JP5232814B2 - Cache system and cache access method - Google Patents

Description

  The present invention relates to a cache system and a cache access method.

  Conventionally, a user's terminal accesses a server connected via a network, and uses a web screen provided from this server to a database on the network (hereinafter referred to as “DB”). Thus, a system for performing data access such as data reference is realized.

  FIG. 25A is a diagram showing a configuration of a service order system which is an example of such an existing system. In the service order system shown in the figure, the web service order device provides a web screen for accessing data in the customer DB managed by the service order management device to a terminal connected via the network. To do. When the user inputs a data reference request to the terminal using this web screen, the web service order device receives the data reference request input to the terminal, converts the format of the received data reference request into a command or the like. Sent to the service order management device. Data read from the customer DB by the service order management device according to the received command is notified to the terminal via the web service order system.

  In the service order system as shown in FIG. 25A, the user terminal cannot acquire the data in the customer DB while the service order management apparatus is stopped due to maintenance or the like. Therefore, as shown in FIG. 25B, it is conceivable to provide a cache system for storing data in synchronization with the customer DB between the web service order apparatus and the service order management apparatus. Thereby, even while the service order management apparatus is stopped, the web service order apparatus can read the data requested to be referenced from the cache system and provide it to the terminal.

  However, as described above, even if the cache system is provided, if the cache system itself stops, data cannot be provided to the terminal. Therefore, it is necessary to ensure the redundancy of the cache system.

When ensuring redundancy, it is common to multiplex devices. FIG. 26 shows a conventional system configuration in which redundancy is ensured by multiplexing cache servers.
FIG. 26A shows a configuration in which a cluster is used and the standby cache server is cold standby (Cold.SBY). In the system having the configuration shown in the figure, a DB is shared by a storage area network (SAN) in an ACT system and a standby system cache server, and system switching is performed using clustering software.
FIG. 26B shows a configuration in which a cluster is used and the standby cache server is set to hot standby (Hot.SBY). Similarly, in the system having the configuration shown in the figure, the DB is shared by the SAN between the act and standby cache servers, and the system is switched using the clustering software.
FIG. 26C illustrates a configuration in which a DAS (Direct Attached Storage) DB is provided in each system cache server and the standby cache server is a cold standby. In FIG. 26C, the quorum server holds the state of each system and performs system switching.

  Further, Patent Documents 1 and 2 disclose techniques for providing data access without interruption. Patent Document 1 describes that a plurality of nodes to which the same data is assigned are connected to the outside via a load balancer, and the load balancer selects the optimum node by monitoring the state of the node. Yes. Also, in Patent Document 2, each of redundant computers operating as a primary system and a secondary system exchanges healthy check information as a result of checking their own state, and switches between the primary system and the secondary system. It is described.

JP 2004-30204 A JP 2008-226153 A

The existing technology for ensuring redundancy as described above has the following problems.
That is, when system switching is performed in the configuration of FIG. 26A, since the standby system is cold standby activated, it may take several tens of minutes for system switching, and data cannot be provided during this time. Further, during SAN firmware upgrade, it is necessary to stop a DBMS (DataBase Management System) accessing the SAN for several tens of minutes, and data cannot be provided during this time. In addition, the DB shared disk becomes a single point of failure (SPOF), and when the shared disk fails, the entire system falls into a failure.

Further, when system switching is performed in the configuration of FIG. 26B, it is necessary to stop the DBMS accessing the SAN for several tens of minutes during the SAN firmware upgrade, as in FIG. The disk becomes SPOF. In addition, when extending the redundant configuration, it is necessary to cancel the relationship between the systems. That is, the DBMS and clustering software of both systems are stopped and the setting change operation is performed, so that the data access service cannot be provided during that time.
As described above, in the form in which the DB is shared between the systems as shown in FIGS. 26A and 26B, it is difficult to provide data continuously.

  In the configuration of FIG. 26 (c), when system switching is performed, the standby system is cold standby activated, so it may take several tens of minutes for system switching. During this time, the data access service cannot be provided. The quorum server that maintains the state of each system and performs system switching becomes SPOF.

  In the technology of Patent Document 1, the load balancer transmits a request received from the outside to both of the duplicated node computers, and instructs the node computer that has returned the temporary execution result for the request to perform the main execution, The other node computer is instructed to be invalid, and each time a request is received by the node computer, a process of executing a temporary execution and discarding the result must be implemented. The technique of Patent Document 2 is for switching between a primary system and a secondary system in a redundant computer, and the redundant computer must always be aware of the state of the other computer.

  The present invention has been made in consideration of such circumstances, and the object thereof is to use a multiplexed cache device and continue constantly without applying a load for system switching to the cache device. To provide a cache system and a cache access method capable of providing data.

The present invention has been made to solve the above-described problem, and includes a plurality of cache devices that cache original data, and an access distribution device that manages to which of the plurality of cache devices a received signal is distributed. In the cache system, each of the plurality of cache devices receives a data access request distributed from the access distribution device, and responds to the original data cached by the own cache device according to the received data access request. A reference unit that performs access processing; and a transfer unit that reads the original data from the original data management device and updates the original data cached by the own cache device, and the access distribution device includes a plurality of act states. Acquire the operational status of each of the reference unit and the transfer unit of the cache device And a cache check device to which a received signal is allocated from among the plurality of cache devices based on the operating states of the reference unit and the transfer unit of the plurality of cache devices acquired by the service check unit. An operating system determination processing unit for determining
According to the above invention, a plurality of cache devices, each realized by a single server (casing), perform access processing to the original data cached in response to a data access request from a higher-level application. The reference unit and the transfer unit that updates the original data cached by the original data held by the original data management device are separated, and the distribution device is configured to transfer the reference unit and the transfer in each cache device. The distribution destination of the signal, that is, the active cache device is determined according to the situation such as whether both of the units are normal or whether one or both of them has failed.
As a result, when the distribution apparatus receives a data access request from a higher-level application, the distribution apparatus first selects a cache apparatus in which the reference unit operates normally, and if the selected cache apparatus is one, the selected cache apparatus is selected. If the device is the distribution destination and there are multiple cache devices selected, you can select the distribution destination from among the cache devices that have the normal transfer unit in the same chassis, and the transfer unit of the selected cache device If there is also a failure, a sorting destination can be arbitrarily selected from the selected ones. Therefore, it is possible to provide access to the original data continuously from a plurality of cache system devices, and to provide the latest original data as much as possible.
In addition, since the reference unit and the transfer unit are separated, it becomes easy to shift to a configuration in which the reference unit and the transfer unit are mounted in separate cases in the future. The traffic of the reference unit increases with the increase in the number of users who use the upper service application and the types of upper service applications, while the traffic item of the transfer unit is added with data items that are referenced from the upper service application. In view of the increase in the number of servers, in the future, the server that executes the reference unit may be shifted to the entire operation system having n units, and the server that executes the transfer unit may be shifted to the operation standby system having 1 + 1 unit. Desirably, the cache system of the present invention can be used as the degenerated configuration.
The n-unit configuration is a redundant configuration in which a plurality of identical systems and functional units are prepared, and all are active. The 1 + 1 configuration is a configuration in which two identical systems and functional units are prepared, Is a redundant configuration with the active system, and the other is a standby system. The active standby system operates when a functional unit is distributed to multiple servers and the functional unit is divided into an active system and a standby system. It is a method.

Also, the present invention is the above-described cache system, wherein the operational system determination processing unit determines that the reference unit is normal when a plurality of the reference units are determined to be normal based on an operational state of the reference unit. The cache device including the transfer unit that is determined to be normal based on an operation state of the transfer unit is determined as a distribution destination from the plurality of cache devices provided.
According to the above invention, the allocating device determines not only that the reference unit is operating normally but also whether the transfer unit that synchronizes the original data is operating normally, so that the operational cache device Select.
If the transfer unit is operating normally, the original data cached in the cache device is correctly synchronized, so the transfer unit also operates normally from among the cache devices in which the reference unit is operating normally. By selecting the cache device as the distribution destination, it is possible to always provide the latest original data among the original data that can be accessed to the upper application.

Further, the present invention is the above-described cache system, wherein each of the transfer units of the plurality of cache devices transmits a system state identification signal to the distribution device, and is transmitted from the distribution device by itself. When the system status identification signal is received, an information transfer processing unit that reads the original data from the original data management device and updates the original data cached by the own cache device, and another cache device caches And an intersystem transfer processing unit that updates the original data read from the original data read by the information transfer processing unit, wherein the operational system determination processing unit of the access distribution device determines the cache determined as the distribution destination The system state identification signal received is transmitted to the apparatus.
According to the above invention, the transfer unit of each cache device transmits a system state identification signal, and the transmitted system state identification signal is distributed to the cache device that the distribution device determines to be the active system. The transfer unit that has received the system state identification signal transmitted by itself determines that the own system is the active system, updates the original data cached by the own system with the original data held by the original data management device, Furthermore, the original data cached by another cache device is also updated.
Since the allocating device determines the signal allocating destination according to the operation status of the reference unit and the transfer unit in each cache device, not the signal type or destination, the allocating device uses the same allocation destination as the data access request from the upper application. A system state identification signal is also distributed. Therefore, when the system state identification signal transmitted by itself is received, that is, the transfer unit of the cache device that processes the data access request first synchronizes with the original data held by the original data management device, so The latest original data can be provided to the host application. For example, when compared with a synchronization procedure in which the transfer unit of the standby cache device synchronizes with the original data management device and subsequently performs intersystem synchronization with the active cache device, the active cache device caches The original data being updated is early and the original data cached by the active cache device cannot be updated due to a failure during synchronization between systems. There is no such thing as offering. Further, after synchronization with the original data management device, inter-system synchronization is performed with other cache devices, so that inconsistency of the original data between the cache devices does not occur.

