividing the server process to be performed and allocating a CPU to each, it is possible to shorten the interruption time of the application service.

  Each server program and each memory synchronization program can be stored in a computer-readable recording medium such as a hard disk, flexible disk, CD-ROM, MO, DVD-ROM, or distributed via a network.

  Next, the memory synchronization method in this embodiment will be described. There are the following four examples of the memory synchronization method. In each of the following embodiments, a case where the memory synchronization processes 1: 102 and 202 synchronize the main memory allocated to the server processes 1: 101 and 201 shown in FIG. 2 will be described as an example.

<First embodiment>
FIG. 3 is a sequence of memory synchronization processing of the first embodiment.

  First, the active server 1 and the standby server 2 each perform initial processing (S11 to S14). That is, the server process 1 requests the OS to secure memory necessary for processing (S11). The OS secures a free area in the main memory and initializes the memory in the secured area (S12).

  Further, the OS activates the process of the memory synchronization process 1 for accepting the memory reference request and the memory update request of the server process 1 along with the memory reservation (S13). At this time, the OS notifies the memory synchronization process 1 of the process ID of the server process 1 and the information (address, etc.) of the allocated memory area so that the memory reference request and the memory update request can be received and controlled.

  Then, the OS notifies the server process 1 that the memory allocation has been completed (S14). At this time, the OS notifies the server process 1 of the process ID of the memory synchronization process 1.

  After the initial processing is completed in each of the active server 1 and the standby server 2, the server process 1 of the active server 1 determines that the process ID notified in S14 is necessary when memory reference is required during application processing. Based on the above, a memory reference request is sent to the memory synchronization process 1 (S15). This memory reference request includes the process ID of the server process 1 and information on the memory area to be referred to.

  The memory synchronization process 1 compares the information notified from the OS in S13 with the information included in the memory reference request, and performs access control to determine whether the memory reference request of the server process 1 is valid. For example, when the process ID notified from the OS matches the process ID included in the memory reference request and the memory area specified by the memory reference request is included in the memory area notified from the OS, it is determined to be valid. And grant access.

  If it is determined that access is permitted, the memory synchronization process 1 acquires the memory contents specified by the memory reference request from the main memory via the OS (S16).

  Then, the memory synchronization process 1 responds and sends the acquired memory contents (memory reference result) to the server process 1 (S17).

  Further, when the memory update is required during the application process, the server process 1 sends a memory update request to the memory synchronization process 1 based on the process ID notified in S14 (S18). This memory update request includes the process ID of the server process 1, information on the memory area to be updated, and the memory update content.

  If the memory synchronization process 1 performs access control similar to S16 based on the information notified from the OS in S13 and determines that access is permitted, the memory synchronization process 1 uses the memory update content of the memory update request via the OS to The main memory of the server 1 is updated (S19).

  Then, the memory synchronization process 1 transmits an update request to the memory synchronization process 1 of the standby server 2 via the NIC and the network, and waits for an update response from the standby server 2 (S20). That is, until the update response is received, the server process 1 is not notified of the memory update result, and enters a wait state. Note that the update request in S20 includes information on the memory area to be updated, memory update contents, and the like.

  The memory synchronization process 1 of the standby server 2 updates the main memory of the standby server 2 with the contents specified by the update request via the OS (S21). Then, the memory synchronization process 1 of the standby server 2 transmits an update response to the memory synchronization process 1 of the active server 1 via the NIC and the network (S22). The update response includes an update result (OK, NG, etc.) of the main memory of the standby server 2.

  The memory synchronization process 1 of the active server 1 receives the update response from the standby server 2, and notifies the server process 1 of the memory update result of S19 and the memory update result based on the update response of S22 (S23). . For example, when both the memory updates in S19 and S22 are normally performed, a memory update result indicating that the memory update has been normally completed is notified. As described above, the memory synchronization process 1 of the active server 1 receives the update response from the standby server 2 and is updated on the condition that the update of the main memory of the active server 1 is completed. Notify the update result.

  In the first embodiment described above, after the memory update of the active server 1 and the memory update of the standby server 2 are synchronized, the server process 1 of the active server 1 is notified of the completion of the memory update. To do.

  In this embodiment, after updating the main memory of the active server 1 in S19, the main memory of the standby server 2 is updated in S21. However, after updating the main memory of the standby server 2, The main memory of the active server 1 may be updated. For example, the main memory of the active server 1 may be updated after receiving an update response from the standby server 2.

<Second embodiment>
FIG. 4 is a sequence of memory synchronization processing according to the second embodiment.

Since the processes from S31 to S39 of the second embodiment shown in FIG. 4 are the same as S11 to S19 of the first embodiment shown in FIG. 3, the description thereof is omitted here.
When the memory update process 1 of the active server 1 completes the memory update of the active server 1 (S39), it notifies the server process 1 of the memory update result (S40).

