JP4812848B2 - Clustering server system and data transfer method - Google Patents

Description

  The present invention relates to an HA (High Availability) clustering server system that is made redundant using an active server and a standby server.

  Data transfer between systems in an HA (High Availability) cluster is known as data replication. Non-Patent Document 1 describes a systematic explanation regarding this data replication.

  As a forum for formulating the use of HA cluster middleware and applications, there is SAF (Service Availability Forum), and an interface called AIS (Application Interface Specification) is defined. As a data replication interface in AIS, CKPT (Check Point Service) described in Non-Patent Document 2 exists.

"Blueprints for High Availability: Timely, Practical, Reliable", Evan Marcus (Author), Hal Stern (Author), Chapter18 Date Replication `` Service Availability Forum application Interface Specification, Check Point Service, SAI-AIS-CKPT-B.02.02 '', Service Availability Forum

  The failover clustering server system in the HA cluster that realizes high availability is equipped with an active server and a standby server, so that if a failure occurs in the active server, the standby server immediately continues processing. It is necessary to synchronize and match data between the active server and the standby server. For this reason, when the data of the active server is updated, the update data is transferred to the standby server, and the data of the standby server is also updated in the same manner as the active server.

  Here, the method for transferring update data from the active server to the standby server includes the transfer method at the time of system switchover in which update data is transferred to the standby server at the time of system switchover (failover), and the update of each data. There are two methods: an in-operation transfer method in which update data is transferred to the standby server each time.

  In order to reduce the processing load when transferring multiple update data with different update frequencies to the standby server in the active server, update data with a low update frequency is used for the update data with a high update frequency by the transfer method during system switchover. Data must be transferred using the in-service transfer method. This is because, when frequently updated data that is frequently updated is transferred using the in-operation transfer method, the load on the active server and standby server increases, resulting in a decrease in the functionality of the entire system. Because there is a fear.

  On the other hand, with the transfer method during system switchover, the amount of data that can be transferred at the time of system switchover is limited due to the limitation of the network bandwidth between the active server and the standby server and the service interruption time due to system switchover. There are limitations.

  However, it is difficult to accurately predict the update frequency of each data updated at various timings due to requests from a plurality of user terminals, and therefore a transfer method suitable for the update frequency of each data is determined in advance. It is also difficult to assign. For example, when the update frequency of each data is predicted using the past traffic pattern, and the predicted data with low update frequency is assigned to the in-operation transfer method, the traffic pattern changes and the data update frequency becomes high after that. Each time data is updated, the data is transferred to the standby server, which increases the processing load and increases the possibility of affecting other service provisions.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to suppress the amount of data transferred by the operating transfer method and to further reduce the processing load of the clustering server system.

  In order to achieve the above object, the present invention provides a clustering server system having an active server and a standby server, and as a data transfer method between the active server and the standby server, The system for transfer during system switchover that transfers update data to the switch destination server in batch and the operation transfer method that transfers update data to the switch destination server for each data update are used together. Data storage means for storing data, transfer list storage means for storing the update frequency of updated data among a plurality of data stored in the data storage means, and between the active server and the standby server Inter-system transfer management means for transferring update data in the inter-system transfer management means, wherein the inter-system transfer management means calculates the total amount of update data stored in the transfer list storage means. Determining whether the total amount of data exceeds a predetermined threshold that can be transferred by the transfer method at the time of system switching, and if the total amount of data exceeds the threshold, until the total amount of data falls below the threshold, Update data with the lowest update frequency stored in the transfer list storage means is transferred to the standby server by the in-operation transfer method.

The present invention is also a data transfer method for a clustering server system having an active server and a standby server, and the clustering server system is a data transfer method between the active server and the standby server. As described above, a combination of the transfer method during system switchover that transfers update data to the switch destination server at the time of system switchover and the in-operation transfer method that transfers update data to the switch destination server for each data update, The server includes a data storage unit that stores a plurality of data, a transfer list storage unit that stores the update frequency of updated data among the plurality of data stored in the data storage unit, and intersystem transfer management comprising a part, the said inter-system transfer management unit, and updating step of updating the forwarding list storage unit, updates data stored in the transfer list storage unit A calculation step for calculating the total amount of data, a determination step for determining whether or not the calculated total amount of data exceeds a predetermined threshold that can be transferred by the transfer method at the time of system switching, and the total amount of data exceeds the threshold A transfer step of transferring update data having the lowest update frequency stored in the transfer list storage unit to the standby server by the in-operation transfer method until the total data amount becomes equal to or less than the threshold value. .