Also, the present invention is the above-described cache system, wherein the information transfer processing unit caches itself from the original data management device when receiving the system state identification signal transmitted by itself. Read the original data whose update date and time in the original data management device is newer than the original data, update the original data cached by the own cache device, and the inter-system transfer processing unit is cached by the own cache device. Of the original data, the original data cached by the other cache device is updated by the original data having a newer update date and time in the original data management device than the original data cached by the other cache device. It is characterized by updating.
According to the above invention, the transfer unit of the active cache device updates the original data cached by itself and the original data cached by another cache device based on the update date and time in the original data management device. Update.
As a result, the active cache device is switched, and the Kish device that is the new allocation destination of the data access request synchronizes with the original data management device or when synchronizing with the original data of another system. There is no update leak. Therefore, it is possible to provide the latest original data while ensuring unity.

The present invention is the above-described cache system, wherein the transfer unit further includes a transfer data storage unit that stores the original data cached by the own cache device, and the information transfer processing unit includes the original data The original data read from the management device is written to the transfer data storage unit, and the reference unit reads the reference data storage unit that stores the original data cached by the own cache device and the transfer data storage unit. An in-system synchronization processing unit that writes the original data to the reference data storage unit, and a reference data access unit that performs an access process to the original data stored in the reference data storage unit in response to the received data access request It is characterized by comprising.
According to the above invention, each of the reference unit and the transfer unit includes the storage unit that stores the cached original data, and the transfer unit writes the original data read from the original data management device to the storage unit that the transfer unit has and caches the original data. Then, the reference unit updates the original data cached in its own storage unit with the original data cached in the storage unit of the transfer unit.
With the above configuration, it is possible to completely separate the reference unit and the transfer unit. Therefore, in the future, the server that executes the reference unit will be changed to all operation systems having n units, and the server that executes the transfer unit will be 1 + 1 in the future. It becomes easy to shift to a configuration standby system.

  The present invention is also a cache access method used in a cache system including a plurality of cache devices that cache original data and an access distribution device that manages to which of the plurality of cache devices a received signal is distributed. Each of the plurality of cache devices receives the data access request distributed from the access distribution device, and performs access processing to the original data cached by the own cache device according to the received data access request And a transfer unit that reads the original data from the original data management device and updates the original data cached by the own cache device. In the access distribution device, the service check unit is in an act state. A plurality of the reference units of the cache device; A status acquisition process for acquiring the operating status of each of the sending units, and an active system determination processing unit, based on the operating statuses of the reference unit and the transfer unit of the cache devices acquired in the status acquiring process, And an operational system determination process for determining the cache device to which a received signal is allocated from among the cache devices.

According to the present invention, the allocating device is configured to distribute the signal according to a situation such as whether both the reference unit and the transfer unit in each cache device are normal, or a failure has occurred in either or both, In other words, since the operating cache device is determined, it is possible to continuously provide access to the original data and to provide the latest original data as much as possible.
Further, according to the present invention, the cache device in which the transfer unit also operates normally is selected as the distribution destination from among the cache devices in which the reference unit operates normally. It is possible to always provide the latest original data to the upper application.
Further, according to the present invention, the transfer unit of the same cache device as the reference system that processes the data access request first synchronizes with the original data held by the original data management device. It can be provided to a host application, and synchronization with the other system is performed after synchronization with the original data management apparatus, so that inconsistency of the original data between the cache apparatuses does not occur.
Further, according to the present invention, the cache device to which the data access request is distributed can update the original data cached by the cache device based on the update date and time in the original data management device, and the other cache device can Since the synchronization with the original data is performed, it is possible to provide the latest original data that ensures unity.
In addition, according to the present invention, since the reference unit and the transfer unit are separated, it becomes easy to move to a configuration in which the reference unit and the transfer unit are mounted in different cases in the future. This separation is further facilitated by having a storage part in each part. Therefore, the cache system according to the present invention can be used as a degenerate configuration when the server that executes the reference unit shifts to the entire operation system of n units and the server that executes the transfer unit shifts to the operation standby system of 1 + 1 unit. It becomes possible.

It is a figure which shows the connection structure of the cache system by one Embodiment of this invention. It is a figure which shows the operation | movement outline | summary of the cache system by the embodiment. It is a block diagram which shows the internal structure of each apparatus by the embodiment. It is a figure which shows the system state confirmation sequence of the cache system by the embodiment. It is a figure which shows the flow of a process between the function parts of each apparatus in the system state confirmation by the embodiment. It is a figure which shows the own system state monitoring process flow in the reference part 11 and the transfer part 15 of the cache server 1 of both systems by the same embodiment. It is a figure which shows the self-system state monitoring result return process flow in the reference part 11 and the transfer part 15 of the cache server 1 of both systems by the embodiment. It is a figure which shows the health check processing flow of the load balancer 2 by the embodiment. It is a figure which shows the distribution process sequence at the time of the data reference request reception by the embodiment. It is a figure which shows the distribution process sequence at the time of the system state identification signal reception by the embodiment. It is a figure which shows the distribution processing flow in the load balancer 2a by the embodiment. It is a figure which shows the distribution state table | surface by the embodiment. It is a figure which shows the processing flow of the distribution process in the load balancer 2b by the embodiment. It is a figure which shows the original data synchronization method which considered application to the embodiment. It is a figure which shows the system synchronization method which considered application to the same embodiment. It is a figure which shows the data synchronization sequence of the cache system by the embodiment. It is a figure which shows the example of data of each apparatus used for the data synchronous process by the embodiment. It is a figure which shows the flow of the data update in the original data synchronous process by the embodiment. It is a figure which shows the flow of the data update in the data synchronization process between systems by the same embodiment. It is a figure which shows the system state identification processing flow in the transfer part 15 of the cache server 1 of both systems by the same embodiment. It is a figure which shows the synchronous processing flow with the original data in the transfer part 15 of the cache server 1 of both systems by the embodiment. It is a figure which shows the data synchronization processing flow in the system in the transfer part 15 of the active system cache server 1 by the embodiment. It is a figure which shows the operation | movement outline | summary using the general purpose interface by the embodiment. It is a figure which shows the general purpose interface processing flow of the cache server 1 by the embodiment. It is a figure which shows the conventional data access system. It is a figure which shows the conventional system structure which ensured the redundancy by duplication.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[1. Constitution]
FIG. 1 is a diagram showing a connection configuration of a cache system according to an embodiment of the present invention. In the figure, the cache system includes a cache interface device multiplexed by the cache server 1 and load balancers 2a and 2b connected to the cache server 1 respectively. The load balancer 2a is further connected to an application device 4 that executes the higher level service application using the original data, and the load balancer 2b is further connected to an original data management device 3 that stores the original data referred to by the higher level service application. The Although only one application device 4 is shown in the figure, a plurality of devices can be connected.

  The cache server 1 internally caches information necessary for the higher-level service application among the original data stored in the original data management device 3. In response to the data reference request transmitted from the application apparatus 4 in accordance with the execution of the higher-level service application, the cache server 1 retrieves the original data read and cached from the original data management apparatus 3 and returns the result. Since the cache server 1 takes the form of a multiplexed configuration, the service can be continuously maintained for the upper service application.

  Hereinafter, one of the cache servers 1 is a 0 system, the other is a 1 system, the 0 system cache server 1 is referred to as a cache server 1-0, and the 1 system cache server 1 is referred to as a cache server 1-1. One system is collectively referred to as “both systems”. Further, the load balancers 2a and 2b are collectively referred to as a load balancer 2. Note that the load balancer 2 is duplicated, and system switching can be performed in a short time even when a failure or the like occurs.

FIG. 2 is a diagram showing an outline of the operation of the cache system shown in FIG.
In this embodiment, the cache servers 1 of both systems do not have an act system and a standby system, and both the cache servers 1-0 and 1-1 operate as an act system, and a load balancer 2 as a network device. Determines whether the received signal is distributed to the cache servers 1-0 and 1-1. Here, the cache server 1 on which the load balancer 2 distributes signals is described as the active system, and the cache server 1 on the non-active system side is described as the standby system.

  Each cache server 1 is a single server including a reference unit 11 and a transfer unit 15. That is, the reference unit 11 and the transfer unit 15 operate in the same casing. The reference unit 11 uses original data cached in a database (hereinafter referred to as “DB”) provided in the reference unit 11 and responds to a data reference request transmitted from the application device 4 by executing a higher-level service application. Respond. The transfer unit 15 identifies whether the own system is determined to be the active system from the load balancer 2, and if it is determined to be the active system, the original data cached in the DB included in the transfer unit 15 and Data synchronization with the original data stored in the original data management device 3 is performed, and further synchronization with the original data cached in the DB of the transfer unit 15 of the other system is performed. The reference unit 11 periodically reads the original data synchronized from the DB of the transfer unit 15 and updates the DB included in the reference unit 11 to perform data synchronization with the DB of the transfer unit 15.

The load balancer 2 periodically sends a health check signal to the cache server 1 to perform a health check that monitors the state of the reference unit 11 and the transfer unit 15 of each system. Based on the monitoring result, the load balancer 2 Determine the standby system.
The load balancer 2a has a virtual IP address A on the application device 4 side, and stores a physical IP address C0 of the cache server 1-0 and a physical IP address C1 of the cache server 1-1. When the destination signal is received, the received signal is sent with the physical IP address C0 or C1 of the cache server 1 on the operation side as the destination. At this time, a port number corresponding to the reference unit 11 is added to the physical IP address. The port numbers of the reference units 11 of the cache servers 1-0 and 1-1 are PS0 and PS1, respectively.
Similarly, the load balancer 2b has a virtual IP address B on the original data management device 3 side, and stores the physical IP address C0 of the cache server 1-0 and the physical IP address C1 of the cache server 1-1. When a signal destined for the virtual IP address B is received, a signal received with the physical IP address C0 or C1 of the cache server 1 on the operation side as the destination is transmitted. At this time, a port number corresponding to the transfer unit 15 is added to the physical IP address. The port numbers of the transfer units 15 of the cache servers 1-0 and 1-1 are PT0 and PT1, respectively.

  In the same cache server 1, the reference unit 11 and the transfer unit 15 can use different physical IP addresses. In this case, when the load balancer 2a receives a signal destined for the virtual IP address A, the load balancer 2a transmits a signal received with the physical IP address corresponding to the reference unit 11 of the cache server 1 on the operation side as the destination. Similarly, when the load balancer 2b receives a signal destined for the virtual IP address B, the load balancer 2b transmits a signal received with the physical IP address corresponding to the transfer unit 15 of the cache server 1 on the operation side as the destination.

