WO2014141393A1 - データベースシステム、プログラムおよびデータ処理方法 - Google Patents
データベースシステム、プログラムおよびデータ処理方法 Download PDFInfo
- Publication number
- WO2014141393A1 WO2014141393A1 PCT/JP2013/056868 JP2013056868W WO2014141393A1 WO 2014141393 A1 WO2014141393 A1 WO 2014141393A1 JP 2013056868 W JP2013056868 W JP 2013056868W WO 2014141393 A1 WO2014141393 A1 WO 2014141393A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- node
- nodes
- unit
- backup
- data set
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2282—Tablespace storage structures; Management thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
- G06F16/275—Synchronous replication
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/27—Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
- G06F16/278—Data partitioning, e.g. horizontal or vertical partitioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/16—Error detection or correction of the data by redundancy in hardware
- G06F11/1658—Data re-synchronization of a redundant component, or initial sync of replacement, additional or spare unit
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/16—Error detection or correction of the data by redundancy in hardware
- G06F11/20—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements
- G06F11/202—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements where processing functionality is redundant
- G06F11/2023—Failover techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/16—Error detection or correction of the data by redundancy in hardware
- G06F11/20—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements
- G06F11/202—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements where processing functionality is redundant
- G06F11/2041—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements where processing functionality is redundant with more than one idle spare processing component
Definitions
- Embodiments described herein relate generally to a database system, a program, and a data processing method.
- a shared nothing type database system includes a plurality of database nodes connected to each other via a network.
- Each of the plurality of database nodes is a computer having an individual resource such as a processor, memory, and storage.
- resources are completely separated for each database node, so that access contention does not occur between database nodes. Therefore, the shared nothing type database system can linearly improve performance by increasing the number of database nodes. That is, the shared nothing type database system is a system that can efficiently realize scale-out by adding database nodes.
- Partitioning In a shared nothing database system, a technique called partitioning can be applied. Partitioning is a technique in which a database is divided into a plurality of data sets, and each divided data set (called a partition) is distributed and stored in a plurality of database nodes. Since the shared nothing type database system to which partitioning is applied can reduce the data set stored in each database node, the access from the client can be distributed to a plurality of database nodes.
- Sharding is a technology that divides data into smaller data and automatically distributes and stores the data in a plurality of database nodes.
- Sharding is a technique similar to partitioning shared-nothing database systems.
- the key-value type database is a database based on a simple data model composed of pairs of keys and values in the same manner as an associative array or Map in a programming language.
- the key-value database Since the key-value database has a simple data model, sharding and partitioning are easy.
- the distributed key-value database can process a large-scale data set that cannot be handled by a single key-value database node using a plurality of key-value database nodes. Therefore, by applying the distributed key-value database to a shared-nothing database system including a plurality of database nodes, a large-scale database system that can be scaled out can be realized.
- a shared-nothing database system (hereinafter referred to as a scale-out database system) to which such a distributed key-value database is applied makes it possible to access data from clients by arranging data evenly on multiple database nodes. Can be distributed to improve throughput and response as a whole.
- a technique called replication in which data replication (replica) is stored in another database node is applied.
- the scale-out type database system can continue the service using the database node storing the replica even if a failure occurs in a certain database node by performing replication.
- the number of database nodes that store replicas of data is called redundancy. In the scale-out database system, for example, 2 or 3 are set as the redundancy.
- the redundancy becomes lower than a preset value. If a scale-out database system continues to operate for a long time with reduced redundancy, there is a high possibility that the entire system will stop operating or a wide range of data will be lost if a further database node failure occurs. . For this reason, in a scale-out type database system, after a failure occurs in a database node, a new database node is added, and a replica is stored in the added database node to restore the redundancy to the original level. .
- the amount of data to be stored may become unbalanced between database nodes depending on the data arrangement.
- the access amount from the client may be unbalanced between the database nodes depending on the data arrangement.
- the scale-out database system performs data relocation processing to multiple database nodes when imbalance occurs in the data arrangement. There must be.
- the database system includes a plurality of nodes that store a database and a management device that manages the plurality of nodes.
- the management device includes a first allocation unit and a second allocation unit.
- the first allocation unit stores a data set and accepts an access request to the data set in response to a change in the status of each of the plurality of nodes, and stores the data set and backs up the first node
- the second node is assigned, and an operation corresponding to the assignment is instructed to each of the plurality of nodes.
- the second assigning unit assigns a third node that is a candidate for the first node or the second node in accordance with the situation of the plurality of nodes and the change of the assignment situation of the first node and the second node, In each case, the third node is instructed to prepare to operate as the first node or the second node.
- Each of the plurality of nodes includes a transfer unit that transmits the data set of the first node or the second node to the third node.
- FIG. 1 is a diagram illustrating a hardware configuration of a database system according to the embodiment.
- FIG. 2 is a diagram illustrating a functional configuration of the database system according to the embodiment.
- FIG. 3 is a diagram illustrating an example of the data holding table.
- FIG. 4 is a diagram illustrating an example of the current role table.
- FIG. 5 is a diagram illustrating an example of the next role table.
- FIG. 6 is a diagram illustrating an example of the target role table.
- FIG. 7 is a diagram illustrating an example of a data set stored in the data store unit.
- FIG. 8 is a diagram for explaining the replication processing.
- FIG. 9 is a diagram for explaining the short-term synchronization processing.
- FIG. 10 is a diagram for explaining the long-term synchronization processing.
- FIG. 11 is a diagram illustrating transition of roles assigned to nodes.
- FIG. 12 is a diagram illustrating a processing flow of the cluster management unit of the management apparatus.
- FIG. 13 is a diagram illustrating a flow of a next role table calculation process by the first assignment unit of the management apparatus.
- FIG. 14 is a diagram illustrating a flow of target role table calculation processing by the second assignment unit of the management device.
- FIG. 15 is a diagram illustrating a processing flow of the access processing unit of the node.
- FIG. 16 is a diagram illustrating a processing flow of the node management unit of the node.
- FIG. 17 is a diagram illustrating a processing flow of the transfer unit of the node.
- FIG. 18 is a diagram illustrating a state of each of the nodes 30-A to 30-C when an owner and a backup node are assigned to each of the three partitions # 1 to # 3.
- FIG. 19 is a diagram showing the states of the nodes 30-A to 30-C after the short-term synchronization processing is executed in response to the occurrence of a failure in the node 30-C from the state of FIG.
- FIG. 20 is a diagram illustrating an example of long-term synchronization processing executed from the state of FIG.
- FIG. 21 is a diagram showing the states of the nodes 30-A to 30-C after the short-term synchronization processing is executed in response to the completion of the long-term synchronization processing of FIG.
- FIG. 19 is a diagram showing the states of the nodes 30-A to 30-C after the short-term synchronization processing is executed in response to the completion of the long-term synchronization processing of FIG.
- FIG. 22 is a diagram illustrating the states of the node 30-A to the node 30-D when the node 30-D is added from the state of FIG.
- FIG. 23 is a diagram illustrating an example of long-term synchronization processing executed from the state of FIG.
- FIG. 24 is a diagram showing the states of the nodes 30-A to 30-D after the short-term synchronization processing is executed in response to the completion of the long-term synchronization processing of FIG.
- FIG. 1 is a diagram illustrating a hardware configuration of a database system 10 according to the embodiment.
- the database system 10 receives an access request (for example, reference, update, or registration) to a database from a client, which is an external computer, via a network, and executes processing according to the received access request.
- an access request for example, reference, update, or registration
- the database system 10 includes a management device 20 and a plurality of nodes 30.
- the management device 20 and the plurality of nodes 30 are connected to each other via a network. Although two nodes 30 are shown in FIG. 1, the database system 10 may include three or more nodes 30.
- the database system 10 is a database system capable of scaling out by applying a distributed key-value database to a shared-nothing database.
- a data set is distributed to a plurality of nodes 30, but the database system 10 is accessed as a huge single database from a client that is an external computer.
- Each of the management device 20 and the plurality of nodes 30 is a computer independent of each other in terms of hardware.
- Each of the plurality of nodes 30 can independently accept an access request from a client, and can independently perform processing according to the access request.
- Each of the management device 20 and the plurality of nodes 30 includes, for example, a CPU (Central Processing Unit) 12, a memory 13, a communication I / F 14, and a storage I / F 15.
- Each of the management device 20 and the plurality of nodes 30 is connected to the external storage device 16 via the storage I / F 15. Note that the external storage device 16 may be provided inside the node 30.
- the database system 10 stores the database by dividing it into a plurality of partitions.
- the number of partitions is predetermined.
- the method for dividing the database may be any method as long as it is a predetermined method.
- Each of the plurality of nodes 30 is assigned to each of the plurality of partitions as to play a role as an owner, a role as a backup, or a role as an owner and a backup.
- the owner stores the data set of the corresponding partition, and accepts and processes the access request from the client to the data set of the corresponding partition.
- the backup stores the data set of the corresponding partition, and backs up the owner so that, for example, when the owner fails, the role of the owner can be assumed instead.
- it may be called a master and a slave instead of an owner and a backup.
- Management device 20 manages a plurality of nodes 30. As an example, the management device 20 assigns a node 30 as an owner and a node 30 as a backup to each of a plurality of partitions.
- the management device 20 assigns the backup node 30 so as to provide redundancy within a predetermined range so that the database is restored as much as possible even if any node 30 fails. Furthermore, the management device 20 assigns the node 30 as the owner and the backup so that the data set included in the database is distributed and arranged in the plurality of nodes 30.
- the management device 20 may be provided in any one of the nodes 30.
- the node 30 functioning as the management device 20 may be selected by an arbitrary algorithm. Further, when a failure occurs in the node 30 that functions as the management apparatus 20, another node 30 may function as the management apparatus 20.