  According to the present invention, the update frequency of each data is dynamically calculated, and the data transferred by the operating transfer method is specified based on the calculated update frequency, so that the amount of data transferred by the operating transfer method is suppressed. Thus, the processing load of the clustering server system can be further reduced.

1 is an overall configuration diagram of a clustering server system according to an embodiment of the present invention. 1 is a block diagram showing a configuration of a clustering server system 1. FIG. It is a 1st explanatory view for explaining data memorized by memory. It is the 2nd explanatory view for explaining data memorized by memory. It is a figure which shows an example of the data structure of a transfer list memory | storage part. It is a flowchart which shows the process at the time of data update.

  Hereinafter, embodiments of the present invention will be described.

  FIG. 1 is an overall configuration diagram showing a clustering server system according to an embodiment of the present invention. The clustering server system 1 according to the present embodiment includes an active server (active server) 10 that operates normally and a standby server 11 that operates when the active server 10 fails. In this embodiment, the clustering server system 1 is connected to a plurality of user terminals 3 via the network 2.

  The clustering server system 1 according to the present embodiment is a system for transferring update data to a switching destination server at the time of system switching (failover) as a method for transferring update data between an active server and a standby server. Two methods are used in combination: a transfer method during switching and an in-operation transfer method in which update data is transferred to the switching destination server each time each data is updated.

  The active server 10 and the standby server 11 can use a general-purpose computer system including at least a CPU, a memory, and an external storage device such as an HDD. In this computer system, each function of each device is realized by the CPU executing a predetermined program loaded on the memory. For example, the functions of the active server 10 and the standby server 11 are the functions of the CPU of the active server 10 in the case of the program for the active server 10, and the standby server 11 in the case of the program for the standby server 11. This is realized by the CPU executing each.

  FIG. 2 is a block diagram showing a configuration of the clustering server system 1. The active server 10 includes a memory 100, an intersystem transfer management unit 101, an application unit 102, and a transfer list storage unit 103. Similarly to the active server 10, the standby server 11 includes a memory 110, an intersystem transfer management unit 111, an application unit 112, and a transfer list storage unit 113. The active server 10 and the standby server 11 are connected by an inter-system network 12.

  A plurality of data is stored in the memory 100 (data storage means) of the active server 10.

  FIG. 3 is an explanatory diagram for explaining each data stored in the memory 100. Each data (X, Y, Z) 301 stored on the memory 100 includes, as management information of the data 301, an update count 302, a last update time 303, an average update frequency 304 as an update frequency, It is assumed that a data amount 305 and an update flag 306 are added.

  The application unit 102 of the active server 10 updates the desired data 301 itself stored in the memory 100 in response to various requests transmitted from the plurality of user terminals 3 or by internal processing of the active server 10. At the same time, the management information 302, 303, 304, 306 of the data 301 is updated.

That is, for example, when the data X on the memory 100 is updated, the application unit 102 adds “1” to the update count 302 of the data X and updates the last update time 303 to the time when the data X is updated. Then, the application unit 102 updates the update count N after the update, the last update time t _N after the update, the last update time t _N-1 before the update (previous), and the average update frequency α before the update (previous). _{Using N−1} , for example, the updated average update frequency α _N is calculated by the following equation 1, and the calculated value α _N is set as the average update frequency 304.

Average update frequency α _N = N / {(t _N −t _N−1 ) + (N−1) / α _N−1 } Equation 1
Further, when the data X on the memory 100 is updated, the application unit 102 sets a flag (for example, “1”) indicating that an update has occurred in the update flag 306. If the flag has already been set, the state is left as it is.

  The data amount 305 is set at the timing when the data X is allocated to the memory 100, and is not changed (not updated) when the data is updated.