  As described above, the load balancer 2a transmits the data reference request received from the application device 4 to the reference unit 11 of the cache server 1 determined to be the active system. Therefore, when the cache server 1 receives a data reference request without being aware of the system, the cache server 1 reads and stores data stored in a DAS (Direct Attached Storage) DB included in its reference unit 11. Good. Since the cache server 1 that is determined to be the standby system by the load balancer 2a is also operating in the same state as the cache server 1 that is determined to be the active system, when the load balancer 2a determines to switch the active system However, it is not necessary to be aware of this, and it is sufficient to respond to the received data reference request. Thus, the two independent cache servers 1 look as if they are one device from the outside, and the determination of the service route is left to the load balancer 2 to be controlled.

  However, in the cache system of the present embodiment, synchronization with the original data stored in the original data management device 3 and data synchronization with the standby cache server 1 are started from the transfer unit 15 of the active cache server 1. To do. Therefore, the transfer unit 15 of the cache server 1 performs a system state identification process that inquires the load balancer 2b about its own system state. In the system state identification process, the load balancer 2 uses the distribution of a signal destined for the virtual IP address to the active cache server 1. That is, when the transfer unit 15 of the cache server 1 transmits a system state identification signal to the virtual IP address B and receives the signal from the load balancer 2b, the transfer unit 15 determines that the system is an active system.

  The load balancer 2 determines the active system and the standby system based on the monitoring results of the statuses of the reference unit 11 and the transfer unit 15 of each system based on the health check, but both the reference unit 11 of the cache server 1 of both systems operate. When it is possible (no failure has occurred), the cache server 1 in which the transfer unit 15 provided in the same casing can be operated is determined as the active system.

For example, when a certain function unit is distributed to multiple servers, regardless of the operation status of the transfer unit 15 as in all operation methods in which all relevant function units on all servers operate as active systems by load balancing. If the reference unit 11 is operable, it is assumed that the cache server 1 can be accessed from the upper service application. In this case, in the cache server 1 in which the operation state of the transfer unit 15 is the standby system, synchronization with the original data acquired from the original data management apparatus 3 may be delayed (synchronization delay between systems), or the cached original data Data may not be updated (intersystem synchronization failure). At this time, if the data reference request is distributed to the cache server 1 in which the transfer unit 15 is a standby system, a problem of returning old data occurs.
Therefore, the load balancer 2a confirms whether or not the reference unit 11 of the cache servers 1 of both systems is operable, and at the same time confirms whether or not the transfer unit 15 of the cache servers 1 of both systems is operable. When the reference unit 11 of both cache servers 1 is operable, the data reference request received from the application device 4 to the reference unit 11 in which the transfer unit 15 provided in the same casing is operating normally. Sort out.

  When the load balancer 2a determines the distribution destination as described above, for example, the reference unit 11 and the transfer unit 15 of the cache server 1-0 are in the operating state, and the reference unit 11 of the cache server 1-1 is in the operating state. When the transfer unit 15 is in a standby state, a data reference request from the application device 4 is distributed to the reference unit 11 of the cache server 1-0. At this time, when a failure such as an application failure, a DB failure, or a network failure with the load balancer occurs in the reference unit 11 of the cache server 1-0, system switching is performed. When the conventional method is applied, it is conceivable that the cache server 1 synchronizes data if the transfer unit 15 is in an operation state regardless of the operation state of the reference unit 11. However, even if the data reference request is assigned to the reference unit 11 of the cache server 1-1, if the system switching is not performed in the operation state of the transfer unit 15 of the cache server 1-0, In comparison, there is a possibility that synchronization with the original data acquired from the original DB may be delayed (lag in intersystem synchronization), or cached original data may not be updated (intersystem synchronization failure). The service provision level will be lowered.

  Therefore, in the present embodiment, the transfer unit 15 of the cache server 1 uses the reference unit 11 in the same housing as the distribution destination of the data reference request at the start of the original data synchronization process with the original data management device 3. Check if it exists. Therefore, the transfer unit 15 of the cache server 1 transmits a system state identification signal, and the load balancer 2b determines a distribution destination in the same manner as the load balancer 2a. The system status identification signal transmitted by itself is received from the load balancer 2b, that is, the transfer unit 15 in the same housing as the cache server 1 to which the data reference request is allocated acquires the original data from the original data management device 3. And synchronize. In this way, the transfer unit 15 in the operating state performs the original data synchronization process when the reference unit 11 in the same casing is the distribution destination, and transfers when the reference unit 11 in another casing is the distribution destination. 15 system switching is performed.

FIG. 3 is a block diagram showing the internal configuration of the cache server 1, the load balancer 2, and the original data management device 3.
The reference unit 11 performs a health check with the load balancer 2 and responds to a data reference request from a higher-level service application, and within the same system, that is, between the transfer unit 15 included in the same cache server 1 To synchronize data. The reference unit 11 includes a timer 111, a monitoring processing unit 112, a monitoring result notification processing unit 113, a reference data management unit 114, an information reference processing unit 115, and an in-system synchronization processing unit 116.

  The monitoring processing unit 112 is periodically started by the timer 111, and confirms the occurrence of each process in the own reference unit 11, and checks whether the DB included in the reference data management unit 114 is active or not. The system state monitoring is performed, and the result is stored in the monitoring result storage unit 121 provided inside. The monitoring result notification processing unit 113 periodically receives a health check signal from the load balancer 2 and transmits a response according to the in-system state monitoring result of the own reference unit 11 held in the monitoring result storage unit 121.

The reference data management unit 114 includes a reference data access unit 122 and a reference data storage unit 123. The reference data access unit 122 can be realized by existing DBMS (DataBase Management System) software, and reads and writes cache data stored in a reference data storage unit 123 as a DB.
The intra-system synchronization processing unit 116 performs data synchronization between the original data stored in the reference data storage unit 123 and the original data stored in the DB of the transfer unit 15.

  The information reference processing unit 115 receives the data reference request transmitted by the application apparatus 4 and outputs the data reference request to the reference data management unit 114, and returns the data read by the reference data management unit 114 based on the data reference request.

  The transfer unit 15 performs a health check with the load balancer 2, confirms whether the own system is an active system, and performs data synchronization with the original data management device 3 if the system is an active system. Further, the transfer unit 15 performs data synchronization with the transfer unit 15 in the same system, that is, in the same cache server 1. The transfer unit 15 includes a timer 151, a monitoring processing unit 152, a monitoring result notification processing unit 153, a transfer data management unit 154, a system state identification unit 155, a synchronization information storage unit 156, an information transfer processing unit 157, and an intersystem transfer processing unit 158. And an in-system synchronization processing unit 159.

  The monitoring processing unit 152 is periodically started by the timer 151, and confirms the occurrence of each process in the transfer unit 15 of the own system and the alive check of the DB included in the transfer data management unit 154. System state monitoring is performed, and the result is stored in the monitoring result storage unit 161 provided therein. The monitoring result notification processing unit 153 periodically receives a health check signal from the load balancer 2 and transmits a response according to the in-system state monitoring result in the own transfer unit 15 held in the monitoring result storage unit 161. .

The transfer data management unit 154 includes a transfer data access unit 162 and a transfer data storage unit 163. The transfer data access unit 162 can be realized by existing DBMS software, and reads and writes cache data stored in the transfer data storage unit 163 as a DB.
The intra-system synchronization processing unit 159 performs original data synchronization between the data stored in the transfer data storage unit 163 and the data held by the reference unit 11.

  The synchronization information storage unit 156 writes the original data read from the original data management device 3 to the transfer data storage unit 163 but has not yet been synchronized with the other cache server 1 and the data. Stores an update information table (hereinafter referred to as “TBL”) in association with the date and time information updated in the original data management apparatus 3. Further, the synchronization information storage unit 156 stores final update information TBL indicating the last look, which is the latest update date / time, of the update date / time in the original data management device 3 of the original data that has been written to the transfer data storage unit 163. To do.

  The system state identification unit 155 inquires whether the load balancer 2 recognizes its own system as an active system or determines that it is a standby system. When the local system is recognized as the active system, the information transfer processing unit 157 converts the original data whose update date and time is after the last look set in the last update information TBL in the synchronization information storage unit 156 to the original data management device 3 and writes the read original data to the update information TBL in the synchronization information storage unit 156 and writes the read original data to instruct the transfer data management unit 154 to update the transfer data storage unit 163. Among the original data stored in the update information TBL in the synchronization information storage unit 156, the inter-system transfer processing unit 158 updates the date and time after the last look set in the final update information TBL of the cache server 1 of the other system. Is written to the transfer data storage unit 163 of the cache server 1 of the other system, and when the writing is successful, the latest update date / time of the original data management device 3 of the original data that has been successfully written is The last look-up information TBL of the other cache server 1 is written as the last look.

  Hereinafter, “−0” is added to each part of the cache server 1 included in the cache server 1-0, and “−1” is added to the part of the cache server 1 included in the cache server 1-1. Add and describe. Further, the fact that each unit is provided in the 0-system cache server 1-0 is simply referred to as “0-system”, and the fact that each unit is provided in the 1-system cache server 1-1 is also simply referred to as “1-system”. . For example, the monitoring processing unit 112 included in the cache server 1-0 is described as “monitoring processing unit 112-0” or “0-system monitoring processing unit 112-0”, and the monitoring processing included in the cache server 1-1. The unit 112 is described as “monitoring processing unit 112-1” or “system 1 monitoring processing unit 112-1”.

In the figure, the load balancer 2 includes a timer 21, a service check unit 22, and an active system determination processing unit 23.
The service check unit 22 is periodically started by the timer 21, and the physical IP address of both systems cache server 1 and the port number of the reference unit 11, the physical IP address of both systems cache server 1 and the port number of the transfer unit 15. Is used as a destination, and a health check signal is transmitted, and a result indicating whether or not a response is returned from the reference unit 11 and the transfer unit 15 of the cache servers 1 of both systems is written in the storage unit 221 provided therein. Whether the response is returned or not can be identified as normal or abnormal.