- Such a database system 10 can disconnect a node 30 in which a failure has occurred when a failure occurs in any one of the plurality of nodes 30.
- the database system 10 can add a new node 30 to improve the performance of the database.
- FIG. 2 is a diagram illustrating a functional configuration of the database system 10 according to the embodiment.
- the management device 20 includes a table storage unit 21, a first allocation unit 22, a second allocation unit 23, and a cluster management unit 24.
- the first allocation unit 22, the second allocation unit 23, and the cluster management unit 24 are realized by the CPU 12 of the management apparatus 20 executing a program.
- the first allocation unit 22, the second allocation unit 23, and the cluster management unit 24 may be partially or entirely implemented by hardware circuits.
- the table storage unit 21 is realized by the memory 13 of the management device 20 or the external storage device 16.
- the table storage unit 21 stores four tables for determining the roles of the plurality of nodes 30 for the plurality of partitions.
- Each of the four tables may be data in a table format or data other than the table format.
- the table storage unit 21 stores a data holding table, a current role table, a next role table, and a target role table.
- the data retention table stores the time stamps of the plurality of nodes 30 for each of the plurality of partitions.
- the time stamp represents the update history of the data set of the corresponding partition stored in the corresponding node 30.
- the time stamp is a value that is incremented every time the data set is updated. Therefore, the node 30 having the largest time stamp in a certain partition means that the node 30 stores the latest data set of the partition.
- the data holding table stores the size or number of data of the data set of the corresponding partition stored in the corresponding node 30 instead of the time stamp. Also good.
- the data holding table for example, as shown in FIG. 3, the node 30 is specified by the row, and the partition is specified by the column.
- the data holding table stores the time stamp of the data set of the partition specified by the column stored in the node 30 specified by the row in each cell in which the row and the column are orthogonal.
- the current role table stores the role in the database assigned to each of the plurality of nodes 30 for each of the plurality of partitions. Specifically, the current role table is assigned to each of a plurality of partitions, each node 30 being assigned as an owner, assigned as a backup, or assigned as an owner or a backup. Stores whether or not.
- the node 30 is specified by a row, and the partition is specified by a column.
- the current role table stores the roles assigned to the node 30 specified by the row for the partition specified by the column in each cell where the row and the column are orthogonal.
- “OWNER” or “O” indicates that the role as the owner is assigned
- “BACKUP” or “B” indicates that the role as the backup is assigned.
- “None” or “N” is entered or blank to indicate that neither the owner nor backup is assigned.
- the next role table stores a role assigned next to each of the plurality of nodes 30 for each of the plurality of partitions. That is, in the next role table, for each of a plurality of partitions, each of the plurality of nodes 30 is next assigned as an owner, assigned as a backup, or assigned as an owner or a backup. Stores whether or not The next role table is a table that is replaced with the current role table, for example, when a failure occurs in the node 30 or when a new node 30 is added.
- next role table As an example, in the next role table, as shown in FIG. 5, the node 30 is specified by the row, and the partition is specified by the column.
- the next role table stores the next role assigned to the node 30 specified by the row with respect to the partition specified by the column in each cell at the position where the row and the column are orthogonal.
- the target role table stores roles as backup candidates assigned to each of the plurality of nodes 30 for each of the plurality of partitions.
- the backup candidate is a role of preparing for the partition 30 to assume the role of owner or backup for the partition in the future.
- the role name is “backup candidate”, but the node 30 assigned to the backup candidate may be the owner in the future.
- the target role table for example, as shown in FIG. 6, the nodes 30 are specified by rows, and the partitions are specified by columns.
- the target role table stores the role as the backup candidate assigned to the node 30 specified by the row for each partition at the position where the row and the column are orthogonal to each other, for the partition specified by the column.
- MID-BACKUP or “M” is described when a role as a backup candidate is assigned, and blank if it is not assigned as a backup candidate.
- the first assigning unit 22 assigns the owner node 30 and the backup node 30 to each of the plurality of partitions in accordance with the change in the status of each of the plurality of nodes 30. For example, when a failure occurs in any one of the plurality of nodes 30, the first assignment unit 22 reassigns the owner and the backup node 30 except for the failed node 30.
- the first assigning unit 22 reassigns the owner and backup node 30 including the new node 30 to each of the plurality of partitions.
- the first assigning unit 22 Re-assign the owner and backup node 30 to each.
- the first allocation unit 22 allocates the owner and the backup node 30 so that at least the owner exists for all of the plurality of partitions.
- the 1st allocation part 22 can function a database at least.
- the first assigning unit 22 assigns the node 30 as the owner and the backup to each of the plurality of partitions so as to have redundancy within a predetermined range on the condition that at least the database functions. . Thereby, even if a failure occurs in any of the nodes 30, the first allocation unit 22 can increase the possibility that the database can function at least.
- the first assigning unit 22 sets each of the plurality of partitions so that the owner and the backup are distributed to each of the plurality of nodes 30 as a condition that the database functions and has redundancy within a predetermined range.
- the node 30 to be the owner and backup is assigned to.
- the 1st allocation part 22 can equalize the load of each process of the some node 30.
- the first assigning unit 22 assigns the owner and the backup node 30 by calculating the next role table.
- An example of a method for calculating the next role table by the first assigning unit 22 will be described later with reference to FIG.
- the second allocation unit 23 allocates a backup candidate node 30 to each of the plurality of partitions in accordance with the status of the plurality of nodes 30 and the change of the owner allocated by the current allocation table and the backup allocation status. . For example, when a failure occurs in any one of the plurality of nodes 30, the second assignment unit 23 reassigns the backup candidate nodes 30 except for the failed node 30. For example, when a new node 30 is added, the second assignment unit 23 reassigns the backup candidate node 30 including the new node 30 to each of the plurality of partitions.
- the second allocation unit 23 allocates a backup candidate node 30 to each of the plurality of partitions so as to have a redundancy within a predetermined range in the future.
- the second allocation unit 23 can increase the possibility that at least the database can function even if a failure occurs in any of the nodes 30 in the future.
- the second assigning unit 23 assigns a backup candidate node 30 to each of a plurality of partitions so that owners and backups are distributed and assigned to the plurality of nodes 30 in the future. Thereby, the 2nd allocation part 23 can equalize the load of each process of the some node 30 in the future.
- the second assigning unit 23 assigns a node 30 as a backup candidate by calculating a target role table.
- An example of a method for calculating the target role table by the second assigning unit 23 will be described later with reference to FIG.
- the cluster management unit 24 manages each of the plurality of nodes 30 by exchanging messages with each of the plurality of nodes 30 via the network. For example, the cluster management unit 24 exchanges a message called a heartbeat with each of the plurality of nodes 30 at regular time intervals. Then, the cluster management unit 24 identifies the node 30 in which the failure has occurred depending on whether or not a heartbeat is returned.
- the cluster management unit 24 periodically receives a data holding table from each of the plurality of nodes 30. Then, the cluster management unit 24 stores the data holding table received from all the nodes 30 together in the table storage unit 21. Further, the cluster management unit 24 identifies the node 30 in which the failure has occurred depending on whether or not the data holding table has been received.
- the cluster management unit 24 allocates each of the plurality of nodes 30 by causing the first allocation unit 22 to calculate the next role table at the time of activation and distributing the calculated next role table to each of the plurality of nodes 30. The operation according to is instructed. In addition, the cluster management unit 24 causes the first assignment unit 22 to calculate the next role table periodically. Then, when the calculated next role table changes from the current role table, the cluster management unit 24 distributes the calculated next role table to each of the plurality of nodes 30 to allocate to each of the plurality of nodes 30. Instruct the appropriate action. Then, after distributing the next role table, the cluster management unit 24 updates the current role table stored in the table storage unit 21 to the contents of the next role table.
- the first assigning unit 22 changes the calculated next role table from the current role table. Therefore, in such a case, the cluster management unit 24 distributes the next role table to each of the plurality of nodes 30.
- the cluster management unit 24 periodically causes the second assignment unit 23 to calculate the target role table, and distributes the calculated target role table to each of the plurality of nodes 30.
- the cluster management unit 24 distributes the target role table to instruct each of the plurality of nodes 30 to prepare the node 30 assigned as a candidate to operate as an owner or a backup.
- the cluster management unit 24 calculates the next role table and the target role table at different periods.
- Each of the plurality of nodes 30 includes a data store unit 31, a table storage unit 32, an access processing unit 33, a node management unit 34, and a transfer unit 35.
- the access processing unit 33, the node management unit 34, and the transfer unit 35 are realized by the CPU 12 of the node 30 executing a program.
- the access processing unit 33, the node management unit 34, and the transfer unit 35 may be partially or entirely realized by a hardware circuit.
- the data store unit 31 and the table storage unit 32 are realized by the memory 13 or the external storage device 16 of the node 30.
- the data store unit 31 stores a data set of partitions in which the own node is assigned to the owner or the backup by the current role table among the plurality of partitions obtained by dividing the database. For example, as shown in FIG. 7, among partitions # 1 to # 3 divided into three, the own node is assigned to the owner for partition # 1 and assigned to the backup for partition # 3. Suppose that In this case, the data store unit 31 stores a data set of the partition # 1 and the partition # 3.
- the node 30 receives a data set from the node 30 assigned to the owner of the partition assigned to the backup candidate by the target role table.
- the data store unit 31 stores a part or all of the data set of the partition that has been assigned to the backup candidate by the target role table by the own node from the owner node 30.
- the table storage unit 32 stores a part corresponding to the own node of the data holding table.
- the table storage unit 32 stores a current role table, a next role table, and a target role table.
- the next role table is received, the current role table is rewritten with the contents of the next role table.
- the target role table is rewritten with the changed contents when the target role table received periodically changes.
- the access processing unit 33 receives an access request from the client via the network. Then, the access processing unit 33 executes a process corresponding to the role assigned to the own node by the current role table for each of the plurality of partitions.