  Even when other data (data Y, data Z) on the memory is updated, the application unit 102 updates the management information 302, 303, 304, 306 of the data, triggered by the data update.

  Instead of the average update frequency 304, a weighted average update frequency may be used as the update frequency.

  FIG. 4 is an explanatory diagram for explaining each data stored in the memory 100, and is different from FIG. 3 in that a weighted average update frequency 404 is used as an update frequency. It is the same. That is, in FIG. 4, each data (X, Y, Z) 401 stored in the memory 100 includes the update count 402, the last update time 403, and the update frequency as management information of the data 401. It is assumed that a weighted average update frequency 404, a data amount 405, and an update flag 406 are added.

  As described with reference to FIG. 3, the application unit 102 updates desired data 401 itself stored in the memory 100 and also updates management information 402, 403, 404, and 406 of the data 401. Since the update count 402, the last update time 403, and the update flag 406 are updated in the same manner as in FIG. 3, the description thereof is omitted here.

The weighted average update frequency 404 is a weighted average of update frequencies obtained by adding a weight ω to the latest update interval time. That is, the weighted average update frequency 404 is calculated based on the weighted average update frequency up to the previous time and the current update interval time that has been weighted in a predetermined manner. The application unit 102 uses the weight ω, the last update time t _N after update, the last update time t _N−1 before update (previous), and the weighted average update frequency α _N−1 before update (previous). Thus, for example, the updated weighted average update frequency α _N is calculated by the following expression 2, and the calculated value α _N is set as the weighted average update frequency 404. As the weight ω, a predetermined value determined in advance is used.

Weighted average update frequency α _N = 1 / {(t _N −t _N−1 ) ω + (1−ω) / α _N−1 } Equation 2
The memory 110 of the standby server 11 is the same as the memory 100 of the active server 10.

FIG. 5 shows an example of the transfer list storage unit 103 managed by the intersystem transfer management unit 101 of the active server 10. The transfer list storage unit 103 may be stored on the memory 100 or may be stored on an external storage device such as an HDD. In the transfer list storage unit 103, an update occurs in the data stored in the memory 100 and a transfer list 501 of update data that needs to be transferred to the standby server 11, a threshold M _th 503, and a total data amount M504 is stored.

  The transfer list 501 shown in the figure has, for each update data, a memory ID (such as an address on the memory) for pointing to the update data, and an update frequency of data stored in the memory ID. As the update frequency, the average update frequency 304 described in FIG. 3 or the weighted average update frequency 404 described in FIG. 4 is used.

The threshold M _th 503 is set with a maximum data amount (threshold) that can be transferred in a batch by the transfer method during system switching. The threshold value M _th 503 is set to a predetermined value in advance due to the limitation of the bandwidth of the inter-system network 12 and the service interruption time due to system switching. The total amount of update data registered in the transfer list 501 is set in the total data amount M504.

  When the processing is switched to the standby server 11 due to a failure of the active server 10 or the like, the same information as in FIG. 5 is generated and stored in the transfer list storage unit 113 of the standby server 11.

  Next, a process when the data stored in the memory 100 of the active server 10 is updated will be described.

  FIG. 6 is a flowchart in the case where the information in the transfer list storage unit 103 is updated at a timing when the data stored in the memory 100 is updated, and the update data exceeding the threshold is transferred by the operation transfer method. .

  First, the application unit 102 of the active server 10 receives predetermined data stored in the memory 100 in response to various requests transmitted from a plurality of user terminals 3 or by internal processing of the active server 10. In addition to updating, the management information (update count, last update time, update frequency, update flag) of the data is updated (S1). The update frequency is the average update frequency in the case of FIG. 3 and the weighted average update frequency in the case of FIG.

  Then, the intersystem transfer management unit 101 of the active server 10 registers (adds) the update data updated in step 1 in the transfer list 501 of the transfer list storage unit 103 in response to the data update of S1 (S2). ). Specifically, the inter-system transfer management unit 101 acquires the memory ID that points to the update data on the memory 100 and the update frequency of the update data updated in S1 from the memory 100, and stores it in the transfer list 501. sign up.

  Then, the inter-system transfer management unit 101 calculates the total amount M504 of update data registered in the transfer list 501 and sets it in the transfer list storage unit 103 (S3). Specifically, the inter-system transfer management unit 101 acquires the amount of each update data registered in the transfer list 501 from the memory 100 based on each memory ID, and calculates the total amount of each acquired data. calculate.

Then, the inter-system transfer management unit 101 compares the calculated total data amount M504 and the threshold value M _th 503 to determine whether or not the total data amount M exceeds the threshold value M _th (M> M _th ) (S4). .

When the total data amount M does not exceed the threshold value M _th , that is, falls below (M ≦ M _th ) (S4: NO), the inter-system transfer management unit 101 ends the process triggered by the data update in S1. In this case, all the update data registered in the transfer list 501 is managed at this time as a transfer method at the time of system switching that is transferred collectively at the time of system switch.

If the data amount M exceeds a threshold value M _th (S4: YES), the inter-system transfer management section 101 specifies the most updated infrequently updated data in the update data registered in the transfer list 501, the update Data is transferred to the memory 110 of the standby server 11 via the intersystem network 12 by the operating transfer method (S5). Thereby, the inter-system transfer management unit 101 deletes the information (record) of the update data transferred from the transfer list 501. Further, the intersystem transfer management unit 101 accesses the memory 100 and deletes the update flags 306 and 407 (see FIGS. 3 and 4) of the transferred update data.

Then, the inter-system transfer management unit 101 returns to S3, recalculates the total data amount of each update data registered in the transfer list 501 after deleting the transferred update data information, and the total data amount M is the threshold value M _th. Is exceeded (S4), and if it exceeds, the process of S5 is performed. Thus, the inter-system transfer management section 101, data amount M of each update data registered in the transfer list 501 repeats the process from S3 to below the threshold M _th S5, minute update exceeding the threshold value M _th The data is transferred to the standby server 11 by the operating transfer method in order from update data with a low update frequency.

  In the processing shown in FIG. 6, the processing from S2 to S5 is performed at the timing when the data on the memory 100 is updated. However, the present invention is not limited to this, and the processing from S2 to S5 is performed at various timings. Processing can be performed.

  For example, the inter-system transfer management unit 101 performs the processing from S2 to S5 at a predetermined cycle (time interval) determined in advance, updates the information in the transfer list storage unit 103, and updates data that exceeds the threshold May be transferred by the transfer method during operation. In this case, the application unit 102 updates data and management information on the memory 100 as needed by a request from the user terminal 3 or internal processing of the active server 10, while the intersystem transfer management unit 101 The processes from S2 to S5 are performed in a cycle.

  Specifically, the intersystem transfer management unit 101 searches for updated data in S2 after performing the processing from S2 to S5 in the previous cycle. For example, the intersystem transfer management unit 101 refers to the management information (FIGS. 3 and 4) of each data held in the memory and searches for data for which an update flag (for example, “1”) is set. . Then, each searched update data is registered (added) in the transfer list 501 of the transfer list storage unit 103 (S2). At this time, if the searched update data is already registered in the transfer list 501, the inter-system transfer management unit 101 updates the update frequency of the update data to the latest value acquired from the memory 100. The processing from S3 to S5 is the same as that from S3 to S5 described in FIG. By performing the processing from S2 to S5 at a predetermined cycle, it is possible to avoid an increase in processing load even when each data in the memory 100 is updated frequently.

  In the present embodiment described above, the amount of data transferred by the in-operation transfer method is selected because the update frequency of each data is dynamically calculated and data to be transferred by the in-operation transfer method is selected based on the calculated update frequency. And the processing load on the clustering server system can be further reduced.

  In other words, in the present embodiment, when data updated in the active server 10 is transferred to the standby server 11, update data with a low update frequency is transferred each time it is updated by the operating transfer method. It is possible to dynamically control so that the amount of data transferred by the system switching transfer method that performs batch transfer at the time of system switching (failover) is always below the threshold value.