  When receiving a signal whose destination is the virtual IP address of the own load balancer 2, the active system determination processing unit 23 stores both the reference unit 11 and the transfer unit 15 of the cache servers 1 of both systems stored in the storage unit 221. The active cache server 1 is determined based on the health check response reception status, that is, the normal or abnormal status. The active system determination processing unit 23 distributes the received signal to the reference unit 11 or the transfer unit 15 of the active cache server 1 by transmitting the received signal to the physical IP address of the cache server 1 determined to be the active system.

Hereinafter, each part of the load balancer 2 included in the load balancer 2a is described by adding “a” to the reference numeral, and each part of the load balancer 2 included in the load balancer 2b is described by adding “b” to the reference numeral. For example, the service check unit 22 included in the load balancer 2a is described as a service check unit 22a, and the service check unit 22 included in the load balancer 2b is described as a service check unit 22b.
Note that the load balancer 2a and the load balancer 2b may be realized by different network devices or may be realized by the same network device.

  In FIG. 1, the original data management device 3 includes a data management unit 31, an update history storage unit 32, and an original data storage unit 33. When the original data stored in the original data storage unit 33 is read / written and the original data is updated, the data management unit 31 associates the updated original data with the update date / time information of the original data. The update history is written into the update history storage unit 32.

[2. System status confirmation process]
The operation of the system state confirmation process in the cache system of this embodiment will be described.

[2.1 System status check sequence]
FIG. 4 is a diagram showing a system state confirmation sequence of the cache system. In the figure, the reference unit 11 and the transfer unit 15 of the cache servers 1-0 and 1-1 periodically perform in-system state monitoring processing of the own function unit (steps S105-1 to S105-4). . Note that it is not necessary to synchronize the timing between the reference units 11-0 and 11-1 and the transfer units 15-0 and 15-1.

  On the other hand, the load balancer 2 periodically starts a health check. That is, the load balancer 2 has the physical IP address C0 of the cache server 1-0 and the port number PS0 of the reference unit 11-0, the physical IP address C0 of the cache server 1-0 and the port number PT0 of the transfer unit 15-0, the cache A health check signal is transmitted with the physical IP address C1 of the server 1-1 and the port number PS1 of the reference unit 11-1, the physical IP address C1 of the cache server 1-1 and the port number PT1 of the transfer unit 15-1 as destinations ( Steps S110 and S115-1 to S115-4).

The 0-system reference unit 11-0 executes an in-system state monitoring result return process (step S120-1), and the internal system state monitoring result for the own system reference unit 11-0 confirmed in step S105-1 is obtained. In response, a response is returned to the load balancer 2 (step S125-1). The 0-system transfer unit 15-0 executes an in-system state monitoring result return process (step S120-2), and the in-system state monitoring result for the own-system transfer unit 15-0 confirmed in step S105-2. In response, a response is returned to the load balancer 2 (step S125-2). Similarly, the 1-system reference unit 11-1 executes an in-system state monitoring result return process (step S120-3), and the in-system state regarding the own system reference unit 11-1 confirmed in step S105-3. A response is returned to the load balancer 2 according to the monitoring result (step S125-3). Further, the 1-system transfer unit 15-1 executes an in-system state monitoring result return process (step S120-4), and the in-system state monitoring of the own system transfer unit 15-1 confirmed in step S105-4. A response is returned to the load balancer 2 according to the result (step S125-4).
The load balancer 2 confirms the response state to the health check signal from each of the reference unit 11 and the transfer unit 15 of the cache servers 1 of both systems, and holds the confirmation result (step S130).

  Thereafter, the reference unit 11 and the transfer unit 15 of the cache servers 1-0 and 1-1 periodically start the in-system state monitoring process of each functional unit as in Steps S105-1 to S105-4 ( Steps S135-1 to S135-4, S140-1 to S140-4). Further, the load balancer 2 periodically transmits a health check signal to the reference unit 11 and the transfer unit 15 of the cache servers 1-0 and 1-1 in the same manner as the processes of steps S110 and S115-1 to S115-4. (Steps S145, S150-1 to S150-4). Thereby, processing similar to steps S120-1 to S120-4, S125-1 to S125-4, and S130 is performed (steps S1555-1 to S155-4, S160-1 to S160-4, S165).

  The processing of the load balancer 2 described above is performed for each of the load balancers 2a and 2b.

[2.2 Process flow between functional units of each device in system status confirmation processing]
FIG. 5 is a diagram illustrating a process flow between functional units of each device in the system state confirmation process.
In the reference unit 11 of the cache servers 1 of both systems, the monitoring processing unit 112 confirms the occurrence of each process, holds the result in the monitoring result storage unit 121 (step S305-1), and then performs dummy SQL. Is issued to confirm the operation of the reference data management unit 114, and the confirmation result is held in the monitoring result storage unit 121 (step S310-1). Similarly, in the transfer unit 15 of the cache servers 1 of both systems, the monitoring processing unit 152 confirms the occurrence of each process and holds the result in the monitoring result storage unit 161 (step S305-2). A dummy SQL is issued to confirm the operation of the transfer data management unit 154, and the confirmation result is held in the monitoring result storage unit 161 (step S310-2).

  The service check unit 22 of the load balancer 2 transmits a health check signal to each of the reference unit 11 and the transfer unit 15 of the cache servers 1 of both systems (steps S315-1 and S315-2). Upon receiving the health check signal, the monitoring result notification processing unit 113 of both cache servers 1 acquires the monitoring result of the system state held in the monitoring result storage unit 121 from the monitoring processing unit 112 (step S320-1). ) If the acquired monitoring result is “normal”, a response is returned (step S325-1). Similarly, upon receiving the health check signal, the monitoring result notification processing unit 153 of the cache servers 1 of both systems acquires the monitoring result of the system state held in the monitoring result storage unit 161 from the monitoring processing unit 152 (Step S1). S320-2) When the acquired monitoring result is “normal”, a response is returned (step S325-2). The service check unit 22 of the load balancer 2 holds in the storage unit 221 the presence / absence of responses from the reference unit 11 and the transfer unit 15 of the cache servers 1 of both systems.

[2.3 Internal State Monitoring Processing in Cache Server 1]
FIG. 6 is executed by the reference unit 11 and the transfer unit 15 of the cache servers 1 in both systems in steps S105-1 to S105-4, S135-1 to S135-4, and S140-1 to S140-4 in FIG. It is a figure which shows a self-system state monitoring process flow.

  First, an internal system state monitoring process flow in the reference unit 11 will be described. In the figure, the timer 111 of the cache server 1 activates the monitoring processing unit 112 every predetermined time (step S405). Thereby, the monitoring processing unit 112 confirms the occurrence of each process of the reference unit 11 in its own system (step S410). If each process has occurred (step S415: normal), it is determined that the process is normal, and then a dummy SQL is issued to the reference data management unit 114 in order to confirm the alive state of the DB (step S420). The reference data access unit 122 of the reference data management unit 114 returns to the monitoring processing unit 112 a result of performing a data operation on the data stored in the reference data storage unit 123 based on the dummy SQL. When the dummy SQL response is normally received from the reference data management unit 114 (step S425: normal), the monitoring processing unit 112 writes the monitoring result “normal” in the monitoring result storage unit 121 of the own system (step S430). .

  On the other hand, when the process has not occurred (Step S415: Abnormal), or when the dummy SQL response is not normally received from the reference data management unit 114 (Step S425: Abnormal), the monitoring processing unit 112 performs monitoring. The result “abnormal” is written in the monitoring result storage unit 121 provided in the own system (step S435).

  In the case of the own system state monitoring process flow in the transfer unit 15, the reference unit 11 is transferred to the transfer unit 15, the timer 111 is set to the timer 151, the monitor processing unit 112 is set to the monitor processing unit 152, and the reference data management unit 114 is transferred to the transfer data management. In this case, the monitoring result storage unit 121 is replaced with the monitoring result storage unit 161, the reference data access unit 122 is replaced with the transfer data access unit 162, the reference data storage unit 123 is replaced with the transfer data storage unit 163, and the same processing as described above is performed. Do.

[2.4 In-system status monitoring result return processing in cache server 1]
7 shows the in-system state monitoring result return process executed by the reference unit 11 and the transfer unit 15 of the cache servers 1 in both systems in steps S120-1 to 120-4 and 155-1 to 155-4 in FIG. It is a figure which shows a flow.

  First, the internal system state monitoring result return processing flow in the reference unit 11 will be described. In the figure, the monitoring result notification processing unit 113 of the cache server 1 receives a health check signal from the load balancer 2 (step S505). The monitoring result notification processing unit 113 acquires the monitoring result of the in-system state stored in the monitoring result storage unit 121 from the monitoring processing unit 112 (step S510). If the acquired in-system status monitoring result is “normal” (step S515: normal), the monitoring result notification processing unit 113 returns a response indicating normality to the load balancer 2 that is the source of the health check signal (step S515). S520) When the system state monitoring result is “abnormal” (step S515: abnormal), the received health check signal is discarded and no response is returned (step S525).

  In the case of the own system state monitoring result return processing flow in the transfer unit 15, the monitoring processing unit 112 is the monitoring processing unit 152, the monitoring result notification processing unit 113 is the monitoring result notification processing unit 153, and the monitoring result storage unit 121 is the monitoring result. The memory 161 is replaced with the same processing as described above.

[2.5 Health check in load balancer 2]
FIG. 8 is a diagram showing a health check process flow of the load balancer 2 executed in steps S110 and S130, S145 and S155 of FIG.
In the figure, the timer 21 of the load balancer 2 activates the service check unit 22 every predetermined time (step S605). First, a case where a health check for the 0-system reference unit 11 is activated will be described. The service check unit 22 transmits a health check signal with the physical IP address C0 of the cache server 1-0 and the port number PS0 of the reference unit 11-0 as destinations (step S610).

When the service check unit 22 receives a response transmitted by the monitoring result notification processing unit 113 of the cache server 1 in step S520 of FIG. 7 in response to the health check signal transmitted in step S610 (step S615: response is present), “Respond” is written in the storage unit 221 in association with the identification information of the reference unit 11-0 (step S620).
On the other hand, if no response is received within a predetermined time with respect to the health check signal transmitted in step S610 (step S615: no response), “no response” is stored in association with the identification information of the reference unit 11-0. The data is written in the part 221 (step S625).