- the access processing unit 33 receives an access request from a client for a partition assigned as an owner in the current role table, and executes a process according to the access request. For example, when receiving a reference request, the access processing unit 33 reads the corresponding data in the data set of the partition and transmits it to the client. For example, when receiving an update request, the access processing unit 33 updates the corresponding data in the data set of the partition. For example, when receiving a registration request, the access processing unit 33 registers new data in the data set of the partition.
- the access processing unit 33 may accept a transaction including a series of access requests. In this case, the access processing unit 33 executes a series of access requests from the client by transaction processing.
- the replication process is a process for generating a replica (replication) of a data set stored in the node 30 assigned to the owner in the node 30 assigned to the backup.
- the access processing unit 33 when the access processing unit 33 receives an update or registration request for the partition assigned to the owner, the access processing unit 33 receives the same access request as a backup. To the node 30. Then, the access processing unit 33 of the other node 30 receives an update or registration request from the node 30 assigned as the owner for the partition assigned as the backup, and executes the update or registration process according to the request. To do.
- the access processing unit 33 can synchronize the data set between the node 30 assigned as the owner and the node 30 assigned as the backup by performing such a replication process.
- the access processing unit 33 of the node 30 assigned as the owner may transmit a transaction including a series of access requests when executing the replication processing.
- the access processing unit 33 of the node 30 assigned as the backup executes the replication processing by transaction processing.
- the access processing unit 33 processes the update request and the registration request, the access processing unit 33 updates the time stamp of the corresponding partition of the data holding table stored in the table storage unit 32.
- the node management unit 34 exchanges messages with the cluster management unit 24 of the management apparatus 20 via the network. In response to receiving a message called heartbeat from the cluster management unit 24, the node management unit 34 transmits a response message if a failure has occurred in the node itself. Further, the node management unit 34 periodically transmits the data holding table stored in the table storage unit 32 to the cluster management unit 24.
- the node management unit 34 receives the next role table from the cluster management unit 24 and stores the received next role table in the table storage unit 32.
- the node management unit 34 accesses from the client to operate each of the plurality of nodes 30 with the role assigned by the next role table.
- Short-term synchronization processing which is synchronization processing performed by temporarily interrupting processing according to a request, is executed.
- the node management unit 34 causes the access processing unit 33 to perform the same processing as the replication processing in order to operate each of the plurality of nodes 30 with the role indicated by the next role table for each of the plurality of partitions. Execute the process. Accordingly, the node management unit 34 can synchronize the data set between the node 30 assigned as the owner and the node 30 assigned as the backup for each of the plurality of partitions. After synchronizing the data sets, the node management unit 34 rewrites the current role table to the contents of the next role table in the access processing unit 33, and assigns the new role table assigned to each of the plurality of partitions by the next role table. Process according to the appropriate role.
- the node management unit 34 updates the contents of the current role table with the contents of the next role table. Thereafter, the access processing unit 33 can accept an access request from a client.
- the node management unit 34 periodically receives the target role table from the cluster management unit 24 and stores the received target role table in the table storage unit 32.
- the transfer unit 35 is a synchronous process that is performed without interrupting the process for the access request from the client in order to operate each of the plurality of nodes 30 with the role assigned by the target role table. Execute long-term synchronization processing. Specifically, the transfer unit 35 assigns the data set of the partition to which the own node is assigned as the owner by the current role table, as a backup candidate by the target role table, and makes the backup to the owner by the current role table. Are transmitted without interruption to the other nodes 30 that are not assigned. In addition, the transfer unit 35 assigns, as an owner, a data set of partitions in which the own node is assigned as a candidate for backup in the target role table and is not assigned to either the owner or the backup in the current role table. The received request is received from the other node 30 without interruption.
- the transfer unit 35 can store a replica of the data set stored in the node 30 assigned to the owner in the node 30 that is not assigned to the owner or backup by executing such long-term synchronization processing. . As a result, the transfer unit 35 can newly generate a node 30 that can be an owner or a backup after the long-term synchronization processing is completed.
- the transfer unit 35 has assigned the data set of the partition assigned to the owner as a backup candidate in the background that does not interfere with the execution of the transaction by the access processing unit 33. Transmit to node 30. In addition, the transfer unit 35 receives a data set of partitions assigned as backup candidates from the node 30 assigned to the owner in the background.
- the background processing is, for example, processing performed using a part of CPUs 12 that are not executing transaction processing when the node 30 includes a plurality of CPUs 12.
- the background process is, for example, a process that is executed in a part of time zone in which the CPU 12 is not executing a transaction process when the CPU 12 is performing a time-sharing process. Thereby, the transfer part 35 can perform a long-term synchronization process, without causing the fall of the response speed etc. with respect to the access request from a client.
- the transfer unit 35 assigns the data set of the partition in which the own node is assigned to the backup in the current role table to other candidates that are assigned to the backup candidate by the target role table and are not assigned to either the owner or the backup. You may transmit with respect to the node 30. FIG. In this case, the transfer unit 35 performs processing on the condition that the same data set is not transmitted from another node 30.
- FIG. 11 is a diagram showing the transition of roles assigned to the nodes 30.
- the node 30 is either in a state where “owner”, “backup” or “backup candidate” is assigned to each of the plurality of partitions, or in a “none” state where no role is assigned. Transition.
- the node 30 makes a transition between the “none” state and the “owner” state by a short-term synchronization process. Further, the node 30 makes a transition between the “owner” state and the “backup” state by a short-term synchronization process. Further, the node 30 transitions from the “backup” state to the “none” state by the short-term synchronization processing.
- the node 30 makes a transition between the “none” state and the “backup candidate” state by the long-term synchronization processing. Then, the node 30 transitions from the “backup candidate” state to the “backup” state by the short-term synchronization processing.
- FIG. 12 is a diagram illustrating a processing flow of the cluster management unit 24 of the management apparatus 20.
- the cluster management unit 24 executes the following processing in steps S111 to S142.
- step S111 the cluster management unit 24 detects a reception event, a first fixed-cycle event, or a second fixed-cycle event in the data holding table.
- the data holding table reception event occurs when the cluster management unit 24 receives a data holding table transmitted from each of the plurality of nodes 30.
- the first fixed-cycle event and the second fixed-cycle event each occur periodically. However, the generation intervals of the first fixed-cycle event and the second fixed-cycle event are different from each other.
- the cluster management unit 24 When the cluster management unit 24 detects a reception event in the data holding table, the cluster management unit 24 proceeds with the process to step S121. When the cluster management unit 24 detects the first periodic event, the cluster management unit 24 proceeds with the process to step S131. When the cluster management unit 24 detects the second fixed-cycle event, the cluster management unit 24 proceeds with the process to step S141.
- step S121 the cluster management unit 24 determines whether or not it is the first reception event of the data holding table after activation.
- step S121 the cluster management unit 24 advances the process to step S122.
- step S122 the cluster management unit 24 registers the received data holding table in the table storage unit 21.
- step S123 the cluster management unit 24 causes the first assignment unit 22 to calculate the next role table.
- the next role table calculation process by the first assigning unit 22 will be described with reference to FIG.
- step S124 the cluster management unit 24 distributes the calculated next role table to each of the plurality of nodes 30. And the cluster management part 24 returns a process to step S111, after finishing the process of step S124.
- step S121 If the received event is not the first data holding table received since activation (No in step S121), the cluster management unit 24 updates the data holding table in the table storage unit 21 in step S125. And the cluster management part 24 returns a process to step S111, after finishing the process of step S125.
- step S131 the cluster management unit 24 sets the plurality of nodes 30 between the previous first fixed-cycle event and the current first fixed-cycle event. It is determined whether or not the data holding table has been received from all.
- the cluster management unit 24 advances the process to step S134.
- the cluster management unit 24 cannot receive the data holding table from any of the nodes 30 (No in Step S131)
- the cluster management unit 24 advances the process to Step S132.
- step S132 the cluster management unit 24 performs a process of separating the node 30 that has not received the data holding table from the database system 10. Subsequently, in step S133, the cluster management unit 24 deletes the contents of the detached node 30 from the data holding table and updates the data holding table. When finishing the process of step S133, the cluster management unit 24 advances the process to step S134.
- step S134 the cluster management unit 24 causes the first assignment unit 22 to calculate the next role table.
- the next role table calculation process by the first assigning unit 22 will be described with reference to FIG.
- step S135 the cluster management unit 24 determines whether or not the next role table has changed.
- the next role table changes when the failed node 30 is disconnected, when a new node 30 is added, or when the node 30 assigned as a backup changes after the long-term synchronization processing is completed.
- step S135 the cluster management unit 24 returns the process to step S111. If the next role table has changed (Yes in step S135), in step S136, the cluster management unit 24 distributes the changed next role table to each of the plurality of nodes 30. And the cluster management part 24 returns a process to step S111, after finishing the process of step S136.
- step S141 the cluster management unit 24 causes the second assignment unit 23 to calculate the target role table.
- the target role table calculation process by the second allocation unit 23 will be described with reference to FIG.
- step S142 the cluster management unit 24 distributes the calculated target role table to each of the plurality of nodes 30. And the cluster management part 24 returns a process to step S111, after finishing the process of step S142.
- FIG. 13 is a diagram illustrating a flow of a next role table calculation process by the first assignment unit 22 of the management apparatus 20.
- the first allocation unit 22 is called by the cluster management unit 24 in step S123 and step S134 of FIG. 12, and executes the processing of the following steps S211 to S220.
- step S211 the first allocation unit 22 initializes the next role table.
- the first assigning unit 22 associates each of the plurality of nodes 30 in the next role table with the node 30 in the data holding table.
- the 1st allocation part 22 can reflect the node 30 separated by the occurrence of a failure, and the newly added node 30 in a next role table.