  Then, the memory synchronization process 1 transmits an update request to the memory synchronization process 1 of the standby server 2 via the NIC and the network, and waits for an update response transmitted from the standby server 2 (S41). Note that the memory synchronization process 1 retains the contents of the memory update request in the memory synchronization process 1 and cannot obtain a response for a certain period in case an abnormality occurs in the subsequent processing of S42 and S43. It is desirable to reliably update the memory of the standby server 2 by resending the memory update request in S41.

  The memory synchronization process 2 of the standby server 2 updates the main memory of the standby server 2 with the content specified in the update request via the OS (S42). Then, the memory synchronization process 1 of the standby server 2 transmits an update response (update result) to the memory synchronization process 1 of the active server 1 (S43).

  In the second embodiment, the memory synchronization process 1 of the active server 1 is asynchronous without waiting for the memory update of the standby server 2, and the memory update of the active server 1 is completed by the server process 1 of the active server 1. Notify you.

<Third embodiment>
FIG. 5 is a sequence of memory synchronization processing according to the third embodiment.

  First, the active server 1 and the standby server 2 respectively perform initial processing (S51 to S54). That is, the server process 1 requests the OS to secure memory necessary for processing (S51). The OS secures a free area in the main memory and initializes the memory in the secured area (S52). Further, the OS activates the process of the memory synchronization process 1 that performs the memory synchronization process in accordance with the memory reservation (S53). Then, the OS notifies the server process 1 that the memory allocation has been completed (S54).

  After the initial processing is completed in each of the active server 1 and the standby server 2, the server process 1 of the active server 1 directly accesses the main server via the OS when memory reference is required during application processing. A memory reference request is sent to the memory (S55). This memory reference request includes information on a memory area to be referred to. The main memory acquires the memory contents of the requested area, and responds / sends the acquired memory contents (memory reference result) to the server process 1 via the OS (S56).

  Further, when a memory update is required during application processing, the server process 1 sends a memory update request directly to the main memory via the OS (S57). This memory update request includes information on the memory area to be updated and the memory update contents.

  When the main memory accepts the memory update request, it sends an update notification including the memory update contents, the update area, etc. to the OS by starting exception processing (S58). The OS sends the received update notification to the memory synchronization process 1 (S59).

  The memory synchronization process 1 detects that the main memory has been updated based on the update notification received from the OS, and sends an update request for updating the main memory of the standby server 2 via the NIC and the network. It transmits to the memory synchronization process 1 of the server 2 and waits for an update response from the standby server 2 (S60).

  The memory synchronization process 1 of the standby server 2 updates the main memory of the standby server 2 with the content specified in the update request via the OS (S61). The memory synchronization process 1 of the standby server 2 transmits an update response to the memory synchronization process 1 of the active server 1 via the NIC and the network (S62). The update response includes an update result (OK, NG, etc.) of the main memory of the standby server 2.

  The memory synchronization process 1 of the active server 1 receives the update response (memory update result) from the standby server 2, and sends the received update response to the OS as a response to the update notification in S59 (S63). The OS updates the main memory of the active server 1 based on the update notification notified in S58 (S64). The main memory notifies the server process 1 of the memory update result via the OS (S65). That is, the main memory notifies the server process 1 of the update result on the condition that the update of the main memory of the active server 1 and the main memory of the standby server has been completed.

  As described above, in the third embodiment, the memory update of the active server 1 is performed after the response of the memory update of the standby server 2 is received, so that the main memory of the active server 1 and the standby server 2 are updated. Synchronize with main memory.

  In this embodiment, after updating the main memory of the standby server 2 in S61, the main memory of the active server 1 is updated in S64. However, after updating the main memory of the active server 1, The main memory of the standby server 2 may be updated. For example, the main memory of the active server 1 may be updated before an update request is transmitted to the standby server 2.

  In this embodiment, the update to the main memory is detected based on the update notifications received from the memory and the OS in S58 and S59. However, the present invention is not limited to this. For example, monitoring for monitoring the update of the main memory. A process may be provided to detect updates to the main memory.

<Fourth embodiment>
FIG. 6 is a sequence of memory synchronization processing according to the fourth embodiment.

  The processing from S71 to S79 of the fourth embodiment shown in FIG. 6 is the same as S51 to S59 of the third embodiment shown in FIG.

  When the OS sends an update notification to the memory synchronization process 1 in S79, the OS updates the main memory of the active server 1 without waiting for an update notification response from the memory synchronization process 1 (S80). Then, the main memory notifies the server process 1 of the memory update result via the OS (S81).