  When the health check for the 0-system transfer unit 15-0 is activated in step S605, in step S610, the service check unit 22 determines the physical IP address C0 of the cache server 1-0 and the port of the transfer unit 15-0. A health check signal is transmitted with the number PT0 as the destination. In step S620, “with response” is associated with the identification information of the transfer unit 15-0, and “no response” is associated with the identification information of the transfer unit 15-0 in step S625. Write to.

  Similarly, when the health check for the 1-system reference unit 11-1 is activated in step S605, in step S610, the service check unit 22 determines that the physical IP address C1 of the cache server 1-1 and the reference unit 11- A health check signal is transmitted with the port number PS1 of 1 as the destination. In step S620, “with response” is associated with the identification information of the reference unit 11-1, and in step S625, “no response” is associated with the identification information of the reference unit 11-1. Write to.

  In step S605, when the health check for the 1-system transfer unit 15-1 is activated, in step S610, the service check unit 22 determines that the physical IP address C1 of the cache server 1-1, the transfer unit 15-1. The health check signal is transmitted with the port number PT1 as the destination. In step S620, “with response” is associated with the identification information of the transfer unit 15-1, and “no response” is associated with the identification information of the transfer unit 15-1 in step S625. Write to.

[3. Sorting process]
The load balancers 2a and 2b hold the health check result for determining whether the reference unit 11 and the transfer unit 15 of the cache servers 1-0 and 1-1 are normal or abnormal by the health check described above. To do. That is, it is possible to determine that the operation state is normal when a response is returned and abnormal when the response is not returned. Then, every time a data reference request is received from the application apparatus 4, the load balancer 2a uses the health check result to determine the active cache server 1, and performs a data reference request distribution process. The load balancer 2b performs the same distribution process every time it receives a system state identification signal from the cache server 1. Hereinafter, the operation of the distribution process will be described.

[3.1 Sorting processing sequence]
FIG. 9 is a diagram illustrating a distribution processing sequence when a data reference request is received from the application apparatus 4.
In the figure, the load balancer 2a receives a data reference request addressed to the virtual IP address A of the load balancer 2a from the application device 4 (step S705). The load balancer 2 determines the active cache server 1 with reference to the health check processing result (step S710). Here, it is assumed that the cache server 1-0 is determined to be an active system. The load balancer 2 transmits the destination of the data reference request as the physical IP address C0 of the active cache server 1-0 and the port number PS0 of the reference unit 11-0 (step S715). When the 0-system reference unit 11-0 returns a response to the received data reference request (step S720), the load balancer 2a transmits a response to the application device 4 that is the transmission source of the data reference request (step S725). .

  Each time the load balancer 2a receives a data reference request addressed to the virtual IP address A of the load balancer 2a from the application device 4, the load balancer 2a refers to the health check processing result to determine the active cache server 1. In other words, as shown in the figure, when the load balancer 2a receives the data reference request from the application device 4 again (step S730), the load balancer 2a refers to the health check processing result to determine the active cache server 1 (step S730). S735). The subsequent steps S745 to S750 are the same as the steps S715 to S720.

  In the distribution process, when the load balancer 2a determines that the active system is the cache server 1-1, the destination of the data reference request received from the application apparatus 4 is the destination of the cache server 1-1 that is the active system. Transmission is performed as the physical IP address C1 and the port number PS1 of the reference unit 11-1. When the 1-system reference unit 11-1 returns a response to the data reference request, the load balancer 2a transmits the response to the application device 4 that is the transmission source of the data reference request.

FIG. 10 is a diagram illustrating a distribution processing sequence when a system state identification signal is received from the cache server 1.
In the figure, the load balancer 2b receives a system state identification signal destined for the virtual IP address B of the load balancer 2b from the 0-system transfer unit 15-0 (step S805). The load balancer 2b refers to the health check process result to determine the active cache server 1 (step S810). Here, it is assumed that the cache server 1-0 is determined to be an active system. The load balancer 2b transmits the destination of the system state identification signal as the physical IP address C0 of the cache server 1-0 that is the active system and the port number PT0 of the transfer unit 15-0 (step S815). The 0-system transfer unit 15-0 recognizes that it is an active system in order to receive the system state identification signal transmitted by the own system.

  Similarly, the load balancer 2b receives a system state identification signal destined for the virtual IP address B of the load balancer 2b from the 1-system transfer unit 15-1 (step S820). It is assumed that the load balancer 2b refers to the health check processing result and determines that the cache server 1-0 is the active system (step S825). The load balancer 2b transmits the destination of the system state identification signal as the physical IP address C0 and the port number PT0 of the cache server 1-0 that is the active system (step S830). The 1-system transfer unit 15-1 recognizes that it is a standby system because it cannot receive the system state identification signal transmitted by its own system.

  In the distribution process, when the load balancer 2b determines that the active system is the cache server 1-1, the destination of the system state identification signal received from the cache servers 1-0 and 1-1 is set to the active system. It is transmitted as the physical IP address C1 of a certain cache server 1-1 and the port number PT1 of the transfer unit 15-1. As a result, the 1-system transfer unit 15-1 receives the system state identification signal transmitted by the own system, and thus recognizes that it is the active system, and the 0-system transfer unit 15-0 transmits the own system. Since the received system state identification signal cannot be received, it is recognized that the system is a standby system.

[3.2 Distributing process in load balancer 2]
FIG. 11 is a diagram showing a distribution processing flow in the load balancer 2a.
In the figure, the load balancer 2a activates the operational system determination processing unit 23a when receiving a data reference request addressed to the virtual IP address A of the load balancer 2a from the application device 4 (step S905). The operational system determination processing unit 23a receives from the storage unit 221a of the service check unit 22a from the 0-system reference unit 11-0 and the transfer unit 15-0, and the 1-system reference unit 11-1 and the transfer unit 15-1. A result of no health check signal response is read (step S910), and an operation flag is read and held (step S915). Note that the operation flag indicates that the response state of the 0-system and 1-system reference unit 11 is the same and the response state of the 0-system and 1-system transfer unit 15 is the same, for example, from the response state. This flag is referred to when the active system cannot be determined. When the 0 system is the distribution destination, “0”, when the 1 system is the distribution destination, “1”, both the 0 system and the 1 system When distribution is impossible, the value “3” is taken. Normally, “0” or “1” is set as an initial value when the system is started.

Subsequently, the active system determination processing unit 23a determines the distribution state according to the distribution state table shown in FIG. 12 (step S920).
In other words, the active system determination processing unit 23a receives a response from both the 0-system and the 1-system reference unit 11, and the response status of the 0-system and the 1-system transfer unit 15 is the same. The minute state is “2”. Also, when there is a response from the 0-system reference unit 11-0 and no response from the 1-system reference unit 11-1, or the 0-system and 1-system reference unit 11 and the 0-system transfer unit 15 When there is a response from −0 and no response is received from the 1-system transfer unit 15-1, the distribution state is set to “0”. Also, when there is no response from the 0-system reference unit 11-0 and there is a response from the 1-system reference unit 11-1, or the 0-system and 1-system reference units 11 and the 1-system transfer unit 15 If there is a response from -1 and no response from the 0-system transfer unit 15-0, the distribution state is set to "1". Further, when there is no response from both the 0-system and 1-system reference units 11, it is determined that the distribution state is “3”.

  When it is determined that the distribution state is “0” (step S920: distribution state 0), the active system determination processing unit 23a sets the destination of the data reference request received in step S905 as the physical IP address C0 of the cache server 1-0. The port number PS0 of the reference unit 11-0 is transmitted (step S925), and the value of the distribution state is set in the operation flag stored in the storage unit 221a (step S930). Thereby, “0” is set in the operation flag.

  When it is determined that the distribution state is “1” (step S920: distribution state 1), the active system determination processing unit 23a sets the destination of the data reference request received in step S905 as the physical IP address C1 of the cache server 1-1. The port number PS1 of the reference unit 11-1 is transmitted (step S935), and the value of the distribution state is set in the operation flag stored in the storage unit 221a (step S930). Thereby, “1” is set in the operation flag.

When it is determined that the distribution state is “2” (step S920: distribution state 2), since both the 0 system and the 1 system are operating normally, whichever of the cache servers 1 is the active system and the standby system from the response state It cannot be judged whether or not. Therefore, the active system determination processing unit 23a further acquires the operation flag held in step S915 (step S940), and determines whether the value is “0” or “1” (step S945).
When the operation flag is “0” or “1” (step S945: true), the active system determination processing unit 23a transmits the data reference request received in step S905 to the cache server 1 of the system indicated by the operation flag. (Step S950). That is, when the operation flag is “0”, the destination address of the data reference request is converted into the physical IP address C0 of the cache server 1-0 and the port number PS0 of the reference unit 11-0 and transmitted, and the operation flag is set. In the case of “1”, the destination address of the data reference request is converted into the physical IP address C1 of the cache server 1-1 and the port number PS1 of the reference unit 11-1 and transmitted. As a result, while both the 0 system and the 1 system are operating normally, the same system becomes the distribution destination.

  On the other hand, when the operation flag is neither “0” nor “1” (step S945: false), the operation system determination processing unit 23a sets the system 0 as the distribution destination. The active system determination processing unit 23a converts the destination address of the data reference request received in step S905 into the physical IP address C0 of the cache server 1-0 and the port number PS0 of the reference unit 11-0 and transmits the converted address (step S955). Then, “0” is set in the operation flag (step S960). Therefore, if an abnormality occurs in both the 0 system and the 1 system and then returns to normal at the same time, the 0 system is the distribution destination.

  When it is determined that the distribution state is “3” (step S920: distribution state 3), since both the 0-system and the 1-system are abnormal, the operational system determination processing unit 23a receives the data reference request received in step S905. Is returned to the application device 4 (step S965), and “3” is set in the operation flag (step S970).