- step S212 the first allocation unit 22 executes a loop process for each partition. For example, when dividing the database into first to third partitions, the first allocation unit 22 targets each of the first partition, the second partition, and the third partition from step S212 to step S220. Execute the process.
- step S213 the first allocation unit 22 selects a set of nodes 30 having the maximum time stamp in the data holding table for the target partition.
- the time stamp is a value that is incremented by 1 each time the data set of the target partition is updated. Accordingly, in step S213, the first allocation unit 22 can select a set of nodes 30 storing the latest data set for the target partition.
- step S214 the first allocation unit 22 selects one node 30 having the smallest number of partitions allocated as owners and backups in the next role table from the set of nodes 30 selected in step S213. Then, the selected node 30 is assigned as the owner. Thereby, the 1st allocation part 22 can allocate the node 30 with the smallest load among the group of the nodes 30 which has memorize
- the first assigning unit 22 gives priority to the node 30 having a high computing capacity over the node 30 having a low computing capacity. May be assigned as the owner. Instead, the first assigning unit 22 may preferentially assign the node 30 that receives a small number of access requests as an owner over the node 30 that receives a large number of access requests.
- step S215 the first assigning unit 22 sets a set of nodes 30 having time stamps within a certain difference from the maximum time stamp, except for the node 30 assigned as the owner for the target partition. Select. Thereby, the first allocation unit 22 can select the latest data set or the set of nodes 30 storing the data set relatively close to the latest data set for the target partition.
- the first allocation unit 22 executes a loop process for the number of replications.
- the number of replications refers to the maximum number of nodes 30 that can perform replication processing with the node 30 assigned to the owner. That is, the number of replications is the same as the number of backups that can be assigned.
- step S217 the first allocation unit 22 determines whether or not there is a node 30 that can be allocated as a backup in the set of nodes 30 selected in step S215. to decide. When there is a node 30 that can be allocated as a backup (Yes in step S217), the first allocation unit 22 advances the process to step S218.
- the first allocation unit 22 forcibly exits the loop processing for the number of replications and advances the process to step S220. That is, therefore, the first allocation unit 22 may calculate a next role table in which no backup exists or the number of backups is less than the number of replications.
- step S218 the first allocation unit 22 allocates, as a backup, the node 30 having the smallest number of partitions allocated as the owner and backup in the next role table from the set of nodes 30 selected in step S215.
- the first assigning unit 22 can assign the node 30 to be backed up in ascending order of load among the set of nodes 30 storing the latest data set or data close to the latest data set.
- the first assigning unit 22 gives priority to the node 30 having a high computing capacity over the node 30 having a low computing capacity. May be assigned as a backup. Instead of this, the first assigning unit 22 may assign the node 30 having a small number of access requests as a backup in preference to the node 30 having a large number of access requests.
- the first assigning unit 22 removes the node 30 assigned as a backup from the selected set of nodes 30, and advances the process to step S219.
- the first assigning unit 22 returns the process to step S216 if the number of processes from step S216 to step S219 is smaller than the number of replications, and if the number of processes matches the number of replications, the first allocation unit 22 performs the process in step S220. Proceed to
- step S220 the first assigning unit 22 returns the process to step S216 when the processes from step S212 to step S220 have not been executed for all partitions. Then, when the processes from step S212 to step S220 are executed for all partitions, the first allocation unit 22 ends the calculation of the next role table and exits this flow.
- the database can be functioned by assigning the node 30 as the owner to all of the plurality of partitions.
- the redundancy of the database can be ensured by assigning the backup node 30 as much as possible.
- the load of each of the some node 30 can be equalized by allocating the node 30 used as an owner and backup in an order from the node 30 with smaller load.
- FIG. 14 is a diagram showing a flow of target role table calculation processing by the second assignment unit 23 of the management apparatus 20.
- the second allocation unit 23 is called by the cluster management unit 24 in step S141 in FIG. 12, and executes the following processing in steps S311 to S324.
- step S311 the second allocation unit 23 initializes the target role table.
- the second assigning unit 23 associates each of the plurality of nodes 30 in the target role table with the node 30 in the data holding table.
- the 2nd allocation part 23 can reflect the node 30 separated by failure occurrence, and the newly added node 30 to a target role table
- step S312 the second assignment unit 23 sorts the partitions of the current role table in ascending order by the total number of assigned owners and backups. As a result, the second allocation unit 23 can allocate backup candidates in order from the partition with the few nodes 30 allocated as backup. That is, the second assigning unit 23 can assign the backup candidate nodes 30 in the current role table in order from the partition with the least redundancy.
- the second allocation unit 23 initializes the load value corresponding to each of the plurality of nodes 30.
- the load value is a value that increases in response to the backup candidate role being assigned to the node 30.
- the second allocation unit 23 initializes each load value to “0”.
- step S314 and step S324 the second allocation unit 23 performs a loop process for each partition.
- the second allocation unit 23 selects a target partition in the order of the partitions of the current role table rearranged in step S312, and executes a loop process.
- step S315 the second allocation unit 23 selects the node 30 with the maximum time stamp in the data holding table for the target partition.
- step S316 the second allocation unit 23 sorts the nodes 30 in the current role table in descending order by the value V.
- the value V is “+1” for the node 30 assigned as the owner, “+1” for the node 30 assigned as the backup, and “+1” for the node 30 with the maximum time stamp. Value. Therefore, the value V of the node 30 that is neither the owner, the backup, nor the time stamp is “0”.
- the second assigning unit 23 can assign the node 30 assigned as the owner and the backup or the node 30 storing the latest data set as the node 30 as the backup candidate first.
- the second assigning unit 23 arranges the nodes 30 at a higher rank so that the node 30 having a high computing capacity is the backup candidate first.
- the second assigning unit 23 can assign the backup candidate node 30 so that the node 30 with high computing ability is preferentially the owner or backup over the node 30 with low computing ability.
- the second allocation unit 23 may be arranged higher so that the node 30 with a small number of access requests received first becomes a backup candidate.
- the second allocation unit 23 assigns the node 30 as a backup candidate so that the node 30 with a small number of access requests can be an owner or a backup with priority over the node 30 with a large number of access requests. Can be assigned.
- step S317 the second allocation unit 23 calculates an upper limit weight MLOAD that is a constant according to the following equation (1).
- MLOAD ⁇ number of partitions ⁇ (number of replications + 1) + (number of nodes ⁇ 1) ⁇ ⁇ number of nodes (1)
- (number of replications + 1) represents the maximum number of owners and backups that can be assigned to one partition.
- Such an upper limit weight MLOAD represents an upper limit criterion of the number that can be assigned to one node 30 as a backup candidate.
- step S318 and step S323 the second allocation unit 23 executes a loop process for each node 30.
- the second allocation unit 23 selects the target node 30 in the order of the current role table rearranged in step S316, and executes a loop process.
- step S319 the second allocation unit 23 determines whether or not the backup candidate node 30 has been allocated to the (number of replications + 1) nodes 30. .
- step S319 the second allocation unit 23 allocates the maximum number of backup candidates nodes 30 to the target partition.
- step S319 If the second allocation unit 23 has not allocated backup candidates to (number of replications + 1) nodes 30 (No in step S319), the process proceeds to step S320.
- step S320 the second allocation unit 23 determines whether the load value of the target node 30 is smaller than the upper limit weight MLOAD. When the load value of the target node 30 is smaller than the upper limit weight MLOAD (Yes in Step S320), the second allocation unit 23 advances the process to Step S321.
- the second allocation unit 23 advances the process to step S323.
- the second assigning unit 23 assigns a role as a backup candidate to a certain node 30 at a reference value or more, it can avoid assigning the node 30 as a backup candidate any more.
- the second assigning unit 23 can distribute and assign roles as backup candidates to the plurality of nodes 30.
- step S321 the second allocation unit 23 allocates the target node 30 as a backup candidate to the target partition. Subsequently, in step S322, the second allocation unit 23 updates the load value of the target node 30 by adding “1”. Thereby, the 2nd allocation part 23 can increase the load value of the node 30 1 each time it allocates as a candidate.
- step S323 the process proceeds to step S323.
- step S323 the second assigning unit 23 returns the process to step S318 when the processes from step S318 to step S323 are not executed for all of the plurality of nodes 30.
- the second allocation unit 23 advances the process to step S324.
- step S324 the second allocation unit 23 returns the process to step S314 when the processes from step S314 to step S324 have not been executed for all of the plurality of partitions. Then, when the processes from step S314 to step S324 are executed for all of the plurality of partitions, the second allocation unit 23 ends the calculation of the target role table and exits this flow.
- backup candidates are allocated in order from the partition with the smallest redundancy (that is, the number of assigned backups is small), so that the redundancy of the database can be efficiently secured. Can do. Further, according to the second allocating unit 23, no backup candidates are further allocated to the nodes 30 in which the number of allocated backup candidates is equal to or greater than the reference value. Can be assigned.
- FIG. 15 is a diagram illustrating a processing flow of the access processing unit 33 of the node 30.
- the access processing unit 33 executes the following processes of Step S411 to Step S433.
- the access processing unit 33 detects a request reception event from a client or a request reception event by a replication process.
- a request reception event from a client occurs when an access request for a database is received from a client via a network.
- the request reception event by the replication process is generated when an update request or a registration request by the replication process is received from another owner node 30 via the network.
- step S421 When the access processing unit 33 detects a request reception event from the client, the process proceeds to step S421. If the access processing unit 33 detects a request reception event due to the replication process, the process proceeds to step S431.
- step S421 the access processing unit 33 calculates and obtains the number of the requested partition indicated in the access request from the client, for example. Subsequently, in step S422, the access processing unit 33 determines whether or not the own node is assigned to the owner for the requested partition by referring to the current role table.
- step S423 the access processing unit 33 notifies the client of the number of the node 30 assigned to the owner. Then, the process returns to step S411.