  On the other hand, the memory synchronization process 1 sends an update request for updating the main memory of the standby server 2 via the NIC and the network based on the update notification (S79) received from the OS. 1 (S82).

  The memory synchronization process 1 of the standby server 2 updates the main memory of the standby server 2 with the contents specified by the update request via the OS (S83). Then, the memory synchronization process 1 of the standby server 2 transmits an update response to the memory synchronization process 1 of the active server 1 via the NIC and the network (S84). The update response includes an update result (OK, NG, etc.) of the main memory of the standby server 2.

  The memory synchronization process 1 of the active server 1 receives the update response (memory update result) from the standby server 2, and sends the received update response to the OS as a response to the update notification in S79 (S65).

  In the fourth embodiment, the main memory of the active server 1 is updated asynchronously without waiting for the memory update of the standby server 2.

  In the present embodiment described above, a memory synchronization process corresponding to the server process is provided separately from the server process that performs application processing. As a result, the application program of the present embodiment does not need to develop a program that takes into account memory synchronization processing, facilitates program development, and improves program development efficiency.

  The active server 1 and the standby server 2 of the present embodiment are multi-core servers each having a plurality of CPUs in each case, and the CPU is assigned to each of the memory synchronization process and the server process separately. Thereby, it is possible to shorten the interruption time of the application service.

  In addition, this invention is not limited to the said embodiment, Many deformation | transformation are possible within the range of the summary.

  For example, in the first to fourth embodiments described above, the memory synchronization process for executing the memory synchronization processing is a process separate from the OS and implemented as middleware as viewed from the server process. However, in these embodiments, The function is not limited, and a function equivalent to the memory synchronization process may be implemented as a part of the OS or hardware.

  In the first to fourth embodiments described above, an example is shown in which the memory synchronization process is started when the server process makes a memory reservation request. However, the present invention is not limited to this, and the memory synchronization process is started in advance. You may keep it.

  In this embodiment, the server process and the memory synchronization process correspond to each other on a one-to-one basis. However, the present invention is not limited to this, and one memory synchronization process is performed for a plurality of server processes. A process may be assigned, or a plurality of memory synchronization processes may be assigned to one server process. Further, the allocation relationship between the server process and the memory synchronization process may not be fixed, but may be dynamically switched according to the processing load status.

  In the third and fourth embodiments, an example in which a memory update request is detected in the main memory which is hardware has been described, but access to the memory is monitored in middleware such as an OS or a memory synchronization process. The memory update request from the server process may be detected.

  In the second and fourth embodiments, the example in which the memory synchronization process performs the memory update process of the standby server 2 asynchronously each time the memory of the active server 1 is updated has been described. However, the present invention is not limited to this. Instead, the main memory of the standby server 2 may be updated synchronously at a time when a certain amount of memory update of the active server 1 has accumulated, or the update status may be checked periodically and synchronized at once. May be.

1: Active server 2: Standby server 11, 21: CPU
12, 22: Main memory 13, 23: Disk controller 14, 24: Disk 15, 25: NIC
101, 201: Server process 1
102, 202: Memory synchronization process 1
103, 203: Server process 2
104, 204: Memory synchronization process 2
105, 205: OS
9: Network

Claims (7)

A duplex system comprising an active server and a standby server,
The active server is
Application processing means for performing application processing;
A second update request for updating the main memory of the standby server is transmitted to the standby server in response to a first update request generated by the application process for updating the main memory of the active server. First synchronizing means to
The standby server is
A duplex system comprising: a second synchronization unit configured to update a main memory of the standby server when receiving the second update request from the active server. The duplex system according to claim 1, wherein
The application processing means sends the first update request to the first synchronization means,
The first synchronization means receives the first update request, updates the main memory of the active server, and transmits the second update request to the standby server. system. The duplex system according to claim 1, wherein
The application processing means sends the first update request to the main memory of the active server,
The duplex system according to claim 1, wherein the first synchronization unit detects an update to the main memory and transmits the second update request to the standby server. The duplex system according to claim 2, wherein
The first synchronization means receives an update response to the second update request from the second synchronization means, and updates the application processing means when the update of the main memory of the active server is completed. A duplex system characterized by notification of results. The duplex system according to claim 3, wherein
The duplex system is characterized in that the main memory notifies the application processing means of an update result when the update of the main memory of the active server and the main memory of the standby server is completed. A duplex system according to any one of claims 1 to 5, wherein
The active server is
A first arithmetic processing unit assigned to the application processing means;
A second arithmetic processing unit assigned to the first synchronization means,
Dual system characterized by.   A memory synchronization method for synchronizing a main memory of the active server and a main memory of the standby server, which is performed by the duplex system according to any one of claims 1 to 6.

Images