FIG. 13 is a diagram illustrating a processing flow of distribution processing in the load balancer 2b.
In the figure, the load balancer 2b activates the active system determination processing unit 23b in response to receiving from the cache server 1 a system state identification signal destined for the virtual IP address B of the load balancer 2b (step S1005).
The processing after step S1010 is the same as the processing after step S910 shown in FIG. 11, and the distribution status table has the same setting contents. However, the operational system determination processing unit 23a, the storage unit 221a, the port number PS0 of the reference unit 11-0, and the port number PS1 of the reference unit 11-1 are respectively the operational system determination processing unit 23b, the storage unit 221b, and the transfer unit 15-. The port number PT0 of 0 and the port number PT1 of the transfer unit 15-1 are read, and the data reference request received from the application apparatus 4 in step S905 is read as the system state identification signal received from the cache server 1 in step S1005. Therefore, in step S1065, an error response is returned to the transfer unit 15 of the cache server 1 that is the transmission source of the system state identification signal.

[4. Data synchronization processing]
In the cache system, it is necessary to synchronize data between the cache servers 1-0 and 1-1 while synchronizing with actual data stored in the original data management device 3. In the following, the reason for adopting the data synchronization method of this embodiment will be described, and then the operation of the synchronization processing in the cache system according to this embodiment will be described.

[4.1 Data synchronization method review process]
FIG. 14 is a diagram showing an original data synchronization method for which application to the cache system according to the present embodiment has been studied.
FIG. 14A shows a method of reflecting actual data from the original data management apparatus 3 to the cache servers 1 of both systems. In this method, it is not necessary to synchronize data between the cache servers 1 of both systems. However, the load on the original data management device 3 is increased, and a function for synchronization must be added. If a mistake occurs during transfer of actual data from the original data management device 3, the cache server 1 There is a problem that the difference from the actual data cannot be reflected. In addition, the consistency of the data stored in the cache servers 1 of both systems cannot be ensured, and there is a risk that the data may not be synchronized at the time of system switching, and there is a problem that the redundancy itself cannot be realized.

  FIG. 14B shows a method of reflecting the actual data from the original data management apparatus 3 to only one cache server 1 and then synchronizing the data between the systems of the cache server 1. In this method, a function for writing original data must be added to the original data management apparatus 3, and if a mistake occurs during transfer of actual data from the original data management apparatus 3, the cache server 1 There is a problem that the difference from the actual data cannot be reflected. Furthermore, in the original data management apparatus 3, the operation when the system switching has occurred in the cache server 1 must be taken into account, and there is a possibility that data may not be synchronized at the time of system switching. There is also a problem that cannot be realized.

FIG. 14C shows a method in which one cache server 1 periodically reads original data from the original data management apparatus 3 to synchronize data between the systems. In the case of this method, although there are points to be considered, the conditions related to the basics of redundancy such as minimization of data lost due to system switching are cleared.
Therefore, in the present embodiment, based on the method of FIG. 14C, a process for determining which of the cache server 1 operating without recognizing the active system and the standby system performs synchronization with the original data. In addition to this, an inter-system synchronization method that can manage how much data synchronization has been performed between systems has been added.

  As to which of the two independent cache servers 1 synchronizes with the original data, as described in FIG. 10 and the like, each cache server 1 sends a dummy request to the virtual IP address of the load balancer 2b, that is, A system state identification signal is transmitted, and when a response is returned from the request, it is recognized that the synchronization is performed, and data synchronization processing is started.

FIG. 15 is a diagram showing a synchronization method between systems that are examined for application to the cache system according to the present embodiment.
FIG. 15A is a diagram illustrating a method of using the data synchronization function of an existing DBMS. In this embodiment, the cache servers 1 of both systems are not aware of the status of their own system or the status of the other system, and the systems are switched by the load balancer 2 that is an external network device. That is, switching by the load balancer 2 is performed regardless of the state of synchronization between systems possessed by the link function of the existing DBMS. Therefore, the link function itself of the existing DBMS cannot recognize that the system has been switched, and cannot perform data synchronization correctly. Further, in the existing DBSM, the function of recovering data loss at the time of failure is not applied, and it may be necessary to restart the DB instance, which is not appropriate.

  FIG. 15B shows a method in which the cache server 1 that is not synchronized with the original data sucks up the updated data from the cache server 1 that is synchronized with the original data. In this method, the synchronization processing with the original data and the synchronization processing between the systems are distributed to both systems. There is a problem that data remaining on the service route side cannot be erased. Further, the cache server 1 of both systems must be aware of the status of the system and is not suitable for the present embodiment.

FIG. 15C shows a method in which the cache server 1 of the system synchronized with the original data writes the updated data to the cache server 1 on the side not synchronized with the original data. In the case of this method, in the cache server 1 of the system that performs synchronization with the original data, the synchronization processing between the systems is also performed, so that the data synchronization processing is concentrated in one system, and data inconsistency occurs during system switching. Can be prevented.
Therefore, the method shown in FIG. 15C is adopted as a synchronization method between systems, and a management method of how far data synchronization has been performed is added as follows.

  In other words, in the method shown in FIG. 15C, the system that newly synchronizes with the original data, such as when a system switch occurs during data transfer between the systems, synchronizes with the original data from any point in time. I can't figure out what to do. Therefore, the extent to which data synchronization has been performed is managed in the original data management device 3 using the time stamp given when the original data was updated. Specifically, when the cache data 1 of the system synchronized with the original data reads the original data updated from the original data management device 3, it reads together with the update date and time of each original data, and the latest update date and time among them Is stored as the synchronization date. When the updated original data is written to the cache server 1 on the side that is not synchronized with the original data, the synchronization date is also notified at the same time. As a result, at the next synchronization timing, the cache server 1 of the system that synchronizes with the original data reads the original data that is updated after the synchronization date and time held by the own system from the original data management device 3. Just put it out.

[4.2 Data synchronization sequence]
FIG. 16 is a diagram showing a data synchronization sequence of the cache system.
This figure shows an example in which the load balancer 2b determines that the cache server 1-0 is the active system and the cache server 1-1 is the standby system.
The transfer units 15 of the cache servers 1 of both systems periodically start the system state identification process and transmit a system state identification signal to the load balancer 2b. In FIG. 10, when the 0-system transfer unit 15-0 transmits a system state identification signal to the load balancer 2b, the processes in steps S805 to S815 in FIG. 10 are performed (step S1105).

  Since the load balancer 2b determines that the cache server 1-0 is the active system, the 0-system transfer unit 15-0 receives the system status identification signal transmitted by the own system. Thereby, the transfer unit 15-0 determines that the own system is the active system, determines to continue the process (step S1110), and starts the synchronization process with the original data (step S1115). The transfer unit 15-0 transmits acquisition of original data to the original data management device 3 (step S1120), and the data management unit 31 of the original data management device 3 reads the original data from the original data storage unit 33 (step S1125). ) And returns to the transfer unit 15-0 (step S1130). The transfer unit 15-0 updates the data stored in the transfer data storage unit 163-0 with the received original data (step S1135), and then performs intersystem synchronization processing (step S1140). That is, the 0-system transfer unit 15-0 transmits the original data to be updated to the 1-system transfer unit 15-1 (step S1145), and the transfer unit 15-1 uses the received original data to transfer the transfer data storage unit 163. The data stored in -1 is updated (step S1150). When the update is normally completed, the 1-system transfer unit 15-1 returns a normal response to the 0-system transfer unit 15-0 (step S1155).

  On the other hand, the 1-system transfer unit 15-1 also transmits a system state identification signal to the load balancer 2b. Thereby, similarly, the processing of steps S820 to S830 in FIG. 10 is performed (step S1160). As described above, since the load balancer 2b determines that the cache server 1-0 is the active system, the system status identification signal is transmitted to the 0-system transfer unit 15-0. Since the transfer unit 15-1 cannot receive the system state identification signal transmitted by the own system, the transfer unit 15-1 determines that the own system is a standby system and ends the process (step S1165).

  The transfer unit 15 of both cache servers 1 periodically starts the processes of steps S1105 and S1160, and repeats the processes of steps S1105 to S1165 described above while the active system and the standby system are not switched. When the system state is switched and the cache server 1-1 becomes the active system, the operations of the 0-system transfer unit 15-0 and the 1-system transfer unit 15-1 in the above processing are switched.

[4.3 Data used for data synchronization processing]
FIG. 17 is a diagram illustrating an example of data of each device used for data synchronization processing. In this example, the load balancer 2b determines that the cache server 1-0 is the active system and the cache server 1-1 is the standby system.

  The original data management device 3 stores, in the original data storage unit 33, “information 1”, “information 2”, “information 3”, “information 4”, which are original data of various types of information used by the application device 4. Is remembered. In the update history storage unit 32, the updated original data “information 1”, “information 2”, “information 3”, “information 4”, their update date and time “00:00:01”, An update history TBL in which “00:00:02”, “00:00:03”, and “00:00:04” are associated is stored. The data management unit 31 of the original data management device 3 updates the original data updated to the update history TBL in the update history storage unit 32 whenever the original data stored in the original data storage unit 33 is updated. And its update date are written in association with each other.

  The transfer data storage unit 163-0 included in the 0-system transfer unit 15-0 considered to be the active system is a copy (cache data) of the original data read from the original data management device 3. Information 1 ”and“ Information 2 ”are stored. In addition, in the update information TBL of the synchronous information storage unit 156-0, the original data “information 2” that has been written in the transfer data storage unit 163-0 but has not yet been synchronously updated with the cache server 1-1, and its data The update date and time “00:00:02” is set in association with each other. Furthermore, in the last update information TBL stored in the synchronization information storage unit 156-0, the last look “00: which is the latest update date / time among the update date / time of the original data that has been written to the transfer data storage unit 163-0. 00:02 "is set.

  The transfer data storage unit 163-1 included in the transfer unit 15-1 of the first system that is regarded as the standby system is a copy (cache data) of the original data read from the original data management device 3. Information 1 "is stored. Further, in the last update information TBL stored in the synchronization information storage unit 156-1, the last look “00: which is the latest update date / time among the update date / time of the original data that has been written to the transfer data storage unit 163-1. 00:01 "is set. In the case of the standby system, since the original data is not read directly from the original data management device 3, the update information TBL in the synchronization information storage unit 156-1 is normally empty.