- step S424 the access processing unit 33 determines whether the access request type is a reference, update, or registration. It is judged whether it is.
- step S425 the access processing unit 33 reads the data requested for reference from the data store unit 31 and transmits it to the client, and the process is stepped. Return to S411.
- step S426 replication processing is executed with another node 30 assigned as a backup to the requested partition. To do. That is, the access processing unit 33 transmits the same access request as the update or registration request received from the client to the node 30 assigned as a backup.
- step S427 the access processing unit 33 updates the data stored in the data store unit 31 or adds a new one to the data store unit 31 in accordance with the update or registration request transmitted from the client. Register the data. Subsequently, in step S428, the access processing unit 33 updates the time stamp of the data holding table by incrementing by 1, and returns the process to step S411.
- step S431 the access processing unit 33 updates the data stored in the data store unit 31 in accordance with the update or registration request transmitted from the owner node 30. Alternatively, new data is registered in the data store unit 31. Subsequently, in step S432, the access processing unit 33 updates the data retention table by incrementing the time stamp by one. Subsequently, in step S433, the access processing unit 33 notifies the owner that registration or update has been completed, and returns the process to step S411.
- the access processing unit 33 may receive a transaction including a series of access requests from the client and execute transaction processing according to the access request from the client. Further, when transmitting an update or registration request by replication processing in step S426, the access processing unit 33 may transmit a transaction including a series of access requests to the node 30 assigned as a backup. In step S431, the access processing unit 33 may receive a transaction including a series of access requests from the owner node 30 and execute a transaction process according to the access request from the owner node 30.
- FIG. 16 is a diagram illustrating a processing flow of the node management unit 34 of the node 30.
- the node management unit 34 executes the following processing in steps S511 to S541.
- step S511 the node management unit 34 detects a third fixed-cycle event, a reception event in the next role table, or a reception event in the target role table.
- the third periodic event occurs periodically.
- the reception event of the next role table occurs when the node management unit 34 receives the next role table.
- the target role table reception event occurs when the node management unit 34 receives the target role table.
- step S521 the node management unit 34 detects the third periodic event. If the node management unit 34 detects a reception event in the next role table, the node management unit 34 advances the process to step S531. If the node management unit 34 detects a reception event in the target role table, the process proceeds to step S541.
- step S521 the node management unit 34 transmits the data holding table stored in the table storage unit 32 to the cluster management unit 24 of the management apparatus 20. Then, the node management unit 34 returns the process to step S511.
- step S531 if the access processing unit 33 is executing a transaction, the node management unit 34 interrupts the processing. Subsequently, in step S532, the node management unit 34 executes short-term synchronization processing according to the received next role table.
- the node management unit 34 accesses the plurality of partitions when there is a difference in the data set stored between the owner node 30 and the backup node 30.
- the processing unit 33 is caused to execute the same process as the replication process.
- the node management unit 34 can synchronize between the node 30 assigned as the owner and the node 30 assigned as the backup for each of the plurality of partitions.
- the node management unit 34 causes the access processing unit 33 to execute processing corresponding to the new role (owner and backup) assigned by the next role table for each of the plurality of partitions.
- step S533 the node management unit 34 rewrites the current role table with the contents of the next role table. Thereby, the access processing unit 33 can accept an access request from a client.
- step S534 the node management unit 34 interrupts and restarts the transaction. Then, the node management unit 34 returns the process to step S511.
- the node management unit 34 updates the target role table stored in the table storage unit 32 in step S541. Then, the node management unit 34 returns the process to step S511.
- FIG. 17 is a diagram illustrating a processing flow of the transfer unit 35 of the node 30.
- the transfer unit 35 executes the following processes in steps S611 to S618 and steps S621 to S622.
- step S611 the transfer unit 35 determines whether or not the load on the CPU 12 is greater than a predetermined reference value. If the load is greater than the reference value (Yes in step S611), in step S612, the transfer unit 35 sleeps for a certain period of time, and then returns the process to step S611.
- step S613 the transfer unit 35 detects a difference between the current role table and the target role table. Subsequently, in step S614, the transfer unit 35 determines whether there is a difference between the current role table and the target role table.
- step S615 When there is no difference between the current role table and the target role table (No in step S614), in step S615, the transfer unit 35 plays a role in the current node in the current role table, the next role table, and the target role table. If there is a partition to which is not assigned, delete the data set. And the transfer part 35 returns a process to step S611 after the process of step S615.
- step S616 the transfer unit 35 selects a partition whose own node is the owner in the current role table and backup candidates in the target role table. In the current role table, another node 30 that is neither the owner nor the backup is detected.
- step S617 the transfer unit 35 determines whether there is another node 30 that is a candidate for backup in the target role table and is neither an owner nor a backup in the current role table. If it does not exist (No in step S617), the transfer unit 35 returns the process to step S611.
- step S618 the transfer unit 35 starts the long-term synchronization process without interrupting the process for the access request from the client. Specifically, the transfer unit 35 assigns the partition whose own node is assigned to the owner in the current role table to the backup candidate according to the target role table and to the owner and the backup according to the current role table. The data set of the partition is transmitted to the other nodes 30 that are not. In this case, the transfer unit 35 transmits the data set in the background that does not interfere with the execution of the transaction by the access processing unit 33. Thereby, the transfer part 35 can perform a long-term synchronization process, without causing the fall of the response speed etc. with respect to the access request from a client.
- step S611 After complete
- the transfer unit 35 of the other node 30 having a partition that is assigned to the backup candidate by the target role table and that is not assigned to either the owner or the backup by the current role table is the following steps S621 and S622. Execute the process.
- step S621 the transfer unit 35 starts the long-term synchronization process without interrupting the process for the access request from the client. Specifically, the transfer unit 35 sets a data set from the node 30 assigned to the owner for a partition that is assigned to the backup candidate by the target role table and that is not assigned to either the owner or the backup by the current role table. Receive. In this case, the transfer unit 35 receives the data set in the background that does not interfere with the execution of the transaction by the access processing unit 33.
- step S622 the transfer unit 35 updates the data holding table.
- the transfer unit 35 updates the data holding table to obtain the time stamp of the partition as the owner's time stamp. Can match the time stamp of the data retention table.
- the transfer unit 35 generates a replica of the data set of the node 30 assigned as the owner to the node 30 assigned as the backup candidate. Can be generated.
- FIG. 18 is a diagram showing the state of each of the nodes 30-A to 30-C when the owner and backup node 30 are assigned to each of the three partitions # 1 to # 3.
- the node 30-A is assigned to the owner and the node 30-B is assigned to the backup for the partition # 1.
- node 30-B is assigned to the owner, and node 30-C is assigned to the backup.
- node 30-C is assigned to the owner, and node 30-A is assigned to the backup.
- Node 30-A and node 30-B are executing replication processing for partition # 1 by transaction.
- the time stamp of the partition # 1 has the same value in the node 30-A and the node 30-B.
- the node 30-B and the node 30-C are executing replication processing for the partition # 2 by transaction.
- the time stamp of the partition # 2 has the same value in the node 30-B and the node 30-C.
- the node 30-C and the node 30-A are executing replication processing for the partition # 3 by transaction.
- the time stamp of the partition # 3 has the same value in the node 30-C and the node 30-A.
- FIG. 19 is a diagram illustrating the respective states of the node 30-A to the node 30-C after executing the short-term synchronization processing in response to a failure occurring in the node 30-C from the state of FIG. It is assumed that the cluster management unit 24 of the management apparatus 20 cannot detect the data holding table from the node 30-C from the state of FIG. 18, and detects that a failure has occurred in the node 30-C.
- the cluster management unit 24 of the management device 20 separates the failed node 30 and calculates the next role table in the first allocation unit 22.
- the first assigning unit 22 reassigns the owner and the backup node 30 except for the failed node 30, and generates the next role table.
- the first allocation unit 22 allocates the node 30 so that at least an owner exists for each of the plurality of partitions so that the database functions at least. For example, when a failure occurs in the node 30 that was the owner, the first allocation unit 22 allocates the node 30 that was the backup to the owner. Furthermore, the first allocation unit 22 allocates the nodes 30 so that backups exist as much as possible for each of the plurality of partitions in order to further increase the redundancy of the database.
- the node 30-A is assigned to the owner and the node 30-B is assigned to the backup for the partition # 1. Further, the node 30-B is assigned to the owner for the partition # 2. Further, the node 30-A is assigned to the owner for the partition # 3.
- the cluster management unit 24 distributes the next role table to the nodes 30-A and 30-B.
- the node managers 34 of the nodes 30-A and 30-B execute short-term synchronization processing and rewrite the contents of the current role table with the contents of the next role table. Thereby, each of the node 30-A and the node 30-B can execute processing according to the newly assigned role.
- FIG. 20 is a diagram showing an example of long-term synchronization processing executed from the state of FIG. Further, the cluster management unit 24 of the management device 20 causes the second assignment unit 23 to calculate the target role table when the failed node 30-C is disconnected.
- the second assignment unit 23 reassigns the backup candidate nodes 30 except for the failed node 30, and generates a target role table.
- the second allocation unit 23 allocates candidate nodes 30 so that at least an owner and a backup exist for each of the plurality of partitions in order to increase the redundancy of the database.
- the second allocation unit 23 allocates the node 30-A as a backup candidate for the partition # 2, and allocates the node 30-B as a backup candidate for the partition # 3.
- the second allocation unit 23 also allocates the node 30 that has already been allocated as the owner and backup according to the current role table as a backup candidate. Therefore, as shown in the target role table of FIG. 20, both the node 30-A and the node 30-B are assigned as backup candidates to each of the partitions # 1 to # 3.
- the cluster management unit 24 distributes the target role table to the nodes 30-A and 30-B.