  Also, in this embodiment, when the total amount of each update data is less than the data amount (threshold value) that can be transferred at the time of system switching, the transfer method is the transfer method at the time of system switching for all update data, and the total amount of each update data If the amount of data that can be transferred during system switchover (threshold) is exceeded, update data with low update frequency is transferred as the operating transfer method, and update data with high update frequency is transferred as the transfer method at system switchover. In addition, the transfer method is optimized according to the update frequency. Thereby, in this embodiment, even in an environment where traffic cannot be predicted, the processing load of the clustering server system can be suppressed, and a decrease in service provision to the user terminal can be prevented as much as possible.

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

DESCRIPTION OF SYMBOLS 1 Clustering server system 2 Network 3 User terminal 10 Active server 11 Standby server 12 Intersystem network 100, 110 Memory 101, 111 Intersystem transfer management part 102, 112 Application part 103, 113 Transfer list memory | storage part

Claims (8)

A clustering server system having an active server and a standby server, and as a data transfer method between the active server and the standby server, the update data is transferred to the switch destination server at the time of system switchover. Use both the transfer method during system switchover and the in-operation transfer method that transfers update data to the switch destination server for each data update.
The active server is
Data storage means for storing a plurality of data;
Among a plurality of data stored in the data storage means, transfer list storage means for storing the update frequency of updated update data;
An intersystem transfer management means for transferring update data between the active server and the standby server,
The inter-system transfer management unit calculates the total data amount of the update data stored in the transfer list storage unit, and determines whether or not the total data amount exceeds a predetermined threshold that can be transferred by the system switching transfer method If the total amount of data exceeds the threshold value, the standby data is transferred to the standby system using the in-operation transfer method with the update data having the lowest update frequency stored in the transfer list storage unit until the total data amount becomes equal to or less than the threshold value. A clustering server system characterized by transferring to a server. The clustering server system according to claim 1,
The inter-system management transfer means calculates the total data amount of the update data every predetermined cycle or when data in the data storage means is updated, and determines whether the total data amount exceeds the threshold value. If the total amount of data exceeds the threshold value, the update data stored in the transfer list storage unit with the lowest update frequency is stored in the standby transfer method in the active transfer method until the total data amount becomes equal to or less than the threshold value. A clustering server system characterized by being transferred to a system server. The clustering server system according to claim 1 or 2,
The clustering server system, wherein the update frequency is an average update frequency calculated based on the number of updates of the update data, a current update interval time, and a previous average update frequency. The clustering server system according to claim 1 or 2,
The clustering server system characterized in that the update frequency is a weighted average update frequency calculated based on a weighted average update frequency up to the previous time and a predetermined weighted current update interval time. A data transfer method for a clustering server system having an active server and a standby server,
The clustering server system is a data transfer method between the active server and the standby server, and a transfer method at the time of system switchover that transfers update data to the switch destination server at the time of system switchover. In conjunction with the in-operation transfer method of transferring update data to the switchover destination server,
The active server is
A data storage unit for storing a plurality of data;
Among a plurality of data stored in the data storage unit, a transfer list storage unit that stores the update frequency of updated update data ;
An intersystem transfer management unit ,
The intersystem transfer management unit
An update step of updating the transfer list storage unit;
A calculation step of calculating the total amount of update data stored in the transfer list storage unit;
A determination step of determining whether or not the calculated total amount of data exceeds a predetermined threshold that can be transferred by the transfer method at the time of system switching;
When the total data amount exceeds the threshold value, the update data with the lowest update frequency stored in the transfer list storage unit is transferred to the standby server by the in-operation transfer method until the total data amount becomes the threshold value or less. A data transfer method comprising: performing a transfer step. The data transfer method according to claim 5, wherein
The intersystem transfer management unit
The data transfer method, wherein the update step, the calculation step, the determination step, and the transfer step are performed at predetermined intervals or when data in the data storage unit is updated. The data transfer method according to claim 5 or 6, wherein
The intersystem transfer management unit
A data transfer method further comprising a calculation step of calculating an average update frequency as the update frequency based on the number of updates of the update data, the current update interval time, and the previous average update frequency. The data transfer method according to claim 5 or 6, wherein
The intersystem transfer management unit
A data transfer method further comprising a calculation step of calculating a weighted average update frequency based on a weighted average update frequency up to the previous time and a predetermined weighted update interval time as the update frequency.

Images