[4.4 Data update flow]
FIG. 18 is a diagram showing the flow of data update in the original data synchronization process, and shows an example in which the update is performed from the state of FIG.
In the figure, the 0-system transfer unit 15-0, which has recognized that the own system is the active system, acquires the last look set in the last update information TBL, and acquires the acquired last look and original data acquisition request. Are transmitted to the original data management apparatus 3 (step S1205). The data management unit 31 of the original data management device 3 searches the update history TBL in the update history storage unit 32 to identify original data whose update date and time is after the last look, and caches the specified original data and its update date and time. Return to server 1-0. In the figure, the original data management apparatus 3 receives the last look “00:00:02” from the cache server 1-0, and the original data “information 3” and its update date “00:00:03”. Data “information 4” and its update date “00:00:04” are returned.

  When the original data “information 3” and “information 4” received from the original data management apparatus 3 are written in the transfer data storage unit 163-0 (step S1210), the 0-system transfer unit 15-0 updates these originals. Of the data update dates and times, the last update date TBL is overwritten with the latest update date and time “00:00:04” (step S1215).

  In the above description, the 0-system transfer unit 15-0 transmits the last look set in the final update information TBL to the original data management apparatus 3 together with the original data acquisition request. You may send the date and time which goes back by time. The data management unit 31 of the original data management device 3 searches for the update history TBL, specifies original data whose update date is after the received date, and returns the specified original data and its update date.

FIG. 19 is a diagram showing the flow of data update in the intersystem data synchronization processing, and shows an example in the case of executing following the processing shown in FIG.
Before or after the process of step S1215 in FIG. 18, the 0-system transfer unit 15-0 adds the original data “information 3” received in step S1210 and its update date and time “00:00:03” to the update information TBL. Original data “information 4” and its update date “00:00:04” are additionally written (step S1305).

  When the last look “00:00:01” is read from the last update information TBL held by the 1-system transfer unit 15-1, the 0-system transfer unit 15-0 is set to the update information TBL held by the own system. It is compared with the update date and time (step S1310). The 0-system transfer unit 15-0 specifies the original data in the update information TBL whose update date and time is after the last look read from the 1-system transfer unit 15-1, and the 1-system transfer unit is determined based on the specified original data. Information stored in the transfer data storage unit 163-1 included in 15-1 is updated. Thereby, “information 2”, “information 3”, and “information 4” are written in the transfer data storage unit 163-1 of the cache server 1-1 (step S1315).

  The 0-system transfer unit 15-0 uses the latest update date and time “00:00:04” among the update dates and times of the original data written in the 1-system transfer data storage unit 163-1 in step S1315 as the last look. The final transfer information TBL stored in the system transfer unit 15-1 is written (step S1320). The 0-system transfer unit 15-0 deletes “information 2”, “information 3”, and “information 4”, which are the original data written in the 1-system transfer unit 15-1, from the update information TBL.

[4.5 Update processing in cache server 1]
FIG. 20 shows a system state identification process flow executed by the transfer unit 15 of the cache servers 1 of both systems in steps S1105 and S1160 of FIG.
In the same figure, the timer 151 of the cache server 1 starts the system state identification process every predetermined time, whereby the system state identification unit 155 sends a system state identification signal to the virtual IP address B of the load balancer 2b. Transmit (step S1405). The load balancer 2b performs the distribution process shown in FIG. 13 and transmits the received system state identification signal to the cache server 1 that is determined to be the active system.

  After transmitting the system state identification signal in step S1405, if the system state identification signal transmitted by the own system cannot be received (step S1410: not reached), the system state identifying unit 155 determines that the own system is the current standby system. It is determined that there is, and the system state is held (step S1415).

  After transmitting the system state identification signal in step S1405, if this system state identification signal transmitted by the own system is received (step S1410: reached), the system state identification unit 155 determines that the own system is the current operating system. The system state is determined and held (step S1420). Whether or not the system status identification signal transmitted by the own system has been received can be determined based on whether or not the source IP address set in the received system status identification signal is the physical IP address of the own system.

FIG. 21 shows a synchronization processing flow with the original data executed by the transfer unit 15 of the two cache servers 1 in steps S1110 to S1135 and S1165 of FIG.
When the system state identification process shown in FIG. 20 ends, the transfer unit 15 of both cache servers 1 starts a synchronization process with the original data. Thereby, the information transfer processing unit 157 determines the system state of the own system held by the system state identification unit 155 (step S1505). When the system status of the own system is the standby system (step S1505: standby system), the information transfer processing unit 157 ends the process.

  When the own system state is the active system (step S1505: active system), the information transfer processing unit 157 reads the last look from the last update information TBL stored in the own synchronization information storage unit 156 (step S1505: active system). S1510). Subsequently, the information transfer processing unit 157 transmits the acquired last look and the original data acquisition request to the original data management device 3, so that the original data having the update date and time after the last look is transmitted from the original data management device 3 and its original data. Information on the update date is acquired (step S1515). When the information transfer processing unit 157 instructs the transfer data management unit 154 to write the original data acquired in step S1515, the transfer data access unit 162 writes the acquired original data in the transfer data storage unit 163 (step S1515). S1520). The information transfer processing unit 157 associates the original data for which writing has been completed with the update date and time of the original data received from the original data management device 3, and stores the update information stored in the own synchronization information storage unit 156. Register in the TBL (step S1525). The information transfer processing unit 157 writes the latest update date / time corresponding to the information registered in step S1525 as the last look in the last update information TBL stored in the own synchronization information storage unit 156 (step S1525). S1530).

FIG. 22 shows an inter-system data synchronization processing flow executed by the transfer unit 15 of the active cache server 1 in step S1140 of FIG.
The transfer unit 15 of the cache server 1 that has determined that its own system state is the active system in step S1505 in FIG. 21, when the synchronization process with the original data shown in FIG. to start. As a result, the inter-system transfer processing unit 158 of the active cache server 1 acquires the original data and its update date and time from the update information TBL stored in the local synchronization information storage unit 156 (step S1605). Further, the inter-system transfer processing unit 158 reads the last look from the final update information TBL stored in the synchronization information storage unit 156 of the cache server 1 that is the other system, that is, the current standby system (step S1610).

  If the update date and time of the original data reflected in the transfer data storage unit 163 acquired in step S1605 is newer than the standby last look acquired in step S1610 (step S1615: the active system is newer), The inter-system transfer processing unit 158 instructs the transfer data management unit 154 of the standby cache server 1 to write the original data acquired in step S1605 (step S1620). As a result, the transfer data access unit 162 of the standby cache server 1 writes the original data instructed to the own transfer data storage unit 163.

  Subsequently, the inter-system transfer processing unit 158 of the active cache server 1 sets the latest update date and time among the update dates and times in the original data management device 3 corresponding to the original data written to the standby cache server 1 in step S1620. The last look-up information TBL stored in the synchronization information storage unit 156 of the standby cache server 1 is written as the last look (step S1625). In step S1620, the inter-system transfer processing unit 158 of the active cache server 1 uses the original data written in the transfer data storage unit 163 of the standby cache server 1 and its update date and time as its own transfer data storage unit 163. Is deleted from the update information TBL stored in (step S1630).

  If the last look of the standby system acquired in step S1610 is newer than the update date and time of the original data already reflected in the local transfer data storage unit 163 acquired in step S1605 (step S1615: the standby system is newer). ), The process is terminated.

Both cache servers 1 periodically perform in-system data synchronization processing as follows.
In the reference unit 11 of the cache server 1, the in-system synchronization processing unit 116 is periodically activated by the timer 111. The activated intra-system synchronization processing unit 116 requests intra-system data synchronization to the transfer unit 15. When the intra-system synchronization processing unit 159 of the transfer unit 15 receives the intra-system data synchronization from the intra-system synchronization processing unit 116 of the reference unit 11, the transfer data access unit 162 returns the original data read from the transfer data storage unit 163. . The intra-system synchronization processing unit 116 of the reference unit 11 instructs the reference data access unit 122 to write the original data received from the intra-system synchronization processing unit 159 of the transfer unit 15, and the reference data access unit 122 The original data received by the unit 116 is written into the reference data storage unit 123.

[5. General-purpose interface]
The cache system according to the present embodiment receives data reference requests from a plurality of application apparatuses 4. However, if a unique interface (hereinafter referred to as IF) is defined for each higher-level service application executed in the application apparatus 4, it is necessary to implement the IF every time a higher-level service application is added. Therefore, a function for performing data access by a general-purpose IF definition is implemented, and development accompanying the addition of a higher-level service application is unnecessary.

  In the general-purpose IF definition, one IF definition corresponds to a plurality of patterns of data reference requests and responses used by a plurality of higher-level service applications, that is, an IN parameter and an OUT parameter. Therefore, even when a new data reference request and response pattern is generated, it is only necessary to change the setting file, and IF development is unnecessary.

  FIG. 23 is a diagram showing an operation outline of the general-purpose IF. In the figure, the application apparatus 4 transmits a data reference request. The data reference request is an IN message described using a general-purpose interface, and an identifier for specifying a conversion rule and an IN item used as a search key are set. The HTTP (Hypertext Transfer Protocol) -SOAP (Simple Object Access Protocol) is set. ). The load balancer 2a distributes the data reference request to the active cache server 1, and the information reference processing unit 115 of the active cache server 1 determines the conversion rule specified by the identifier set in the data reference request as the information Obtained from the rule storage unit 124 included in the reference processing unit 115 (step S1705).

The rule storage unit 124 stores conversion rules for each identifier, and the conversion rules include normality check data, DB access statement conversion rules, and OUT message generation rules.
The normality check data indicates a condition for checking the normality of the IN message. The normality check data is described in, for example, an XML (extensible markup language) schema, and is an XSD file that describes conditions for checking the normality of the data type, length, setting value, etc. of IN item parameters. is there.
The DB access statement conversion rule indicates a conversion rule for generating a DB access statement using the IN item of the general-purpose interface as a parameter. The DB access statement conversion rule is, for example, an XSLT file for deriving SQL that describes a conversion rule for generating an SQL statement using an IN item as a search key, which is described in XML Stylesheet Language Transformations (XSLT).
The OUT message generation rule indicates a conversion rule for generating an OUT message of a general-purpose interface in which data read from the reference data management unit 114 is set as an OUT item. The OUT message generation rule is, for example, an OUT message generation XSLT file that describes a conversion rule that is described in XSLT and that formats the read data into an XML format as an OUT item.