- Each transfer unit 35 of the node 30-A and the node 30-B performs long-term synchronization processing on the difference between the current role table and the target role table. That is, the transfer unit 35 of the node 30-B transmits the data set of the partition # 2 to the node 30-A in the background. Further, the transfer unit 35 of the node 30-A transmits the data set of the partition # 3 to the node 30-B in the background.
- the transfer unit 35 executes the long-term synchronization process without interrupting the process for the access request from the client.
- the node 30-A can store a replica of the data set of the partition # 2. Further, as shown in the data holding table of FIG. 20, the time stamp of the partition # 2 at this time has the same value in the node 30-A and the node 30-B.
- the node 30-B can store a replica of the data set of the partition # 3. Further, as shown in the data holding table of FIG. 20, the time stamp of the partition # 3 has the same value in the node 30-A and the node 30-B.
- FIG. 21 is a diagram showing the states of the nodes 30-A to 30-C after the short-term synchronization processing is executed in response to the completion of the long-term synchronization processing of FIG.
- the cluster management unit 24 of the management device 20 periodically calls the first allocation unit 22 to calculate the next role table.
- the first allocation unit 22 uses the replica generated by the long-term synchronization processing for each of the plurality of partitions to increase redundancy. In this manner, the owner and the backup node 30 are reassigned.
- the node 30-A is assigned as the owner and the node 30-B is assigned as the backup to the partition # 1.
- the node 30-B is assigned as the owner, and the node 30-A is assigned as the backup.
- the node 30-A is assigned as the owner to the partition # 3, and the node 30-B is assigned as the backup.
- the cluster management unit 24 distributes the next role table to the nodes 30-A and 30-B.
- the node managers 34 of the nodes 30-A and 30-B execute short-term synchronization processing and rewrite the contents of the current role table with the contents of the next role table. Thereby, each of the node 30-A and the node 30-B can execute processing according to the newly assigned role.
- FIG. 22 is a diagram showing each state of the node 30-A to the node 30-D when the node 30-D is added from the state of FIG. For example, it is assumed that the cluster management unit 24 of the management apparatus 20 detects that a node 30-D has been newly added from the state of FIG.
- the cluster management unit 24 generates a data holding table to which the node 30-D is added as shown in the data holding table of FIG.
- the first assignment unit 22 In response to the addition of the node 30-D to the data holding table, the first assignment unit 22 generates a next role table to which the node 30-D has been added.
- the node 30-D does not store the data set. Accordingly, the first allocation unit 22 does not change the allocation of the owner and the backup immediately after the node 30-D is added.
- FIG. 23 is a diagram showing an example of long-term synchronization processing executed from the state shown in FIG.
- the second allocation unit 23 reallocates the node 30 that is a backup candidate including the new node 30 to each of the plurality of partitions.
- the second assigning unit 23 performs backup candidates for each of the plurality of partitions so that the difference between the number of assigned owners and the number of backups is within a predetermined value range between the nodes 30. Node 30 is assigned.
- the node 30-A and the node 30-B are assigned as the backup candidates to the partition # 1. Further, the node 30-B and the node 30-D are assigned as the backup candidates to the partition # 2. Further, the node 30-A and the node 30-D are assigned as the backup candidates to the partition # 3.
- the cluster management unit 24 distributes the target role table to the nodes 30-A, 30-B, and 30-D.
- Each transfer unit 35 of the node 30-A, the node 30-B, and the node 30-D executes long-term synchronization processing on the difference portion between the current role table and the target role table.
- the transfer unit 35 of the node 30-B transmits the data set of the partition # 2 to the node 30-D in the background.
- the transfer unit 35 of the node 30-A transmits the data set of the partition # 3 to the node 30-D in the background.
- the transfer unit 35 executes the long-term synchronization process without interrupting the process for the access request from the client.
- the node 30-D can store a replica of the data set of the partition # 2. Further, as shown in the data retention table of FIG. 23, the time stamp of the partition # 2 has the same value in the node 30-B and the node 30-D.
- the node 30-D can store a replica of the data set of the partition # 3. Furthermore, as shown in the data retention table of FIG. 23, the time stamp of partition # 3 has the same value at node 30-A and node 30-D.
- FIG. 24 is a diagram showing the states of the nodes 30-A to 30-D after the short-term synchronization processing is executed in response to the completion of the long-term synchronization processing of FIG.
- the first allocation unit 22 uses a replica generated by the long-term synchronization processing for each of the plurality of partitions to further connect the nodes 30.
- the node 30 as the owner and the backup are reassigned so that the load is equalized.
- the node 30-A is assigned as the owner and the node 30-B is assigned as the backup to the partition # 1.
- the node 30-B is assigned as the owner to the partition # 2
- the node 30-D is assigned as the backup.
- the node 30-A is assigned as the owner to the partition # 3, and the node 30-D is assigned as the backup.
- the cluster management unit 24 distributes the next role table to the nodes 30-A, 30-B, and 30-D.
- each node management unit 34 of the node 30-A, node 30-B, and node 30-D executes short-term synchronization processing and rewrites the contents of the current role table with the contents of the next role table. .
- each of the node 30-A, the node 30-B, and the node 30-D can execute processing according to the newly assigned role.
- the node 30 as the backup candidate is assigned and the data set is transferred from the node 30 as the owner to the node 30 as the candidate. It is possible to provide redundancy and equalize the processing load of the node 30. Furthermore, according to the database system 10, such a process can be executed without interrupting the process for the access request from the client. Therefore, even when the node 30 is disconnected or a new node 30 is added. Data can be rearranged without stopping the system.
- the programs executed in the management apparatus 20 and the node 30 according to this embodiment are files in an installable format or an executable format, and are CD-ROM, flexible disk (FD), CD-R, DVD (Digital Versatile Disk). And the like recorded on a computer-readable recording medium.
- the program executed by the management device 20 and the node 30 of the present embodiment may be provided by being stored on a computer connected to a network such as the Internet and downloaded via the network. Further, the program executed by the management apparatus 20 and the node 30 of the present embodiment may be configured to be provided or distributed via a network such as the Internet.
- management apparatus 20 and the node 30 program of this embodiment may be configured to be provided by being incorporated in advance in a ROM or the like.
- the program executed by the management apparatus 20 of the present embodiment has a module configuration including the above-described units (the table storage unit 21, the first allocation unit 22, the second allocation unit 23, and the cluster management unit 24).
- a CPU processor
- the table storage unit 21, the first allocation unit 22, the second allocation unit 23, and A cluster management unit 24 is generated on the main storage device.
- the program executed in the node 30 of the present embodiment has a module configuration including the above-described units (data store unit 31, table storage unit 32, access processing unit 33, node management unit 34, and transfer unit 35).
- a CPU processor
- the node management unit 34 and the transfer unit 35 are generated on the main storage device.
- Database system 10
- CPU 13 Memory 14
- Communication I / F 15 Storage I / F 16 external storage device
- management device 21 table storage unit 22
- first allocation unit 23 second allocation unit 24
- cluster management unit 30 node 31 data store unit 32
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Hardware Redundancy (AREA)
Abstract
Description
従来より、シェアードナッシング型のデータベースシステムが知られている。シェアードナッシング型のデータベースシステムは、互いがネットワークで接続された複数のデータベースノードを備える。
シェアードナッシング型のデータベースシステムでは、パーティショニングと呼ばれる技術を適用できる。パーティショニングは、データベースを複数に分割し、分割したそれぞれのデータ集合(パーティションと呼ばれる。)を複数のデータベースノードに分散して記憶させる技術である。パーティショニングを適用したシェアードナッシング型のデータベースシステムは、それぞれのデータベースノードに記憶させるデータ集合を小さくできるので、クライアントからのアクセスを複数のデータベースノードに分散させることができる。
データベース技術の一つに、シャーディングと呼ばれる技術がある。シャーディングは、データをより小さなデータに分割し、複数のデータベースノードに自動的に分散して保存する技術である。シャーディングは、シェアードナッシング型のデータベースシステムのパーティショニングと類似した技術である。
また、キーバリュー型データベースが知られている。キーバリュー型データベースは、プログラミング言語の連想配列またはMapと同様にキーとバリューとのペアからなる単純なデータモデルに基づくデータベースである。また、キーバリュー型データベースをキーによりシャーディングすることで分散化した分散キーバリュー型データベースも知られている。
MLOAD={パーティションの数×(レプリケーション数+1)+(ノード数-1)}÷ノード数 …(1)
12 CPU
13 メモリ
14 通信I/F
15 ストレージI/F
16 外部記憶装置
20 管理装置
21 表記憶部
22 第1割当部
23 第2割当部
24 クラスタ管理部
30 ノード
31 データストア部
32 表記憶部
33 アクセス処理部
34 ノード管理部
35 転送部
Claims (16)
- データベースを記憶する複数のノードと、前記複数のノードを管理する管理装置と、を備えるデータベースシステムであって、
前記管理装置は、
前記複数のノードのそれぞれの状況の変更に応じて、データ集合を記憶して前記データ集合へのアクセス要求を受け付ける第1ノード、および、前記データ集合を記憶して前記第1ノードのバックアップとなる第2ノードを割り当てて、前記複数のノードのそれぞれに割り当てに応じた動作を指示する第1割当部と、
前記複数のノードの状況と、前記第1ノードおよび前記第2ノードの割り当て状況の変更に応じて、前記第1ノードまたは前記第2ノードの候補となる第3ノードを割り当てて、前記複数のノードのそれぞれに対して、前記第3ノードに前記第1ノードまたは前記第2ノードとして動作させるための準備を指示する第2割当部と、
を有し、
前記複数のノードのそれぞれは、
前記第1ノードまたは前記第2ノードのデータ集合を、前記第3ノードに送信する転送部、
を有する
データベースシステム。 - 前記複数のノードは、前記データベースを複数のパーティションに分割して記憶し、
前記第1割当部は、前記複数のパーティションのそれぞれに対して、前記第1ノードおよび前記第2ノードを割り当て、
前記第2割当部は、前記複数のパーティションのそれぞれに対して、前記第3ノードを割り当てる
請求項1に記載のデータベースシステム。 - 前記第1割当部は、複数のパーティションのそれぞれに対して、前記複数のノードのそれぞれが前記第1ノードか前記第2ノードかを示す第1テーブルを生成し、前記第1テーブルを前記複数のノードのそれぞれに配布することにより割り当てに応じた動作を指示し、
前記第2割当部は、複数のパーティションのそれぞれに対して、前記複数のノードのそれぞれが前記第3ノードかを示す第2テーブルを生成し、前記第2テーブルを前記複数のノードのそれぞれに配布することにより前記準備を指示する
請求項2に記載のデータベースシステム。 - 前記第1割当部は、前記第3ノードへのデータ集合の送信が完了した場合、前記複数のパーティションのそれぞれに対して前記第1ノードおよび前記第2ノードを再度割り当てる
請求項2に記載のデータベースシステム。 - 前記複数のノードのうちの何れかのノードで障害が発生した場合、
前記第1割当部は、前記障害が発生したノードを除いて、前記第1ノードおよび前記第2ノードを再度割り当て、
前記第2割当部は、前記障害が発生したノードを除いて、前記第3ノードを再度割り当てる
請求項4に記載のデータベースシステム。 - 前記第2割当部は、前記複数のパーティションのそれぞれに対して、少なくとも前記第1ノードおよび前記第2ノードが存在するように、前記第3ノードを割り当てる
請求項5に記載のデータベースシステム。 - 新たなノードが追加された場合、前記第2割当部は、前記複数のパーティションのそれぞれに対して、前記新たなノードを含めて前記第3ノードを再度割り当てる
請求項4に記載のデータベースシステム。 - 前記第2割当部は、割り当てられた前記第1ノードおよび前記第2ノードの数の差がノード間で予め定められた値の範囲内となるように、前記複数のパーティションのそれぞれに対して、前記第3ノードを割り当てる
請求項7に記載のデータベースシステム。 - 前記第2割当部は、演算能力の高いノードが演算能力の低いノードよりも優先的に前記第1ノードまたは前記第2ノードとなるように、前記第3ノードを割り当てる
請求項8に記載のデータベースシステム。 - 前記第2割当部は、前記アクセス要求を受け付ける数が少ないノードが前記アクセス要求を受け付ける数が多いノードよりも優先的に前記第1ノードまたは前記第2ノードとなるように、前記第3ノードを割り当てる
請求項8に記載のデータベースシステム。 - 前記複数のノードのそれぞれは、前記第1ノードのパーティションに対する一連のアクセス要求を含むトランザクションを受け付けてクライアントからのアクセス要求に応じた処理を実行し、前記パーティションのデータ集合の更新または登録のアクセス要求を実行する場合、前記第2ノードに一連のアクセス要求を含むトランザクションを送信して実行させるアクセス処理部をさらに有し、
前記転送部は、前記第1ノードまたは前記第2ノードの前記パーティションのデータ集合を、前記トランザクションの実行を妨害しないバックグラウンドで、前記第3ノードに送信する
請求項1に記載のデータベースシステム。 - データベースを記憶する複数のノードを管理する管理装置であって、
前記管理装置は、
前記複数のノードのそれぞれの状況の変更に応じて、データ集合を記憶して前記データ集合へのアクセス要求を受け付ける第1ノード、および、前記データ集合を記憶して前記第1ノードのバックアップとなる第2ノードを割り当てて、前記複数のノードのそれぞれに割り当てに応じた動作を指示する第1割当部と、
前記複数のノードの状況と、前記第1ノードおよび前記第2ノードの割り当て状況との変更に応じて、前記第1ノードまたは前記第2ノードの候補となる第3ノードを割り当てて、前記複数のノードのそれぞれに対して、前記第3ノードに前記第1ノードまたは前記第2ノードとして動作させるための準備を指示する第2割当部と、
を有し、
前記複数のノードのそれぞれは、
前記第1ノードまたは前記第2ノードのデータ集合を、前記第3ノードに送信する
管理装置。 - 請求項1に記載のデータベースシステムに備えられるノード。
- 請求項1に記載のデータベースシステムに備えられる管理装置としてコンピュータを機能させるためのプログラム。
- 請求項1に記載のデータベースシステムに備えられるノードとしてコンピュータを機能させるためのプログラム。
- データベースを記憶する複数のノードと、前記複数のノードを管理する管理装置と、を備えるデータベースシステムにおけるデータ処理方法であって、
前記管理装置が、
前記複数のノードのそれぞれの状況の変更に応じて、データ集合を記憶して前記データ集合へのアクセス要求を受け付ける第1ノード、および、前記データ集合を記憶して前記第1ノードのバックアップとなる第2ノードを割り当てて、前記複数のノードのそれぞれに割り当てに応じた動作を指示する第1割当ステップと、
前記複数のノードの状況と、前記第1ノードおよび前記第2ノードの割り当て状況の変更に応じて、前記第1ノードまたは前記第2ノードの候補となる第3ノードを割り当てて、前記複数のノードのそれぞれに対して、前記第3ノードに前記第1ノードまたは前記第2ノードとして動作させるための準備を指示する第2割当ステップと、
を実行し、
前記複数のノードのそれぞれが、
前記第1ノードまたは前記第2ノードのデータ集合を、前記第3ノードに送信する転送ステップ
を実行する
データ処理方法。
Priority Applications (8)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP13834378.5A EP2975523A4 (en) | 2013-03-12 | 2013-03-12 | Database system, program, and data processing method |
EP23152630.2A EP4191431A1 (en) | 2013-03-12 | 2013-03-12 | Database system, program, and data processing method |
JP2014504113A JP5698865B2 (ja) | 2013-03-12 | 2013-03-12 | データベースシステム、プログラムおよびデータ処理方法 |
CN201380003048.1A CN104185841B (zh) | 2013-03-12 | 2013-03-12 | 数据库系统、管理装置、节点以及数据处理方法 |
PCT/JP2013/056868 WO2014141393A1 (ja) | 2013-03-12 | 2013-03-12 | データベースシステム、プログラムおよびデータ処理方法 |
AU2013381504A AU2013381504B2 (en) | 2013-03-12 | 2013-03-12 | Database system, program, and data processing method |
US14/206,819 US20140279902A1 (en) | 2013-03-12 | 2014-03-12 | Database system, computer program product, and data processing method |
US18/167,959 US20230244694A1 (en) | 2013-03-12 | 2023-02-13 | Database system, computer program product, and data processing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2013/056868 WO2014141393A1 (ja) | 2013-03-12 | 2013-03-12 | データベースシステム、プログラムおよびデータ処理方法 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US14/206,819 Continuation US20140279902A1 (en) | 2013-03-12 | 2014-03-12 | Database system, computer program product, and data processing method |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014141393A1 true WO2014141393A1 (ja) | 2014-09-18 |
Family
ID=51532972
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/056868 WO2014141393A1 (ja) | 2013-03-12 | 2013-03-12 | データベースシステム、プログラムおよびデータ処理方法 |
Country Status (6)
Country | Link |
---|---|
US (2) | US20140279902A1 (ja) |
EP (2) | EP4191431A1 (ja) |
JP (1) | JP5698865B2 (ja) |
CN (1) | CN104185841B (ja) |
AU (1) | AU2013381504B2 (ja) |
WO (1) | WO2014141393A1 (ja) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6382819B2 (ja) | 2013-08-21 | 2018-08-29 | 株式会社東芝 | データベースシステム、ノード、管理装置、プログラムおよびデータ処理方法 |
JP6122126B2 (ja) | 2013-08-27 | 2017-04-26 | 株式会社東芝 | データベースシステム、プログラムおよびデータ処理方法 |
EP3035595A1 (en) * | 2014-12-17 | 2016-06-22 | Alcatel Lucent | Routable distributed database for managing a plurality of entities of a telecommunication network |
US10977276B2 (en) * | 2015-07-31 | 2021-04-13 | International Business Machines Corporation | Balanced partition placement in distributed databases |
US10303557B2 (en) * | 2016-03-09 | 2019-05-28 | Commvault Systems, Inc. | Data transfer to a distributed storage environment |
CN107800551A (zh) * | 2016-08-31 | 2018-03-13 | 北京优朋普乐科技有限公司 | Redis集群系统及其提高可靠性的方法、客户端 |
CN110300188B (zh) * | 2019-07-25 | 2022-03-22 | 中国工商银行股份有限公司 | 数据传输系统、方法和设备 |
CN111309805B (zh) * | 2019-12-13 | 2023-10-20 | 华为技术有限公司 | 数据库的数据读写方法及装置 |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0392942A (ja) * | 1989-09-06 | 1991-04-18 | Hitachi Ltd | ファイルの格納方法およびアクセス方法 |
JPH09146812A (ja) * | 1995-11-27 | 1997-06-06 | Sanyo Electric Co Ltd | データベース装置 |
JPH11506556A (ja) * | 1995-05-26 | 1999-06-08 | クラストラ アクティーゼルスカブ | データベースフラグメントレプリカの共通部分が最小の複数のノードのグループを有する連続的に使用可能なデータベースサーバ |