  The information reference processing unit 115 uses the read conversion rule to check the normality of the IN item set in the data reference request, and then generates a DB access statement that searches the original data using the IN item as a key. The data is output to the transfer data management unit 154 (step S1710). Using the read general-purpose IF definition, the information reference processing unit 115 generates an XML OUT message in which the original data read by the DB access statement from the reference data management unit 123 is set as an OUT item, and sends it to the application apparatus 4. Return it (step S1715).

  As described above, when adding a data access request and a response, it is only necessary to add a conversion rule, and IF development associated with the addition to the higher-level service application becomes unnecessary.

  FIG. 24 is a diagram illustrating a general-purpose interface processing flow in the information reference processing unit 115. In the figure, the information reference processing unit 115 extracts an identifier from an IN message that is a data reference request transmitted from the application apparatus 4. The information reference processing unit 115 identifies the conversion rule stored in the rule storage unit 124 by the read identifier, that is, the message normality confirmation XSD file, the SQL derivation XSLT file, and the OUT message generation XSLT file to be applied. (Step S1805). The information reference processing unit 115 applies the message normality confirmation XSD file specified in step S1805 and confirms the normality of the IN item, which is a parameter set in the IN message (step S1810). The information reference processing unit 115 uses the SQL derivation XSLT file specified in step S1805 to generate SQL for reading original data that is a necessary element as an OUT item (step S1815).

  The information reference processing unit 115 outputs the SQL sentence generated in step S1815 to the reference data management unit 114. The reference data access unit 122 of the reference data management unit 114 reads the original data from the reference data storage unit 123 according to the received SQL sentence, and outputs it to the information reference processing unit 115 (step S1820). The information reference processing unit 115 applies the OUT message generation XSLT file specified in step S1805 to the original data acquired in step S1820, and creates an OUT message (step S1825). The information reference processing unit 115 returns the OUT message generated in step S1825 to the application device 4.

[6. effect]
According to the cache system of the present embodiment described above, even if the original data management device 3 breaks down or performs planned stoppage, the original data that ensures the latest and unity in the application device 4 Can be provided continuously.

  In the cache system of the present embodiment, both cache servers 1 operate independently and asynchronously without being conscious of whether the own system is an active system or a standby system, and are connected to the load balancer 2 that is an external network device. I am entrusting the system switchover. Therefore, when the cache server 1 receives the data access request, the cache server 1 may provide access to the cache data stored in the reference data storage unit 123 included in the own system. Therefore, it is possible to solve the problem of a single point of failure when a shared disk is used as in the prior art. Furthermore, when the load balancer 2 determines the necessity of system switching, the system can be switched quickly. Can be done.

Further, the cache server 1 can make an inquiry as to whether or not the own system is a synchronized system by using the system switching function of the load balancer 2, and the cache server 1 communicates with the original data management apparatus 3 according to the inquiry result. Data synchronization and intersystem data synchronization with the other cache server 1 are started. Therefore, it is possible to realize synchronization while maintaining data consistency even between independent systems.
In addition, even if a system switchover is performed or an error occurs during data transfer between systems by using the update date and time of the original data assigned in the original data management apparatus, the original data management It is possible to prevent data synchronization with the apparatus and synchronization between systems.

  The cache system is a system located in an intermediate layer between the application device 4 that executes the higher-level service application and the original data management device 3 that holds the original data. For this reason, the reference unit 11 and the transfer unit 15 do not necessarily match the traffic increase factors and timing. For example, the traffic of the reference unit 11 increases with an increase in the number of users who use the upper service application and an increase in the types of upper service applications. On the other hand, the traffic of the transfer unit 15 increases when a data item referred to by the upper service application is added. In the present embodiment, since the reference unit 11 and the transfer unit 15 are separated in one housing (server), a server that executes the reference unit 11 and a server that executes the transfer unit 15 in the future. It is easy to shift to a configuration in which the number of units is set in accordance with traffic. In general, since the rate of increase in traffic from the higher-level service application is considered to be higher, for example, the reference unit 11 is executed in order to enable load distribution by expanding the performance of the reference unit 11 and scaling out. It is easy to shift the server to be operated to an all-operation system having n units and the server to execute the transfer unit 15 to an operation standby system having 1 + 1 units.

[7. Others]
In the above embodiment, the cache interface apparatus is multiplexed by two cache servers 1, but may be multiplexed by three or more. When the cache interface device is multiplexed by three or more cache servers 1, at least two of them are assumed to be operating as an act system, and the load balancer 2 performs the above-described processing for the act cache server 1. Then, one of the act cache servers 1 is determined as the active system.
Moreover, it is good also as a structure provided with the some reference part 11 in the one housing | casing (server) which operate | moves as the cache server 1. FIG.

  Note that the cache server 1, the load balancer 2, the original data management device 3, and the application device 4 described above have a computer system therein. Then, the monitoring processing unit 112, the monitoring result notification processing unit 113, the reference data management unit 114, the information reference processing unit 115, the intra-system synchronization processing unit 116, the monitoring processing unit 152, the monitoring result notification processing unit 153 of the cache server 1, transfer Data management unit 154, system state identification unit 155, information transfer processing unit 157, inter-system transfer processing unit 158 and intra-system synchronization processing unit 159, service check unit 22 and operational system determination processing unit 23 of load balancer 2, original data management The operation process of the data management unit 31 of the apparatus 3 and the application apparatus 4 is stored in a computer-readable recording medium in the form of a program, and the computer system reads out and executes this program, whereby the above processing is performed. Is done. The computer system here includes a CPU, various memories, an OS, and hardware such as peripheral devices.

Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

DESCRIPTION OF SYMBOLS 1, 1-0, 1-1 ... Cache server 11, 11-0, 11-1 ... Reference part 111, 151 ... Timer 112, 152 ... Monitoring processing part 113, 153 ... Monitoring result notification processing part 114 ... Reference data management Unit 115 ... information reference processing unit 116, 159 ... intra-system synchronization processing unit 121, 161 ... monitoring result storage unit 122 ... reference data access unit 123 ... reference data storage unit 15, 15-0, 15-1 ... transfer unit 154 ... Transfer data management unit 155 ... System state identification unit 156 ... Synchronization information storage unit 157 ... Information transfer processing unit 158 ... Inter-system transfer processing unit 162 ... Transfer data access unit 163 ... Transfer data storage units 2, 2a, 2b ... Load balancer 21 ... Timer 22 ... Service check part 221 ... Storage part 23 ... Active system judgment processing part 3 ... Original data management device 31 ... Data management part 32 ... Update history storage part 3 ... original data storage unit 4 ... application device

Claims (6)

A cache system comprising: a plurality of cache devices that cache original data; and an access distribution device that manages to which of the plurality of cache devices a received signal is distributed,
Each of the plurality of cache devices
A reference unit that receives the data access request distributed from the access distribution device, and performs access processing to the original data cached by the own cache device according to the received data access request;
A transfer unit that reads the original data from the original data management device and updates the original data cached by the own cache device;
The access distribution device
A service check unit that obtains an operation state of each of the reference unit and the transfer unit of the plurality of cache devices in an act state;
Operation for determining the cache device that is a distribution destination of the received signal from among the plurality of cache devices based on the operation states of the reference unit and the transfer unit of the plurality of cache devices acquired by the service check unit A system determination processing unit,
A cache system characterized by that. When the operational system determination processing unit determines that the plurality of reference units are normal based on the operational state of the reference unit, the transfer unit from among the plurality of cache devices including the normal reference unit The cache device including the transfer unit that is determined to be normal based on the operation state of
The cache system according to claim 1. The transfer unit of each of the plurality of cache devices is
A system state identification unit for transmitting a system state identification signal to the distribution device;
An information transfer processing unit that reads the original data from the original data management device and updates the original data cached by the own cache device when the system state identification signal transmitted from the distribution device is received;
An intersystem transfer processing unit that updates original data cached by another cache device with the original data read by the information transfer processing unit;
The operational system determination processing unit of the access distribution device transmits the received system state identification signal to the cache device determined as a distribution destination.
The cache system according to claim 1, wherein the cache system is a cache system. When the information transfer processing unit receives the system state identification signal transmitted by itself, the information transfer processing unit updates the original data management device from the original data management device rather than the original data cached by the own cache device. Read the original data with the new date and time, update the original data cached by its own cache device,
The inter-system transfer processing unit uses the original data cached by the own cache device based on the original data whose update date and time in the original data management device is newer than the original data cached by the other cache device. Updating the original data cached by the other cache device;
The cache system according to claim 3. The transfer unit further includes a transfer data storage unit that stores the original data cached by the own cache device,
The information transfer processing unit writes the original data read from the original data management device to the transfer data storage unit,
The reference section is
A reference data storage unit for storing the original data cached by the own cache device;
An in-system synchronization processing unit that writes the original data read from the transfer data storage unit to the reference data storage unit;
A reference data access unit that performs access processing to the original data stored in the reference data storage unit in response to the received data access request,
The cache system according to any one of claims 1 to 4, characterized in that: A cache access method used in a cache system including a plurality of cache devices that cache original data and an access distribution device that manages to which of the plurality of cache devices a received signal is distributed,
Each of the plurality of cache devices
A reference unit that receives the data access request distributed from the access distribution device, and performs access processing to the original data cached by the own cache device according to the received data access request;
A transfer unit that reads the original data from the original data management device and updates the original data cached by the own cache device;
In the access distribution device,
A service check unit that acquires an operation state of each of the reference unit and the transfer unit of the plurality of cache devices in an act state; and
Based on the operational states of the reference unit and the transfer unit of the plurality of cache devices acquired in the state acquisition process, the active system determination processing unit becomes a distribution destination of the received signal from the plurality of cache devices. An operational determination process for determining the cache device;
A cache access method comprising:

Images