JP2001508900A (ja) * | 1997-01-20 | 2001-07-03 | テレフオンアクチーボラゲツト エル エム エリクソン(パブル) | 分散データ処理システムにおけるデータの分配および複写 |
JP2002522845A (ja) * | 1998-08-11 | 2002-07-23 | テレフオンアクチーボラゲツト エル エム エリクソン(パブル) | フォールトトレラント・コンピュータシステム |
JP2003345640A (ja) * | 2002-05-28 | 2003-12-05 | Mitsubishi Electric Corp | データベースバックアップシステム |
JP2005196602A (ja) | 2004-01-09 | 2005-07-21 | Hitachi Ltd | 無共有型データベース管理システムにおけるシステム構成変更方法 |
JP2009157785A (ja) | 2007-12-27 | 2009-07-16 | Hitachi Ltd | 待機系計算機の追加方法、計算機及び計算機システム |
Family Cites Families (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5884326A (en) * | 1996-11-18 | 1999-03-16 | Weinger; Ralph | Method and apparatus for data access |
CA2159269C (en) * | 1995-09-27 | 2000-11-21 | Chaitanya K. Baru | Method and apparatus for achieving uniform data distribution in a parallel database system |
JPH10247181A (ja) * | 1997-03-05 | 1998-09-14 | Mitsubishi Electric Corp | 拠点バックアップ計算機システム |
US7013316B1 (en) * | 2000-07-13 | 2006-03-14 | Microsoft Corporation | System and method for synchronizing multiple database files |
US7809384B2 (en) * | 2002-11-05 | 2010-10-05 | Microsoft Corporation | User-input scheduling of synchronization operation on a mobile device based on user activity |
CN1317658C (zh) * | 2002-12-31 | 2007-05-23 | 联想(北京)有限公司 | 利用机群节点相互备份的容错方法 |
JP4374953B2 (ja) * | 2003-09-09 | 2009-12-02 | 株式会社日立製作所 | データ処理システム |
US20050144316A1 (en) * | 2003-12-06 | 2005-06-30 | William Loo | Method and system for service node redundancy |
US7120769B2 (en) * | 2004-03-08 | 2006-10-10 | Hitachi, Ltd. | Point in time remote copy for multiple sites |
US8266406B2 (en) * | 2004-04-30 | 2012-09-11 | Commvault Systems, Inc. | System and method for allocation of organizational resources |
US7490205B2 (en) * | 2005-03-14 | 2009-02-10 | International Business Machines Corporation | Method for providing a triad copy of storage data |
JP4107676B2 (ja) * | 2006-07-21 | 2008-06-25 | インターナショナル・ビジネス・マシーンズ・コーポレーション | トランザクション引継ぎシステム |
US7725764B2 (en) * | 2006-08-04 | 2010-05-25 | Tsx Inc. | Failover system and method |
US8849746B2 (en) * | 2006-12-19 | 2014-09-30 | Teradata Us, Inc. | High-throughput extract-transform-load (ETL) of program events for subsequent analysis |
US20080288630A1 (en) * | 2007-05-18 | 2008-11-20 | Motorola, Inc. | Device management |
JP4386932B2 (ja) * | 2007-08-17 | 2009-12-16 | 富士通株式会社 | ストレージ管理プログラム、ストレージ管理装置およびストレージ管理方法 |
JP5222617B2 (ja) * | 2008-04-28 | 2013-06-26 | 株式会社日立製作所 | 情報システム及びi/o処理方法 |
JP4612715B2 (ja) * | 2008-09-05 | 2011-01-12 | 株式会社日立製作所 | 情報処理システム、データ更新方法およびデータ更新プログラム |
JP5359201B2 (ja) * | 2008-11-06 | 2013-12-04 | 富士通株式会社 | コンテンツの削除更新プログラム |
JP5381336B2 (ja) * | 2009-05-28 | 2014-01-08 | 富士通株式会社 | 管理プログラム、管理装置および管理方法 |
US8724449B2 (en) * | 2009-06-10 | 2014-05-13 | Cisco Technology, Inc. | Failure protection for access ring topology |
US9325802B2 (en) * | 2009-07-16 | 2016-04-26 | Microsoft Technology Licensing, Llc | Hierarchical scale unit values for storing instances of data among nodes of a distributed store |
US8495044B2 (en) * | 2009-09-02 | 2013-07-23 | Microsoft Corporation | File system node updates |
US20110283277A1 (en) * | 2010-05-11 | 2011-11-17 | International Business Machines Corporation | Virtualization and dynamic resource allocation aware storage level reordering |
US8595184B2 (en) * | 2010-05-19 | 2013-11-26 | Microsoft Corporation | Scaleable fault-tolerant metadata service |
US8515915B2 (en) * | 2010-09-24 | 2013-08-20 | Hitachi Data Systems Corporation | System and method for enhancing availability of a distributed object storage system during a partial database outage |
US8627024B2 (en) * | 2010-11-30 | 2014-01-07 | International Business Machines Corporation | Snapshot based replication |
US8572031B2 (en) * | 2010-12-23 | 2013-10-29 | Mongodb, Inc. | Method and apparatus for maintaining replica sets |
US8554730B2 (en) * | 2011-07-29 | 2013-10-08 | International Business Machines Corporation | Adding a key column to a table to be replicated |
US20130149678A1 (en) * | 2011-12-12 | 2013-06-13 | Yukie J. Tokuda | System and methods for virtual cooking with multi-course planning |
US8880565B2 (en) * | 2011-12-23 | 2014-11-04 | Sap Se | Table creation for partitioned tables |
-
2013
- 2013-03-12 CN CN201380003048.1A patent/CN104185841B/zh active Active
- 2013-03-12 WO PCT/JP2013/056868 patent/WO2014141393A1/ja active Application Filing
- 2013-03-12 EP EP23152630.2A patent/EP4191431A1/en active Pending
- 2013-03-12 EP EP13834378.5A patent/EP2975523A4/en not_active Ceased
- 2013-03-12 JP JP2014504113A patent/JP5698865B2/ja active Active
- 2013-03-12 AU AU2013381504A patent/AU2013381504B2/en active Active
-
2014
- 2014-03-12 US US14/206,819 patent/US20140279902A1/en not_active Abandoned
-
2023
- 2023-02-13 US US18/167,959 patent/US20230244694A1/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0392942A (ja) * | 1989-09-06 | 1991-04-18 | Hitachi Ltd | ファイルの格納方法およびアクセス方法 |
JPH11506556A (ja) * | 1995-05-26 | 1999-06-08 | クラストラ アクティーゼルスカブ | データベースフラグメントレプリカの共通部分が最小の複数のノードのグループを有する連続的に使用可能なデータベースサーバ |
JPH09146812A (ja) * | 1995-11-27 | 1997-06-06 | Sanyo Electric Co Ltd | データベース装置 |
JP2001508900A (ja) * | 1997-01-20 | 2001-07-03 | テレフオンアクチーボラゲツト エル エム エリクソン(パブル) | 分散データ処理システムにおけるデータの分配および複写 |
JP2002522845A (ja) * | 1998-08-11 | 2002-07-23 | テレフオンアクチーボラゲツト エル エム エリクソン(パブル) | フォールトトレラント・コンピュータシステム |
JP2003345640A (ja) * | 2002-05-28 | 2003-12-05 | Mitsubishi Electric Corp | データベースバックアップシステム |
JP2005196602A (ja) | 2004-01-09 | 2005-07-21 | Hitachi Ltd | 無共有型データベース管理システムにおけるシステム構成変更方法 |
JP2009157785A (ja) | 2007-12-27 | 2009-07-16 | Hitachi Ltd | 待機系計算機の追加方法、計算機及び計算機システム |
Non-Patent Citations (2)
Title |
---|
MONDODB MANUAL CONTENTS, Retrieved from the Internet <URL:http://docs.mongodb.org/manual/contents> |
See also references of EP2975523A4 |
Also Published As
Publication number | Publication date |
---|---|
JP5698865B2 (ja) | 2015-04-08 |
CN104185841B (zh) | 2018-08-17 |
JPWO2014141393A1 (ja) | 2017-02-16 |
AU2013381504B2 (en) | 2016-06-23 |
EP2975523A4 (en) | 2017-02-08 |
CN104185841A (zh) | 2014-12-03 |
EP4191431A1 (en) | 2023-06-07 |
AU2013381504A1 (en) | 2015-02-12 |
US20140279902A1 (en) | 2014-09-18 |
EP2975523A1 (en) | 2016-01-20 |
US20230244694A1 (en) | 2023-08-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6382819B2 (ja) | データベースシステム、ノード、管理装置、プログラムおよびデータ処理方法 | |
JP6122126B2 (ja) | データベースシステム、プログラムおよびデータ処理方法 | |
JP5698865B2 (ja) | データベースシステム、プログラムおよびデータ処理方法 | |
CN107005596B (zh) | 用于在集群重新配置后的工作负载平衡的复制型数据库分配 | |
EP3764244B1 (en) | System and method for massively parallel processing database | |
KR101544480B1 (ko) | 복수 개의 프락시 서버를 포함하는 분산 저장 시스템 및 그 오브젝트 관리 방법 및 컴퓨터에 의하여 독출가능한 저장 매체 | |
EP2643771B1 (en) | Real time database system | |
US20150169718A1 (en) | System and method for supporting persistence partition discovery in a distributed data grid | |
JP2014232483A (ja) | データベースシステム、検索方法およびプログラム | |
US10437797B1 (en) | In-memory distributed database with a remote data store | |
US11321283B2 (en) | Table and index communications channels | |
JP5969315B2 (ja) | データ移行処理システムおよびデータ移行処理方法 | |
Malkhi et al. | From paxos to corfu: a flash-speed shared log | |
US11182362B2 (en) | Calculating device, data base system, calculation system, calculation method, and storage medium | |
KR101681651B1 (ko) | 데이터베이스 관리 시스템 및 방법 | |
CN114416116A (zh) | 一种软件单元化部署的多活方法和系统 | |
CN114791850A (zh) | 任务处理方法及其装置、计算机可读存储介质 | |
Yang et al. | Pyxis+: A Scalable and Adaptive Data Replication Framework |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2014504113 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2013834378 Country of ref document: EP |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13834378 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2013381504 Country of ref document: AU Date of ref document: 20130312 Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |