US20140279902A1 - Database system, computer program product, and data processing method - Google Patents
Database system, computer program product, and data processing method Download PDFInfo
- Publication number
- US20140279902A1 US20140279902A1 US14/206,819 US201414206819A US2014279902A1 US 20140279902 A1 US20140279902 A1 US 20140279902A1 US 201414206819 A US201414206819 A US 201414206819A US 2014279902 A1 US2014279902 A1 US 2014279902A1
- Authority
- US
- United States
- Prior art keywords
- node
- nodes
- serve
- assigned
- assigning unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G06F17/30584—
-
- 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
Abstract
Description
- This application is a continuation of PCT international application Ser. No. PCT/JP2013/056868 filed on Mar. 12, 2013 which designates the United States, the entire contents of which are incorporated herein by reference.
- Embodiments described herein relate generally to a database system, a computer program product, and a data processing method.
- Shared-Nothing Type Database
- Conventionally, a shared-nothing type database system is known in which a plurality of database nodes is connected to each other by a network.
- Each of a plurality of database nodes is a computer that has individual resources such as a processor, a memory, and a storage. In a shared-nothing type database system, total separation of resources is achieved for each database node. Hence, there occurs no access competition among the database nodes. As a result, in a shared-nothing type database system, if the number of database nodes is increased, the performance can be enhanced in a linear manner. That is, in a shared-nothing type database system, it is possible to efficiently implement the scale-out capability in which database nodes are added.
- Partitioning
- In a shared-nothing type database system, it is possible to implement a technology called partitioning in which a database is divided into a plurality of data pieces (called partitions), and each of the divided data pieces is stored in a plurality of database nodes in a distributed manner. In a shared-nothing type database system in which partitioning is implemented, it becomes possible to store smaller data pieces in each database node. As a result, an access from a client can be distributed among a plurality of database nodes.
- Sharding
- There is one database technology called sharding. In sharding, data is divided into smaller pieces of data which are automatically distributed and stored in a plurality of database nodes. Thus, sharding is a similar technology to the partitioning technology implemented in a shared-nothing type database system.
- Distributed Key-Value Type Database
- Key-value type databases are known. A key-value type database is a database that is based on a simple data model made of pairs of keys and values in an identical manner to the associative arrays or Map of a programming language. Moreover, a distributed key-value type database is also known in which a key-value type database is distributed by performing sharding with the use of keys.
- In a key-value type database, since the data model is simple in nature, it is easy to perform sharding and partitioning operations. Moreover, in a distributed key-value type database, a plurality of key-value type database nodes can be used to process large-scale data pieces which cannot be handled in a single key-value type database node. Hence, if a distributed key-value type database is implemented in a shared-nothing type database containing a plurality of database nodes, then it becomes possible to implement a large-scale database system having the scale-out capability.
- Moreover, in a shared-nothing type database system in which a distributed key-value type database is implemented (hereinafter, called a scale-out type database system); if the data is placed in an equal manner in a plurality of database nodes, then an access from a client can be distributed so as to enhance the throughput and the response on the whole.
- Meanwhile, in a scale-out type database system, a technology called replication is implemented in which a copy (replica) of the data is stored in other database nodes. In a scale-out type database system, as a result of performing replication, even if a failure occurs in a particular database node, the services can be continued using a database node in which the replica has been stored. The number of database nodes in which the replica of the data is stored is called redundancy. In a scale-out type database system, the redundancy is set to, for example, 2 or 3.
- Herein, in a scale-out type database system, if a failure occurs in a database node, and if the database node in which a failure has occurred is separated off; then the redundancy decreases from the value which has been set in advance. However, if the scale-out type database system is kept running with a decreased redundancy for a long period of time, then any further database node failure is highly likely to cause the entire system to stop working or cause a wide range of data loss. For that reason, in a scale-out type database system, it is necessary to have a mechanism by which, after a failure occurs in a database node, a new database node is added and a data replica is stored in the added database node so as to restore the redundancy to the original.
- Moreover, in a scale-out type database system, depending on the manner in which data is placed; there are times when the volume of stored data is not equal among the database nodes. Moreover, in a scale-out type database system, depending on the manner in which data is placed; there are times when the traffic from clients is not equal among the database nodes.
- If such inequality in the placement of data grows large; then, in a scale-out type database system, the database nodes having a large volume of data or the database nodes that are accessed from clients take a relatively long period of time to perform operations. Since one of the merits of a scale-out type database system is to enable distribution of an access from a client, inequality in the placement of data results in losing that merit.
- With the aim of solving the abovementioned issues of a decrease in the redundancy and inequality in the placement of data; in a scale-out type database system, at the point of time when inequality occurs in the placement of data, it is necessary to perform an operation of relocating the data among a plurality of database nodes.
- However, in a conventional database system, such an operation of relocation needs to be performed with the services stopped. Moreover, in a conventional database system, the operation of relocating the data has an extremely high processing load. Hence, unless that operation is performed in an efficient manner, the processing efficiency undergoes a decline in a significant way.
-
FIG. 1 is a diagram illustrating a hardware configuration of a database system according to an 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 a data retention table; -
FIG. 4 is a diagram illustrating an example of a current-role table; -
FIG. 5 is a diagram illustrating an example of a next-role table; -
FIG. 6 is a diagram illustrating an example of a target-role table; -
FIG. 7 is a diagram illustrating an example of data pieces stored in a data storing unit; -
FIG. 8 is a diagram for explaining a replication application; -
FIG. 9 is a diagram for explaining short-term synchronization processing; -
FIG. 10 is a diagram for explaining long-term synchronization processing; -
FIG. 11 is a diagram illustrating a transition in the roles assigned to nodes; -
FIG. 12 is a diagram illustrating an operation flowchart of a cluster managing unit of a management device; -
FIG. 13 is a flowchart for explaining the operations by which a first assigning unit of the management device calculates the next-role table; -
FIG. 14 is a flowchart for explaining the operations by which a second assigning unit of the management device calculates the target-role table; -
FIG. 15 is a diagram illustrating an operation flowchart of an access processing unit of the node; -
FIG. 16 is a diagram for explaining an operation flowchart of a node managing unit of the node; -
FIG. 17 is a diagram illustrating an operation flowchart of a transferring unit of the node; -
FIG. 18 is a diagram illustrating the states of a node 30-A to a node 30-C in a case in which each of threepartitions # 1 to #3 is assigned with the nodes serving as the owner node and the backup node; -
FIG. 19 is a diagram illustrating the states of the node 30-A to the node 30-C after short-term synchronization processing is performed in response to a failure occurring in the node 30-C in the state illustrated inFIG. 18 ; -
FIG. 20 is a diagram illustrating an example of long-term synchronization processing performed in the state illustrated inFIG. 19 ; -
FIG. 21 is a diagram illustrating the states of the node 30-A to the node 30-C after short-term synchronization processing is performed upon completion of the long-term synchronization processing explained with reference toFIG. 20 ; -
FIG. 22 is a diagram illustrating the states of the node 30-A to a node 30-D in a case in which the node 30-D is added to the state illustrated inFIG. 21 ; -
FIG. 23 is a diagram illustrating an example of long-term synchronization processing performed in the state illustrated inFIG. 22 ; and -
FIG. 24 is a diagram illustrating the states of the node 30-A to the node 30-D after short-term synchronization processing is performed upon completion of the long-term synchronization processing explained with reference toFIG. 23 . - According to an embodiment, a database system includes a plurality of nodes in which a database is stored; and a management device configured to manage the plurality of nodes. The management device includes a first assigning unit and a second assigning unit. Depending on a change in state of each of the plurality of nodes, the first assigning unit assigns a first node, which stores therein data pieces and receives an access request with respect to the data pieces, and assign a second node, which stores therein the data pieces and serves as a backup node for the first node, and instructs each of the plurality of nodes to perform operations according to assignment. Depending on a state of the plurality of nodes and depending on a change in assignment state of the first node and the second node, the second assigning unit assigns a third node which is a candidate node to serve as the first node or the second node, and instructs each of the plurality of nodes to make preparation for causing the third node to operate as the first node or the second node. Each of the plurality of nodes includes a transferring unit configured to send the data pieces of the first node or the second node to the third node.
-
FIG. 1 is a diagram illustrating a hardware configuration of adatabase system 10 according to an embodiment. Thedatabase system 10 receives a database access request (such as a reference request, an update request, or a registration request) from a client that is an external computer via a network, and performs operations according to the received access request. - The
database system 10 includes amanagement device 20 and a plurality ofnodes 30. Themanagement device 20 and a plurality ofnodes 30 are connected to each other via a network. InFIG. 1 , although only twonodes 30 are illustrated, thedatabase system 10 can also include three ormore nodes 30. - As an example, the
database system 10 is a database system in which a distributed key-value type database is implemented in a shared-nothing type database and which has the scale-out capability. In thisdatabase system 10, although the data pieces are distributed to a plurality ofnodes 30, thedatabase system 10 is accessed as a single massive database by a client that is an external computer. - The
management device 20 as well as each of a plurality ofnodes 30 is a computer that is independent from each other as far as the hardware is concerned. Each of a plurality ofnodes 30 independently receives an access request from a client, and is capable of independently performing operations according to the access request. - As an example, the
management device 20 as well as each of a plurality ofnodes 30 includes a CPU (Central Processing Unit) 12, amemory 13, a communication I/F 14, and a storage I/F 15. Moreover, themanagement device 20 as well as each of a plurality ofnodes 30 is connected to anexternal memory device 16 via the corresponding storage I/F 15. Alternatively, theexternal memory device 16 can also be installed inside thenodes 30. - In the
database system 10, a database is stored so as to be partitioned into a plurality of partitions. The number of partitions is set in advance. Regarding the method of partitioning, any method can be implemented as long as it is determined in advance. - With respect to each of a plurality of partitions, each of a plurality of
nodes 30 is assigned either to serve as an owner node, or to serve as a backup node, or neither to serve as an owner node nor to serve as a backup node. - An owner node stores therein the data pieces present in the corresponding partitions; and receives access requests from clients with respect to the data pieces of the corresponding partitions and processes the access requests. A backup node stores therein the data pieces present in the corresponding partitions; and, for example, if a failure occurs in the owner node, backs the owner node up by taking over the role of the owner node. Meanwhile, instead of the terminology such as owner nodes and backup nodes, there are times when the terminology such as master nodes and slave nodes is also used.
- The
management device 20 manages a plurality ofnodes 30. As an example, with respect to each of a plurality of partitions, themanagement device 20 assigns thenode 30 that would serve as the owner node and assigns thenode 30 that would serve as the backup node. - In this case, regarding the
nodes 30 assigned to serve as the backup nodes, themanagement device 20 does the assignment in such a way that the redundancy is maintained within a predetermined range. With that, even if a failure occurs in anynode 30, the database can be restored to the original to the extent possible. Moreover, regarding thenodes 30 assigned to serve as the owner nodes and thenodes 30 assigned to serve as the backup nodes, themanagement device 20 does the assignment in such a way that the data pieces included in the database are placed in a plurality ofnodes 30 in a distributed manner. - Meanwhile, the
management device 20 can also be configured in any one of thenodes 30. Alternatively, thenode 30 that would function as themanagement device 20 may be selected according to an arbitrary algorithm. In case a failure occurs in thenode 30 functioning as themanagement device 20, anothernode 30 may function as themanagement device 20. - In the
database system 10, if a failure occurs in any one of a plurality ofnodes 30, it becomes possible to separate off thenode 30 in which a failure has occurred. Besides, in thedatabase system 10, it is possible to newly add thenodes 30 and enhance the database performance. -
FIG. 2 is a diagram illustrating a functional configuration of thedatabase system 10 according to the embodiment. - The
management device 20 includes atable memory unit 21, a first assigningunit 22, a second assigning unit 23, and acluster managing unit 24. The first assigningunit 22, the second assigning unit 23, and thecluster managing unit 24 are implemented when theCPU 12 of themanagement device 20 runs programs. Alternatively, the first assigningunit 22, the second assigning unit 23, and thecluster managing unit 24 can be implemented either partially or entirely using hardware circuitry. Thetable memory unit 21 is implemented using thememory 13 of themanagement device 20 or using theexternal memory device 16. - The
table memory unit 21 is used to store four tables that are created for the purpose of deciding the role of each of a plurality ofnodes 30 with respect to each of a plurality of partitions. Each of the four tables can be data in the table form or can be data in a form other than the table form. - More particularly, the
table memory unit 21 is used to store a data retention table, a current-role table, a next-role table, and a target-role table. - The data retention table is used to store the time stamp of each of a plurality of
nodes 30 with respect to each of a plurality of partitions. A time stamp represents the update history about the data pieces in corresponding partitions stored in a correspondingnode 30. As an example, a time stamp is a value that is incremented after every instance of updating the corresponding data pieces. Thus, thenode 30 having the largest time stamp with respect to a particular partition represents thenode 30 that stores therein the latest data pieces with respect to the particular partition. - Meanwhile, in the case when the size of the database increases with time; then, instead of storing the time stamps, the data retention table may store the size or the data count about the data pieces in the corresponding partitions stored in the corresponding
node 30. - As an example, as illustrated in
FIG. 3 , in the data retention table, thenodes 30 are identified by rows, while the partitions are identified by columns. In this case, in the data retention table, in each cell having an intersection between a row and a column is specified the time stamp which is stored in thenode 30 identified by that row and which is of the data pieces of the partition identified by that column. - The current-role table stores therein the assigned role in the database for each of a plurality of
nodes 30 with respect to each of a plurality of partitions. More particularly, the current-role table specifies, with respect to each of a plurality of partitions, whether each of a plurality ofnodes 30 is assigned either to serve as an owner node, or to serve as a backup node, or neither to serve as an owner node nor to serve as a backup node. - As an example, as illustrated in
FIG. 4 , in the current-role table, thenodes 30 are identified by rows, while the partitions are identified by columns. In this case, in the current-role table, in each cell having an intersection between a row and a column is specified the role assigned to thenode 30 identified by that row and assigned with respect to the partition identified by that column. In the drawings, the case in which a node is assigned to serve as an owner node is illustrated as “OWNER” or “O”; the case in which a node is assigned to serve as a backup node is illustrated as “BACKUP” or “B”; and the case in which a node is neither assigned to serve as an owner node nor assigned to serve as a backup node is illustrated as “NONE” or “N” or a blank space. - The next-role table stores therein the next role assigned to each of a plurality of
nodes 30 with respect to each of a plurality of partitions. Thus, the next-role table specifies, with respect to each of a plurality of partitions, whether each of a plurality ofnodes 30 is next assigned either to serve as an owner node, or to serve as a backup node, or neither to serve as an owner node nor to serve as a backup node. For example, when a failure occurs in anode 30 or when anew node 30 is added, then the current-role table is replaced with the next-role table. - As an example, as illustrated in
FIG. 5 , in the next-role table, thenodes 30 are identified by rows, while the partitions are identified by columns. In this case, the next-role table stores, in each cell having an intersection between a row and a column, the next role assigned to thenode 30 identified by the row with respect to the partition identified by that column. - The target-role table stores therein an assigned role of a backup candidate node assigned to each of a plurality of
nodes 30 with respect to each of a plurality of partitions. A backup candidate has a role to make preparations for serving as an owner node or a backup node in future with respect to a partition. In the explanation of the present embodiment, although this particular role is named as “backup candidate”, the node assigned to serve as a backup candidate node can also serve as an owner node in future. - As an example, as illustrated in
FIG. 6 , in the target-role table, thenodes 30 are identified by rows, while the partitions are identified by columns. In this case, the target-role table stores, in each cell having an intersection between a row and a column, the role of a backup candidate node assigned to thenode 30 identified by the row and assigned with respect to the partition identified by the column. In the drawings, the case in which a node is assigned to serve as a backup candidate node is illustrated as “MID-BACKUP” or “M”; and the case in which a node is not assigned to serve as a backup candidate node is illustrated as a blank space. - The first assigning
unit 22 assigns, with respect to each of a plurality of partitions, thenode 30 that would serve as the owner node and thenode 30 that would serve as the backup node depending on the state of each of a plurality ofnodes 30. Moreover, for example, if a failure occurs in any onenode 30 of a plurality ofnodes 30; then, with the exclusion of thenode 30 in which a failure has occurred, the first assigningunit 22 reassigns thenodes 30 that would serve as the owner nodes and thenodes 30 that would serve as the backup nodes. - Furthermore, if a
new node 30 is added; then, while including the newly addednode 30, the first assigningunit 22 reassigns, with respect to each of a plurality of partitions, thenode 30 that would serve as the owner node and thenode 30 that would serve as the backup node. Moreover, in the case in which, due to the completion of sending the data pieces to anode 30 that is assigned to serve as a backup candidate node, thenode 30 becomes newly available to serve as an owner node or a backup node; the first assigningunit 22 reassigns, with respect to each of a plurality of partitions, thenode 30 that would serve as the owner node and thenode 30 that would serve as the backup node. - Herein, regarding the
nodes 30 assigned to serve as the owner nodes and thenodes 30 assigned to serve as the backup nodes, the first assigningunit 22 does the assignment with respect to all of a plurality of partitions in such a way that at least the owner nodes are present. With that, the first assigningunit 22 can at least make the database work. - Moreover, subject to at least making the database work, the first assigning
unit 22 assigns, with respect to each of a plurality of partitions, thenode 30 that would serve as the owner node and thenode 30 that would serve as the backup node in such a way that the redundancy within a predetermined range is achieved. With that, even if a failure occurs in any one of thenodes 30, the first assigningunit 22 can increase the likelihood of at least making the database work. - Moreover, subject to at least making the database work as well as subject to achieving the redundancy within a predetermined range, the first assigning
unit 22 assigns, with respect to each of a plurality of partitions, thenode 30 that would serve as the owner node and thenode 30 that would serve as the backup node in such a way that the owner nodes and the backup nodes are assigned in a distributed manner among a plurality ofnodes 30. With that, the first assigningunit 22 can even out the processing load of each of a plurality ofnodes 30. - Meanwhile, in the present example, in order to assign the
nodes 30 that would serve as the owner nodes and thenodes 30 that would serve as the backup nodes, the first assigningunit 22 calculates the next-role table. Regarding an example of the method by which the first assigningunit 22 calculates the next-role table, the explanation is given later with reference toFIG. 13 . - The second assigning unit 23 assigns, with respect to each of a plurality of partitions, the
node 30 that would serve as the backup candidate node depending on the state of a plurality ofnodes 30 and depending on the changes in the assignment state of the owner nodes and the backup nodes assigned according to the current-role table. Moreover, for example, if a failure occurs in any onenode 30 of a plurality ofnodes 30; then, with the exclusion of thenode 30 in which a failure has occurred, the second assigning unit 23 reassigns thenodes 30 that would serve as the backup candidate nodes. Furthermore, if anew node 30 is added; then, while including the newly addednode 30, the second assigning unit 23 reassigns, with respect to each of a plurality of partitions, thenodes 30 that would serve as the backup candidate nodes. - Herein, with respect to each of a plurality of partitions, the second assigning unit 23 assigns the
node 30 that would serve as the backup candidate node in such a way that the redundancy within a predetermined range is achieved in future. With that, even if a failure occurs in any onenode 30 in future, the second assigning unit 23 can increase the likelihood of at least making the database work. - Moreover, with respect to each of a plurality of partitions, the second assigning unit 23 assigns the
node 30 that would serve as the backup candidate node in such a way that future assignment of the owner nodes and the backup nodes is done in a distributed manner among a plurality ofnodes 30. With that, the second assigning unit 23 can even out the processing load of each of a plurality ofnodes 30 in future. - the
nodes 30 that would serve as the backup candidate nodes, the second assigning unit 23 calculates the target-role table. Regarding an example of the method by which the second assigning unit 23 calculates the target-role table, the explanation is given later with reference toFIG. 14 . - The
cluster managing unit 24 communicates messages with each of a plurality ofnodes 30 via a network, and manages each of a plurality ofnodes 30. For example, at regular time intervals, thecluster managing unit 24 communicates a message called heartbeat to each of a plurality ofnodes 30. Then, depending on whether or not a response to the heartbeat is received, thecluster managing unit 24 identifies thenode 30 in which a failure has occurred. - Moreover, at regular intervals, the
cluster managing unit 24 receives the data retention table from each of a plurality ofnodes 30. Then, thecluster managing unit 24 stores, in thetable memory unit 21, the data retention tables received from all of thenodes 30. Furthermore, depending on whether or not the data retention table is received, thecluster managing unit 24 identifies thenode 30 in which a failure has occurred. - Moreover, at the time of startup, the
cluster managing unit 24 causes the first assigningunit 22 to calculate the next-role table and distributes the calculated next-role table to each of a plurality ofnodes 30, to thereby instruct each of a plurality ofnodes 30 to perform operations according to the assignment. Thecluster managing unit 24 causes the first assigningunit 22 on a periodic basis to calculate the next-role table. In the case when the calculated next-role table changes from the current-role table, thecluster managing unit 24 distributes the calculated next-role table to each of a plurality ofnodes 30 to thereby instruct each of a plurality ofnodes 30 to perform operations according to the assignment. Once the next-role table is distributed; thecluster managing unit 24 updates the current-role table, which is stored in thetable memory unit 21, with the contents of the next-role table. - For example, if a failure occurs in any one
node 30, or if anew node 30 is added, or if sending of the data pieces to anode 30 that is assigned to serve as a backup candidate node is completed and thatnode 30 becomes newly available for assignment as a backup node; then the calculated next-role table changes from the current-role table. Accordingly, in such a case, thecluster managing unit 24 distributes the next-role table to each of a plurality ofnodes 30. - Meanwhile, the
cluster managing unit 24 causes the second assigning unit 23 on a periodic basis to calculate the target-role table and distributes the calculated target-role table to each of a plurality ofnodes 30. By distributing the target-role table, thecluster managing unit 24 instructs each of thenodes 30 to make preparations for causing thenode 30 assigned to serve as the backup candidate node to operate as the owner nodes or the backup nodes. Meanwhile, thecluster managing unit 24 instructs calculation of the next-role table and calculation of the target-role table at different cycles. - Each of a plurality of
nodes 30 includes adata storing unit 31, atable memory unit 32, anaccess processing unit 33, anode managing unit 34, and a transferringunit 35. Theaccess processing unit 33, thenode managing unit 34, and the transferringunit 35 are implemented when theCPU 12 of the correspondingnode 30 runs programs. Alternatively, theaccess processing unit 33, thenode managing unit 34, and the transferringunit 35 can be implemented either partially or entirely using hardware circuitry. Thedata storing unit 31 and thetable memory unit 32 are implemented using thememory 13 of the correspondingnode 30 or theexternal memory device 16. - The
data storing unit 31 is used to store the data pieces of the partitions, from among a plurality of partitions obtained by partitioning the database, with respect to which the node is assigned to serve as the owner node or the backup node. For example, as illustrated inFIG. 7 , from among threepartitions # 1 to #3 obtained by partitioning the database, assume that the concerned node is assigned to serve as the owner node with respect to thepartition # 1 and assigned to serve as the backup node with respect to thepartition # 3. In this case, the correspondingdata storing unit 31 stores the data pieces of thepartition # 1 and the data pieces of thepartition # 3. - Regarding a partition for which a
node 30 is assigned to serve as the backup candidate node according to the target-role table, thatnode 30 receives the data pieces from anothernode 30 which is assigned to serve as the owner node with respect to the partition. Regarding the data pieces present in a partition for which the concerned node is assigned to serve as the backup candidate node according to the target-role table, the correspondingdata storing unit 31 stores therein some or all of the data that is already received from thenode 30 serving as the owner node with respect to the partition. - The
table memory unit 32 is used to store the portion in the data retention table which corresponds to the corresponding node. Besides, thetable memory unit 32 is used to store the current-role table, the next-role table, and the target-role table. Regarding the current-role table; when the next-role table is received, the current-role table is replaced with the next-role table. Regarding the target-role table; when there is a change in the target-role table received on a periodic basis, it is rewritten with the changed contents. - The
access processing unit 33 receives an access request from a client via a network. Then, with respect to each of a plurality of partitions, theaccess processing unit 33 performs operations according to the role assigned to the corresponding node in the current-role table. - More particularly, the
access processing unit 33 receives from a client an access request with respect to the partition for which the corresponding node is assigned to serve as the owner node; and performs operations according to that access request. As an example, when a reference request is received, theaccess processing unit 33 reads the corresponding data from the data pieces in the partition, and sends the read data to the client. Moreover, as an example, when an update request is received, theaccess processing unit 33 updates the corresponding data in the data pieces in the partition. Furthermore, as an example, when a registration request is received, theaccess processing unit 33 registers new data in the data pieces in the partition. - Meanwhile, the
access processing unit 33 can also receive a transaction that contains a sequence of access requests. In that case, theaccess processing unit 33 performs transaction processing to process the sequence of access requests received from a client. - Moreover, when an update request or a registration request is processed, the
access processing unit 33 performs a replication operation with thenode 30 that is assigned in the current-role table to serve as the backup node with respect to the concerned partition. A replication operation points to an operation of generating, in thenode 30 that is assigned to serve as the backup node, a replica of the data pieces stored in thenode 30 that is assigned to serve as the owner node. - More particularly, as illustrated in
FIG. 8 , when an update request or a registration request is received with respect to a partition for which the corresponding node is assigned to serve as the owner node, theaccess processing unit 33 sends an identical access request to anothernode 30 that is assigned to serve as the backup node. Then, theaccess processing unit 33 of theother node 30 receives, from thenode 30 assigned to serve as the owner node, an update request or a registration request with respect to the partition for which the correspondingnode 30 is assigned to serve as the backup node; and performs an update operation or a registration operation according to the received request. - By performing such a replication operation, the
access processing unit 33 can achieve synchronization of data pieces between thenode 30 assigned to serve as the owner node and thenode 30 assigned to serve as the backup node. - In the case in which the replication operation is performed by the
access processing unit 33 of thenode 30 that is assigned to serve as the owner node, theaccess processing unit 33 can also send the transaction that contains a sequence of access requests. In this case, in thenode 30 that is assigned to serve as the backup node, theaccess processing unit 33 performs a replication operation by means of transaction processing. - Meanwhile, when an update request or a registration request is processed, the
access processing unit 33 updates the time stamp of the corresponding partition in the data retention table that is stored in thetable memory unit 32. - The
node managing unit 34 communicates messages with thecluster managing unit 24 of themanagement device 20 via a network. Upon receiving a message called heartbeat from thecluster managing unit 24, thenode managing unit 34 sends back a response message in case a failure has occurred in the corresponding node. Moreover, thenode managing unit 34 sends the data retention table, which is stored in thetable memory unit 32, to thecluster managing unit 24 on a periodic basis. - Furthermore, the
node managing unit 34 receives the next-role table from thecluster managing unit 24, and stores the next-role table in thetable memory unit 32. Herein, upon receiving the next-role table, as illustrated inFIG. 9 , thenode managing unit 34 performs short-term synchronization processing that is a synchronization processing in which the operations being performed according to the access request received from a client are temporarily discontinued for the purpose of making each of a plurality ofnodes 30 operate according to the role assigned in the next-role table. - More particularly, with respect to each of a plurality of partitions, the
node managing unit 34 causes theaccess processing unit 33 to perform an identical operation to the replication operation for the purpose of making each of a plurality ofnodes 30 operate according to the role assigned in the next-role table. As a result, with respect to each of a plurality of partitions, thenode managing unit 34 can achieve synchronization of data pieces between thenode 30 assigned to serve as the owner node and the node assigned to serve as the backup node. Upon achieving synchronization of data pieces, thenode managing unit 34 causes theaccess processing unit 33 to rewrite the current-role table with the contents of the next-role table and to perform operations with respect to each of a plurality of partitions according to the new roles assigned in the next-role table. - Once the short-term synchronization processing is completed, the
node managing unit 34 updates the contents of the current-role table with the contents of the next-role table. After that, theaccess processing unit 33 can receive an access request from a client. - The
node managing unit 34 receives the target-role table from thecluster managing unit 24 on a periodic basis, and stores that target-role table in thetable memory unit 32. - As illustrated in
FIG. 10 , the transferringunit 35 performs long-term synchronization processing that is a synchronization processing for the purpose of making each of a plurality ofnodes 30 operate in the role assigned in the next-role table without discontinuing the operations being performed according to the access request received from a client. More particularly, the transferringunit 35 sends, without discontinuing the operations being performed according to the access request, the data pieces of the partition, for which the corresponding node is assigned to serve as the owner node according to the current-role table, to theother node 30 that is assigned to serve as the backup candidate node according to the target-role table and that is neither assigned to serve as the owner node nor assigned to serve as the backup node. Moreover, the transferringunit 35 sends, without discontinuing the operations being performed according to the access request, the data pieces of the partition, for which the corresponding node is assigned to serve as the backup candidate node according to the target-role table and for which the corresponding node is neither assigned to serve as the owner node nor assigned to serve as the backup node according to the current-role table, from theother node 30 that is assigned to serve as the owner node according to the current-role table. - By performing such long-term synchronization processing, a replica of the data pieces, which are stored in the
node 30 that is assigned to serve as the owner node, can be stored by the transferringunit 35 in thenode 30 that is not assigned to serve as the owner node or the backup node. As a result, upon the completion of the long-term synchronization processing, the transferringunit 35 can newly generate anode 30 that can serve as the owner node or the backup node. - In the case of performing long-term synchronization processing, the transferring
unit 35 sends, in the background while not interrupting the transaction execution by theaccess processing unit 33, the data pieces of the partition for which the corresponding node is assigned to serve as the owner node to thenode 30 that is assigned to serve as the backup candidate node. Moreover, the transferringunit 35 receives, in the background, the data pieces of the partition for which the corresponding node is assigned to serve as the backup candidate node from thenode 30 that is assigned to serve as the owner node. Herein, regarding an operation performed in the background; as an example, when thenode 30 includes a plurality ofCPUs 12, an operation performed using some of theCPUs 12 not performing the transaction operation represents an operation performed in the background. Alternatively, regarding an operation performed in the background; as an example, when theCPU 12 performs operations in a time-shared manner, an operation performed in some of the time slots in which theCPU 12 does not perform the transaction operation represents an operation performed in the background. With that, the transferringunit 35 becomes able to perform long-term synchronization processing without causing a decrease in the response speed with respect to an access request from a client. - Meanwhile, regarding the data pieces of the partition for which the corresponding node is assigned to serve as the backup node according to the current-role table, the transferring
unit 35 can send those data pieces to theother node 30 that is assigned to serve as the backup candidate node according to the target-role table and that is neither assigned to serve as the owner node nor assigned to serve as the backup node according to the target-role table. In this case, the transferringunit 35 performs operations conditional upon the fact that the same data pieces have not been sent from anothernode 30. -
FIG. 11 is a diagram illustrating a transition in the roles assigned to thenodes 30. With respect to each of a plurality of partitions, thenodes 30 make transition between the state of being assigned to serve as the “owner node”, or the state of being assigned to serve as the “backup node”, or the state of being assigned to serve as the “backup candidate node”, or the state of being assigned “no role”. - A
node 30 makes mutual transition between the state of being assigned “no role” and the state of being assigned to serve as the “owner node” as a result of short-term synchronization processing. Similarly, anode 30 makes mutual transition between the state of being assigned to serve as the “owner node” and the state of being assigned to serve as the “backup node” as a result of short-term synchronization processing. Moreover, anode 30 makes mutual transition between the state of being assigned to serve as the “backup node” and the state of being assigned “no role” as a result of short-term synchronization processing. - Furthermore, a
node 30 makes mutual transition between the state of being assigned “no role” and the state of being assigned to serve as the “backup candidate node” as a result of long-term synchronization processing. Besides, anode 30 makes transition from the state of being assigned to serve as the “backup candidate node” to the state of being assigned to serve as the “backup node” as a result of short-term synchronization processing. -
FIG. 12 is a diagram illustrating an operation flowchart of thecluster managing unit 24 of themanagement device 20. Thecluster managing unit 24 performs operations from Step S111 to Step S142 explained below. - Firstly, at Step S111, the
cluster managing unit 24 detects a data-retention-table receive event, a first fixed-cycle event, or a second fixed-cycle event. A data-retention-table receive event occurs when thecluster managing unit 24 receives the data retention table sent by each of a plurality ofnodes 30. A first fixed-cycle event as well as a second fixed-cycle event occurs on a periodic basis. However, the interval of occurrence of the first fixed-cycle events is different from the interval of occurrence of the second fixed-cycle events. - When a data-retention-table receive event is detected, the
cluster managing unit 24 proceeds to the operation at Step S121. When a first fixed-cycle event is detected, thecluster managing unit 24 proceeds to the operation at Step S131. When a second fixed-cycle event is detected, thecluster managing unit 24 proceeds to the operation at Step S141. - When a data-retention-table receive event is detected; at Step S121, the
cluster managing unit 24 determines whether or not a data-retention-table receive event has occurred for the first time since the startup. - If the data-retention-table receive event has occurred for the first time since the startup (Yes at Step S121), then the
cluster managing unit 24 proceeds to the operation at Step S122. At Step S122, thecluster managing unit 24 registers the received data retention table in thetable memory unit 21. - Then, at Step S123, the
cluster managing unit 24 causes the first assigningunit 22 to calculate the next-role table. The operation by which the first assigningunit 22 calculates the next-role table is explained with reference toFIG. 13 . - Subsequently, at Step S124, the
cluster managing unit 24 distributes the next-role table to each of a plurality ofnodes 30. Once the operation at Step S124 is completed, thecluster managing unit 24 returns to the operation at Step S111. - Meanwhile, if it is not the first time since the startup that the data-retention-table receive event has occurred (No at Step S121); then, at Step S125, the
cluster managing unit 24 updates the data retention table stored in thetable memory unit 21. Once the operation at Step S125 is completed, thecluster managing unit 24 returns to the operation at Step S111. - Meanwhile, if a first fixed-cycle event is detected; then, at Step S131, the
cluster managing unit 24 determines whether or not, during the period between the previous first fixed-cycle event and the current first fixed-cycle event, the data retention table is received from each of a plurality ofnodes 30. If the data retention table is received from each of a plurality of nodes 30 (Yes at Step S131), then thecluster managing unit 24 proceeds to the operation at Step S134. However, if the data retention table is not received from any one node 30 (No at Step S131), then thecluster managing unit 24 proceeds to the operation at Step S132. - At Step S132, the
cluster managing unit 24 performs an operation to separate off thenode 30, from which the data retention table could not be received, from thedatabase system 10. Then, at Step S133, thecluster managing unit 24 updates the data retention table by deleting the contents of the separatednode 30 from the data retention table. Once the operation at Step S133 is completed, thecluster managing unit 24 proceeds to the operation Step S134. - At Step S134, the
cluster managing unit 24 causes the first assigningunit 22 to calculate the next-role table. Regarding the operations by which the first assigningunit 22 calculates the next-role table, the explanation is given later with reference toFIG. 13 . - Subsequently, at Step S135, the
cluster managing unit 24 determines whether or not there is a change in the next-role table. For example, when anode 30 in which a failure has occurred is separated off, or when anew node 30 is added, or when long-term synchronization processing is completed and there is a change in thenode 30 assigned to serve as the backup node; the next-role table undergoes a change. - If the next-role table has not changed (No at Step S135), then the
cluster managing unit 24 returns to the operation at Step S111. However, when there is a change in the next-role table (Yes at Step S135); then, at Step S136, thecluster managing unit 24 distributes the changed next-role table to each of a plurality ofnodes 30. Once the operation at Step S136 is completed, thecluster managing unit 24 returns to the operation at Step S111. - Meanwhile, when a second fixed-cycle event is detected; then, at Step S141, the
cluster managing unit 24 causes the second assigning unit 23 to calculate the target-role table. Regarding the operations by which the second assigning unit 23 calculates the target-role table, the explanation is given later with reference toFIG. 14 . - Subsequently, at Step S142, the
cluster managing unit 24 distributes the calculated target-role table to each of a plurality ofnodes 30. Once the operation at Step S142 is completed, thecluster managing unit 24 returns to the operation at S111. -
FIG. 13 is a flowchart for explaining the operations by which the first assigningunit 22 of themanagement device 20 calculates the next-role table. The first assigningunit 22 is called by thecluster managing unit 24 at Step S123 and Step S134 illustrated inFIG. 12 , and performs the operations from Step S211 to Step S220 explained below. - Firstly, at Step S211, the first assigning
unit 22 initializes the next-role table. At that time, the first assigningunit 22 associates each of a plurality ofnodes 30 specified in the next-role table with thenodes 30 specified in the data retention table. With that, in the next-role table, the first assigningunit 22 can reflect thenode 30 that has been separated off due to a failure and thenode 30 that has been newly added. - Then, from Step S212 to Step S220, the first assigning
unit 22 performs a loop operation on a partition-by-partition basis. For example, if the database is partitioned into a first partition to a third partition, then the first assigningunit 22 performs the operations from Step S212 to Step S220 with respect to each of the first partition, the second partition, and the third partition. - In the loop operation performed on a partition-by-partition basis, firstly, at Step S213, the first assigning
unit 22 selects, for the target partition, the set ofnodes 30 having the largest time stamp specified in the data retention table. In this example, a time stamp is a value that is incremented after every instance of updating the data pieces of the target partition. Thus, at Step S213, the first assigningunit 22 can select, for the target partition, the set ofnodes 30 in which the latest data pieces are stored. - Then, at Step S214, from the set of
nodes 30 selected at Step S213, the first assigningunit 22 selects asingle node 30 that is assigned to serve as the owner node and the backup node for the least number of partitions according to the next-role table; and assigns the selectednode 30 to serve as the owner node. With that, from among the set ofnodes 30 in which the latest data pieces are stored, the first assigningunit 22 can assign thenode 30 having the least load as the owner node. - Meanwhile, if more than one
node 30 is assigned to serve as the owner node and the backup node for the least number of partitions, then the first assigningunit 22 can assign anode 30 having a higher computing power to serve as the owner node on a priority basis as compared to anode 30 having a lower computing power. Alternatively, the first assigningunit 22 can assign anode 30 receiving a smaller number of access requests to serve as the owner node on a priority basis as compared to anode 30 receiving a greater number of access requests. - Subsequently, at Step S215, for the target partition, the first assigning
unit 22 selects such a set ofnodes 30 that does not include thenode 30 assigned to serve as the owner node but that includes thenodes 30 having respective time stamps within a predetermined difference from the largest time stamp. With that, for the target partition, the first assigningunit 22 can select a set ofnodes 30 in which either the latest data pieces are stored or the data pieces relatively closer to the latest data pieces are stored. - Then, from Step S216 to Step S219, the first assigning
unit 22 performs a loop operation for a number of times equal to the number of replications. Herein, the number of replications represents the largest number ofnodes 30 for which the replication operation can be performed with thenode 30 that is assigned to serve as the owner node. Thus, the number of replications is identical to the number of assignable backup nodes. - In the loop operation performed for a number of times equal to the number of replications; firstly, at Step S217, from the set of
nodes 30 selected at Step S215, the first assigningunit 22 determines whether or not there is anode 30 which can be assigned to serve as the backup node. If there is anode 30 which can be assigned to serve as the backup node (Yes at Step S217), then the first assigningunit 22 proceeds to the operation at Step S218. - On the other hand, if there is no node which can be assigned as the backup node (No at Step S217), then the first assigning
unit 22 mandatorily exits the loop operation performed for a number of times equal to the number of replications, and proceeds to the operation at Step S220. In view of that, sometimes the first assigningunit 22 calculates the next-role table in which the backup nodes are not present or in which the number of backup nodes is smaller than the number of replications. - At Step S213, from among the set of
nodes 30 selected at Step S215, the first assigningunit 22 assigns, as the backup node, thenode 30 that is assigned to serve as the owner node and the backup node for the least number of partitions according to the next-role table. With that, from among the set ofnodes 30 in which either the latest data pieces are stored or the data pieces relatively closer to the latest data pieces are stored, the first assigningunit 22 can assign thenodes 30 to serve as the backup nodes in ascending order of the processing load. - Meanwhile, if more than one
node 30 is assigned to serve as the owner node and the backup node for the least number of partitions, then the first assigningunit 22 can assign anode 30 having a higher computing power to serve as the backup node on a priority basis as compared to anode 30 having a lower computing power. Alternatively, the first assigningunit 22 can assign anode 30 receiving a smaller number of access requests to serve as the backup node on a priority basis as compared to anode 30 receiving a greater number of access requests. - Subsequently, the first assigning
unit 22 excludes, from the set ofnodes 30, thenode 30 assigned to serve as the backup node; and proceeds to the operation at Step S219. - At Step S219, if the number of operations from Step S216 to Step S219 is smaller than the number of replications, then the first assigning
unit 22 returns to the operation at Step S216. However, if the number of operations from Step S216 to Step S219 is equal to the number of replications, then the first assigningunit 22 proceeds to the operation at Step S220. - Then, at Step S220, if the operations from Step S212 to Step S220 are not yet performed with respect to all partitions, then the first assigning
unit 22 returns to the operation at Step S216. When the operations from Step S212 to Step S220 are performed with respect to all partitions, the first assigningunit 22 ends the calculation of the next-role table and exits the present flowchart. - In this way, the first assigning
unit 22 assigns thenodes 30 in such a way that each of a plurality of partitions has an owner node assigned thereto. With that, the first assigningunit 22 can at least make the database work. Along with that, the first assigningunit 22 assigns thenodes 30 in such a way that backup nodes are present to the extent possible. With that, the first assigningunit 22 can guarantee the redundancy of the database. Moreover, the first assigningunit 22 assigns thenodes 30 to serve as the owner nodes and the backup nodes in ascending order of the processing load. With that, the first assigningunit 22 can even out the processing load of each of a plurality ofnodes 30. -
FIG. 14 is a flowchart for explaining the operations by which the second assigning unit 23 of themanagement device 20 calculates the target-role table. The second assigning unit 23 is called by thecluster managing unit 24 at Step S141 illustrated inFIG. 12 , and performs the operations from Step S311 to Step S324 explained below. - Firstly, at Step S311, the second assigning unit 23 initializes the target-role table. At that time, the second assigning unit 23 associates each of a plurality of
nodes 30 specified in the target-role table with thenodes 30 specified in the data retention table. With that, in the target-role table, the second assigning unit 23 can reflect thenode 30 that has been separated off due to a failure and thenode 30 that has been newly added. - Then, at Step S312, the second assigning unit 23 sorts the partitions specified in the current-role table in ascending order of the total number of owner nodes and backup nodes assigned to each partition. With that, the second assigning unit 23 can assign the backup candidate nodes to the partitions in ascending order of the number of
nodes 30 assigned to serve as the backup nodes for each partition. That is, in the current-role table, the second assigning unit 23 can assign thenodes 30 to serve as the backup candidate nodes to the partitions in ascending order of the redundancy of each partition. - Subsequently, at Step S313, the second assigning unit 23 initializes a load value with respect to each of a plurality of
nodes 30. A load value is a value that increases in response to assigning the correspondingnode 30 to serve as the backup candidate node. Herein, as an example, the second assigning unit 23 initializes each load value to “0”. - Then, from Step S314 to Step S324, the second assigning unit 23 performs a loop operation on a partition-by-partition basis. In this case, the second assigning unit 23 selects the target partition in the order of partitions sorted in the current-role table at Step S312, and performs the loop operation.
- In the loop operation performed on a partition-by-partition basis; firstly, at Step S315, the second assigning unit 23 selects, for the target partition, the
node 30 having the largest time stamp specified in the data retention table. - Then, at Step S316, the second assigning unit 23 sorts the
nodes 30 specified in the current-role table in descending order of values V. Herein, regarding thenodes 30 assigned to serve as the owner nodes, the value V is set to “+1”. Regarding the nodes assigned to serve as the backup nodes, the value V is set to “+1”. Regarding the nodes having the largest time stamp, the value V is set to “+1”. Thus, regarding anode 30 that neither is an owner node nor is a backup node nor has the largest time stamp, the value V is equal to “0”. - With that, as the
nodes 30 that would serve as the backup candidate nodes, the second assigning unit 23 can firstly assign thenodes 30 that are assigned to serve as the owner nodes and the backup nodes or thenodes 30 in which the latest data pieces are stored. - Meanwhile, if more than one
node 30 has the same value V, then the second assigning unit 23 arranges thenodes 30 having higher computing power near the top of the list so that they are assigned to serve as the backup candidate nodes on a priority basis. With that, the second assigning unit 23 can assign thenodes 30 to serve as the backup candidate nodes in such a way that anode 30 having a higher computing power serves as the owner node or the backup node on a priority basis as compared to anode 30 having a lower computing power. Alternatively, the second assigning unit 23 can arrange thenodes 30 receiving a smaller number of access requests near the top of the list so that they are assigned to serve as the backup candidate node on a priority basis. With that, the second assigning unit 23 can assign thenodes 30 to serve as the backup candidate nodes in such a way that anode 30 receiving a smaller number of access requests to serve as the owner node or the backup node on a priority basis as compared to anode 30 receiving a greater number of access requests. - Subsequently, at Step S317, the second assigning unit 23 calculates an upper limit load MLOAD, which is a constant number, using Equation (1) given below.
-
MLOAD={number of partitions (number of replications+1)+(number of nodes−1)}/(number of nodes) (1) - Herein, (number of replications+1) represents the maximum number of owner nodes and backup nodes that can be assigned to a single partition. The upper limit load MLOAD represents the upper limit standard of the number of partitions for which a
single node 30 can be assigned to serve as the backup candidate node. - Subsequently, from Step S318 to Step S323, the second assigning unit 23 performs a loop operation for each
node 30. In this case, the second assigning unit 23 selects thetarget node 30 according to the order of nodes specified in the current-role table after the sorting performed at Step S316, and performs the loop operation. - In the loop operation performed for each
node 30; firstly, at Step S319, the second assigning unit 23 determines whether or not thenodes 30 equal in number to (number of replications+1) are assigned to serve as the backup candidate nodes. If thenodes 30 equal in number to (number of replications+1) are assigned to serve as the backup candidate nodes (Yes at Step S319), then the second assigning unit 23 proceeds to the operation at Step S324. In this case, with respect to the target partition, the second assigning unit 23 assigns the maximum number ofnodes 30 to serve as the backup candidate nodes. - However, if the
nodes 30 equal in number to (number of replications+1) are not assigned to serve as the backup candidate nodes (No at Step S319), then the second assigning unit 23 proceeds to the operation at Step S320. - At Step S320, the second assigning unit 23 determines whether the load value of the
target node 30 is smaller than the upper limit load MLOAD. If the load value of thetarget node 30 is smaller than the upper limit load MLOAD (Yes at Step S320), then the second assigning unit 23 proceeds to the operation at Step S321. - On the other hand, if the load value of the
target node 30 is equal to or greater than the upper limit load MLOAD (No at Step S320), then the second assigning unit 23 proceeds to the operation at Step S323. With that, in case aparticular node 30 has been assigned to serve as the backup candidate node for a number of times equal to or greater than a reference value, the second assigning unit 23 can avoid assigning thatnode 30 anymore to serve as the backup candidate node. As a result, the second assigning unit 23 can assign, in a distributed manner, the roles of the backup candidate nodes to a plurality ofnodes 30. - At Step S321, with respect to the target partition, the second assigning unit 23 assigns the
target node 30 to serve as the backup candidate node. Then, at Step S322, the second assigning unit 23 updates the load value of thetarget node 30 by adding “1”. With that, every time thenode 30 is assigned to serve as the backup candidate node, the second assigning unit 23 can increment the load value by one. Once the operation at Step S322 is completed, the second assigning unit 23 proceeds to the operation at Step S323. - Then, at Step S323, if the operations from Step S318 to Step S323 are not yet performed for each of a plurality of
nodes 30, then the second assigning unit 23 returns to the operation at Step S318. When the operations from Step S318 to Step S323 are performed for each of a plurality ofnodes 30, the second assigning unit 23 proceeds to the operation at Step S324. - Subsequently, at Step S324, if the operations from Step S314 to Step S324 are not yet performed for each of a plurality of partitions, then the second assigning unit 23 returns to the operation at Step S314. When the operations from Step S314 to Step S324 are performed for each of a plurality of partitions, then the second assigning unit 23 ends the calculation of the target-role table and exits the present flowchart.
- In this way, the second assigning unit 23 assigns the backup candidate nodes to the partitions in ascending order of the redundancy (i.e., in ascending order of the number of assigned backup nodes). Hence, the redundancy of the database can be guaranteed in an efficient manner. Moreover, with respect to a
node 30 that has been assigned to serve as the backup candidate node for a number of times equal to or greater than a reference value, the second assigning unit 23 does not assign thenode 30 anymore to serve as the backup candidate node. As a result, a plurality ofnodes 30 can be assigned in a distributed manner to serve as the backup candidate nodes. -
FIG. 15 is a diagram illustrating an operation flowchart of theaccess processing unit 33 of thenode 30. Theaccess processing unit 33 performs operations from Step S411 to Step S433 explained below. - Firstly, at Step S411, the
access processing unit 33 detects a request receive event from a client or detects a request receive event attributed to a replication operation. A request receive event from a client occurs in the case when an access request with respect to the data is received from a client via a network. A request receive event attributed to a replication operation occurs in the case when an update request or a registration request attributed to a replication operation is received via a network from anothernode 30 serving as the owner node. - When a request receive event from a client is detected, the
access processing unit 33 proceeds to the operation at Step S421. When a request receive event attributed to a replication operation is detected, theaccess processing unit 33 proceeds to the operation at Step S431. - When a request receive event from a client is received; at Step S421, for example, the
access processing unit 33 calculates and obtains the number given to a request destination partition specified in the access request from the client. Then, at Step S422, theaccess processing unit 33 refers to the current-role table and determines whether or not the corresponding node has been assigned to serve as the owner node for the request destination partition. - If the corresponding node has not been assigned to serve as the owner node for the request destination partition (No at Step S422); then, at Step S423, the
access processing unit 33 notifies the client about the number given to thenode 30 which has been assigned to serve as the owner node, and returns to the operation at Step S411. - On the other hand, if the corresponding node has been assigned to serve as the owner node for the request destination partition (Yes at Step S422); then, at Step S424, the
access processing unit 33 determines whether the type of the access request points to a reference request, or an update request, or a registration request. - If the type of the access request points to a reference request (reference request at Step S424); then, at Step S425, the
access processing unit 33 reads from thedata storing unit 31 data for which the reference request is issued and sends the data to the client, and returns to the operation at Step S411. - If the type of the access request points to an update request or a registration request (update request/registration request at Step S424); then, at Step S426, the
access processing unit 33 performs a replication operation with theother node 30 that has been assigned to serve as the backup node for the request destination partition. That is, theaccess processing unit 33 sends an access request, which is identical to the update request or the registration request received from the client, to thenode 30 which has been assigned to serve as the backup node. - Once the replication operation is completed; then, at Step S427, according to the update request or the registration request issued by the client, the
access processing unit 33 either updates the data stored in thedata storing unit 31 or registers new data in thedata storing unit 31. Then, at Step S428, theaccess processing unit 33 updates the corresponding time stamp in the data retention table by incrementing the time stamp by one, and returns to the operation at Step S411. - Meanwhile, if a request receive event attributed to a replication operation is detected; then, at Step S431, according to an update request or a registration request issued by the
node 30 which has been assigned to serve as the owner node, theaccess processing unit 33 either updates the data stored in thedata storing unit 31 or registers new data in thedata storing unit 31. Then, at Step S432, theaccess processing unit 33 updates the corresponding time stamp in the data retention table by incrementing the time stamp by one. Subsequently, at Step S433, theaccess processing unit 33 notifies the owner node about the completion of updating or registration, and returns to the operation at Step S411. - Meanwhile, at Step S425 and Step S428, the
access processing unit 33 can also receive from a client a transaction that contains a sequence of access requests, and can perform transaction processing according to the access requests received from the client. Moreover, at Step S426, in the case of sending an update request or a registration request by means of a replication operation, theaccess processing unit 33 can send the transaction containing a sequence of access requests to thenode 30 that is assigned to serve as the backup node. Furthermore, at Step S431, theaccess processing unit 33 can receive the transaction, which contains a sequence of access requests, from thenode 30 that is assigned to serve as the owner node; and can perform transaction processing according to the access requests received from thenode 30 that is assigned to serve as the owner node. -
FIG. 16 is a diagram for explaining an operation flowchart of thenode managing unit 34 of thenode 30. Thenode managing unit 34 performs operations from Step S511 to Step S541 explained below. - Firstly, at Step S511, the
node managing unit 34 detects a third fixed-cycle event, a next-role-table receive event, or a target-role-table receive event. A third fixed-cycle receive event occurs on a periodic basis. A next-role-table receive event occurs when thenode managing unit 34 receives the next-role table. A target-role-table receive event occurs when thenode managing unit 34 receives the target-role table. - When a third fixed-cycle event is detected, the
node managing unit 34 proceeds to the operation at Step S521. When a next-role-table receive event is detected, thenode managing unit 34 proceeds to the operation at Step S531. When a target-role-table receive event is detected, thenode managing unit 34 proceeds to the operation at Step S541. - When a third fixed-cycle event is detected; then, at Step S521, the
node managing unit 34 sends the data retention table, which is stored in thetable memory unit 32, to thecluster managing unit 24 of themanagement device 20. Then, thenode managing unit 34 returns to the operation at Step S511. - When a next-role-table receive event is detected; then, at Step S531, if the
access processing unit 33 is executing a transaction, thenode managing unit 34 discontinues the transaction. Then, at Step S532, thenode managing unit 34 performs short-term synchronization processing according to the next-role table that is received. - More particularly, with respect to each of a plurality of partitions, if the data pieces stored in the
node 30 assigned to serve as the owner node differ from the data pieces stored in thenode 30 assigned to serve as the backup node, then thenode managing unit 34 causes theaccess processing unit 33 to perform an operation identical to the replication operation. With that, with respect to each of a plurality of partitions, thenode managing unit 34 can achieve synchronization between thenode 30 assigned to serve as the owner node and thenode 30 assigned to serve as the backup node. Then, with respect to each of a plurality of nodes, thenode managing unit 34 cases theaccess processing unit 33 to perform operations according to the new roles (as the owner node and the backup node) assigned in the next-role table. - Once the short-term synchronization processing is completed; then, at Step S533, the
node managing unit 34 rewrites the current-role table with the contents of the next-role table. After that, theaccess processing unit 33 can receive an access request from a client. - Subsequently, at Step S534, the
node managing unit 34 resumes the transaction that was discontinued. Then, thenode managing unit 34 returns to the operation at Step S511. - If a target-role-table receive event is detected; then, at Step S541, the
node managing unit 34 updates the target-role table stored in thetable memory unit 32. Then, thenode managing unit 34 returns to the operation at Step S511. -
FIG. 17 is a diagram illustrating an operation flowchart of the transferringunit 35 of thenode 30. The transferringunit 35 performs operations from S611 to Step S618 as well as operations at Step S621 and Step S622 explained below. - Firstly, at Step S611, the transferring
unit 35 determines whether or not the processing load of theCPU 12 is greater than a predetermined reference value. If the processing load of theCPU 12 is greater than the reference value (Yes at Step S611); then, at Step S612, the transferringunit 35 switches to a sleep mode for a certain amount of time and then returns to the operation at Step S611. - However, if the processing load of the
CPU 12 is equal to or smaller than the reference value (No at Step S611); then, at Step S613, the transferringunit 35 detects the difference between the current-role table and the target-role table. Subsequently, at Step S614, the transferringunit 35 determines whether or not there is difference between the current-role table and the target-role table. - If there is no difference between the current-role table and the target-role table (No at Step S614); then, at Step S615, the transferring
unit 35 deletes the data pieces included in a partition with respect to which the corresponding node has been assigned a role in none of the current-role table, the next-role table, and the target-role table. After performing the operation at Step S615, the transferringunit 35 returns to the operation at Step S611. - On the other hand, if there is difference between the current-role table and the target-role table (Yes at Step S614); then, at Step S616, for the partition with respect to which the corresponding node is assigned to serve as the owner node in the current-role table, the transferring
unit 35 detects anothernode 30 that is assigned to serve as the backup candidate node in the target-role table as well as that is neither assigned to serve as the owner node nor assigned to serve as the backup node in the current-role table. - Subsequently, at Step S617, the transferring
unit 35 determines whether or not anothernode 30 is present that is assigned to serve as the backup candidate node in the target-role table as well as that is neither assigned to serve as the owner node nor assigned to serve as the backup node in the current-role table. If such anode 30 is not present (No at Step S617), then the transferringunit 35 returns to the operation at Step S611. - On the other hand, if such a node is present (Yes at Step S617); then, at Step S618, the transferring
unit 35 starts long-term synchronization processing without discontinuing the operations with respect to the access request from the client. More particularly, regarding the partition for which the corresponding node is assigned to serve as the owner node in the current-role table, the transferringunit 35 sends the data pieces of that partition to theother node 30 that is assigned to serve as the backup candidate node in the target-role table as well as that is neither assigned to serve as the owner node nor assigned to serve as the backup node in the current-role table. In this case, the transferringunit 35 sends the data pieces in the background so as not to interrupt the transaction execution by theaccess processing unit 33. With that, the transferringunit 35 becomes able to perform the long-term synchronization processing without causing a decrease in the response speed with respect to the access request from the client. - Once the operation at Step S618 is completed, the transferring
unit 35 returns to the operation at Step S611. - Meanwhile, in the
other node 30 that is assigned to serve as the backup candidate node in the target-role table as well as that is neither assigned to serve as the owner node nor assigned to serve as the backup node in the current-role table; the corresponding transferringunit 35 performs operations at Step S621 and Step S622 explained below. - At Step S621, the transferring
unit 35 starts long-term synchronization processing without discontinuing the operations with respect to the access request from the client. More particularly, regarding the partition with respect to which the corresponding node is assigned to serve as the backup candidate node in the target-role table as well as is neither assigned to serve as the owner node nor assigned to serve as the backup node in the current-role table, the transferringunit 35 receives the data pieces of that partition from thenode 30 that is assigned to serve as the owner node with respect to that partition. In this case, the transferringunit 35 receives the data pieces in the background so as not to interrupt the transaction execution by theaccess processing unit 33. - Subsequently, at Step S622, the transferring
unit 35 updates the data retention table. Once the long-term synchronization processing is completed, a replica of the data pieces, which are stored in thenode 30 assigned to serve as the owner node, gets stored. Thus, by updating the data retention table, the transferringunit 35 can match the time stamp for the concerned partition with the time stamp in the data retention table of the owner node. Once the operation at Step S622 is completed, the transferringunit 35 exits the present flowchart. - In this way, a replica of the data pieces of the
node 30 assigned to serve as the owner node is generated by the transferringunit 35 in thenode 30 that is assigned to serve as the backup candidate node. As a result, the transferringunit 35 can newly generate anode 30 that can be assigned to serve either as the owner node or as the backup node. - Explained below with reference to
FIG. 18 toFIG. 24 is an example of operations performed in thedatabase system 10.FIG. 18 is a diagram illustrating the states of a node 30-A to a node 30-C in a case in which each of the threepartitions # 1 to #3 is assigned with thenodes 30 serving as the owner node and the backup node. - In the present example, as illustrated in
FIG. 18 , with respect to thepartition # 1, the node 30-A is assigned to serve as the owner node and the node 30-B is assigned to serve as the backup node. Moreover, with respect to thepartition # 2, the node 30-B is assigned to serve as the owner node and the node 30-C is assigned to serve as the backup node. Furthermore, with respect to thepartition # 3, the node 30-C is assigned to serve as the owner node and the node 30-A is assigned to serve as the backup node. - The nodes 30-A and 30-B perform replication operations with respect to the
partition # 1 by means of transactions. As a result, as illustrated in data retention table in FIG. 18, the time stamps of the nodes 30-A and 30-B have the same value for thepartition # 1. - The nodes 30-B and 30-C perform replication operations with respect to the
partition # 2 by means of transactions. As a result, as illustrated in data retention table inFIG. 18 , the time stamps of the nodes 30-B and 30-C have the same value for thepartition # 2. - The nodes 30-C and 30-A perform replication operations with respect to the
partition # 3 by means of transactions. As a result, as illustrated in data retention table inFIG. 18 , the time stamps of the nodes 30-C and 30-A have the same value for thepartition # 3. -
FIG. 19 is a diagram illustrating the states of the node 30-A to the node 30-C after short-term synchronization processing is performed in response to a failure occurring in the node 30-C in the state illustrated inFIG. 18 . Herein, it is assumed that thecluster managing unit 24 of themanagement device 20 cannot detect the data retention table from the node 30-C in the state illustrated inFIG. 18 , and detects that a failure has occurred in the node 30-C. - When a failure occurs in any one
node 30 of a plurality ofnodes 30, thecluster managing unit 24 of themanagement device 20 separates off thenode 30 in which a failure has occurred and then causes the first assigningunit 22 to calculate the next-role table. - In response to being called by the
cluster managing unit 24; the first assigningunit 22 reassigns, with the exclusion of thenode 30 in which a failure has occurred, thenodes 30 that would serve as the owner nodes and thenodes 30 that would serve as the backup nodes, to thereby generate the next-role table. In this case, with the aim of at least making the database work, the first assigningunit 22 assigns thenodes 30 in such a way that each of a plurality of partitions has at least the owner node assigned thereto. For example, if a failure occurs in thenode 30 assigned to serve as the owner node; then the first assigningunit 22 assigns thenode 30 which was assigned to serve as the backup node to now serve as the owner node. Moreover, with the aim of further enhancing the redundancy of the database, the first assigningunit 22 assigns thenodes 30 in such a way that each of a plurality of partitions has the backup node assigned thereto to the extent possible. - In the present example, as a result of the reassignment, as illustrated in the next-role table in
FIG. 19 ; with respect to thepartition # 1, the node 30-A is assigned to serve as the owner node and the node 30-B is assigned to serve as the backup node. Moreover, with respect to thepartition # 2, the node 30-B is assigned to serve as the owner node. Furthermore, with respect to thepartition # 3, the node 30-A is assigned to serve as the owner node. - Then, the
cluster managing unit 24 distributes the next-role table to the node 30-A and the node 30-B. Upon receiving the next-role table, thenode managing unit 34 of each of the node 30-A and the node 30-B performs short-term synchronization processing and rewrites the contents of the current-role table with the contents of the next-role table. With that, each of the node 30-A and the node 30-B can perform operations according to its newly-assigned role. -
FIG. 20 is a diagram illustrating an example of long-term synchronization processing performed in the state illustrated inFIG. 19 . Thecluster managing unit 24 of themanagement device 20 separates off the node 30-C in which a failure has occurred, and then causes the second assigning unit 23 to calculate the target-role table. - In response to being called by the
cluster managing unit 24; the second assigning unit 23 reassigns, with the exclusion of thenode 30 in which a failure has occurred, thenodes 30 that would serve as the backup candidate nodes, to thereby generate the target-role table. In this case, with the aim of further enhancing the redundancy of the database, the second assigning unit 23 assigns thenodes 30 that would serve as the backup candidate nodes in such a way that each of a plurality of partitions at least has the owner node and the backup node assigned thereto. - In the state illustrated in
FIG. 19 , with respect to thepartition # 2 as well as thepartition # 3, nonode 30 is assigned to serve as the backup node. Thus, the second assigning unit 23 assigns the node 30-A to serve as the backup candidate node for thepartition # 2, and assigns the node 30-B to serve as the backup candidate node for thepartition # 3. - Meanwhile, in the present example, the second assigning unit 23 assigns, as the backup candidate nodes,
such nodes 30 too that are already assigned to serve as the owner nodes and the backup nodes in the current-role table. For that reason, as illustrated in the target-role table inFIG. 20 , with respect to each of thepartitions # 1 to #3, the node 30-A as well as the node 30-B is assigned to serve as the backup candidate node. - Then, the
cluster managing unit 24 distributes the target-role table to the node 30-A and the node 30-B. Subsequently, the transferringunit 35 of the node 30-A as well as the node 30-B performs long-term synchronization processing with respect to the portion of difference between the current-role table and the target-role table. That is, the transferringunit 35 of the node 30-B sends the data pieces of thepartition # 2 to the node 30-A in the background. Moreover, the transferringunit 35 of the node 30-A sends the data pieces of thepartition # 3 to the node 30-B in the background. - At that time, each transferring
unit 35 performs the long-term synchronization processing without discontinuing the operations with respect to the access request from the client. - Once the long-term synchronization processing is completed, the node 30-A can store therein a replica of the data pieces of the
partition # 2. Moreover, as illustrated in the data retention table inFIG. 20 , the time stamps of the nodes 30-A and 30-B have the same value for thepartition # 2. - Similarly, the node 30-B can store therein a replica of the data pieces of the
partition # 3. Moreover, as illustrated in the data retention table inFIG. 20 , the time stamps of the nodes 30-A and 30-B have the same value for thepartition # 3. -
FIG. 21 is a diagram illustrating the states of the node 30-A to the node 30-C after short-term synchronization processing is performed upon completion of the long-term synchronization processing explained with reference toFIG. 20 . Thecluster managing unit 24 of themanagement device 20 periodically calls the first assigningunit 22 and causes it to calculate the next-role table. - Upon being called by the
cluster managing unit 24 after the completion of the long-term synchronization processing, the first assigningunit 22 reassigns, with respect to each of a plurality of partitions, thenodes 30 to serve as the owner node and the backup node in such a way that the redundancy increases using the replicas generated in the long-term synchronization processing. - In the present example, as illustrated in the next-role table in
FIG. 21 , with respect to thepartition # 1, the node 30-A is assigned to serve as the owner node and the node 30-B is assigned to serve as the backup node. Moreover, with respect to thepartition # 2, the node 30-B is assigned to serve as the owner node and the node 30-A is assigned to serve as the backup node. Furthermore, with respect to thepartition # 3, the node 30-A is assigned to serve as the owner node and the node 30-B is assigned to serve as the backup node. - Then, the
cluster managing unit 24 distributes the next-role table to the node 30-A and the node 30-B. Upon receiving the next-role table, thenode managing unit 34 of each of the node 30-A and the node 30-B performs short-term synchronization processing and rewrites the contents of the current-role table with the contents of the next-role table. With that, each of the node 30-A and the node 30-B becomes able to perform operations according to its newly-assigned role. -
FIG. 22 is a diagram illustrating the states of the node 30-A to a node 30-D in a case in which the node 30-D is added to the state illustrated inFIG. 21 . Herein, for example, it is assumed that thecluster managing unit 24 of themanagement device 20 detects that the node 30-D is newly added in the state illustrated inFIG. 21 . - As illustrated in the data retention table in
FIG. 22 , thecluster managing unit 24 generates the data retention table in which the node 30-D is added. In response to the addition of the node 30-D in the data retention table, the first assigningunit 22 generates the next-role table in which the node 30-D is added. - Of course, the node 30-D does not have any data pieces stored therein. Thus, immediately after the addition of the node 30-D, the first assigning
unit 22 does not change the assignment of the owner nodes and the backup nodes. -
FIG. 23 is a diagram illustrating an example of long-term synchronization processing performed in the state illustrated inFIG. 22 . When anew node 30 is added, the second assigning unit 23 reassigns, with respect to each of a plurality of partitions, thenodes 30 including thenew node 30 to serve as the backup candidate nodes. In this case, with respect to each of a plurality of partitions, the second assigning unit 23 assigns thenodes 30 to serve as the backup candidate nodes in such a way that the difference in the number of assigned owner nodes and the number of assigned backup nodes is within a range of values determined in advance among thenodes 30. - In the present example, as a result of the reassignment, as illustrated in the target-role table in
FIG. 23 ; with respect to thepartition # 1, the node 30-A and the node 30-B are assigned to serve as the backup candidate nodes. Moreover, with respect to thepartition # 2, the node 30-B and the node 30-D are assigned to serve as the backup candidate nodes. Furthermore, with respect to thepartition # 3, the node 30-A and the node 30-D are assigned to serve as the backup candidate nodes. - Then, the
cluster managing unit 24 distributes the target-role table to the node 30-A, the node 30-B, and the node 30-D. Subsequently, the transferringunit 35 of each of the node 30-A, the node 30-B, and the node 30-D performs long-term synchronization processing with respect to the portion of difference between the current-role table and the target-role table. - That is, the transferring
unit 35 of the node 30-B sends the data pieces of thepartition # 2 to the node 30-D in the background. Moreover, the transferringunit 35 of the node 30-A sends the data pieces of thepartition # 3 to the node 30-D in the background. At that time, each transferringunit 35 performs the long-term synchronization processing without discontinuing the operations with respect to the access request from the client. - As a result, the node 30-D can store therein a replica of the data pieces of the
partition # 2. Moreover, as illustrated in the data retention table inFIG. 23 , the time stamps of the nodes 30-B and 30-D have the same value for thepartition # 2. - Besides, the node 30-D can store therein a replica of the data pieces of the
partition # 3. Moreover, as illustrated in the data retention table inFIG. 23 , the time stamps of the nodes 30-A and 30-D have the same value for thepartition # 3. -
FIG. 24 is a diagram illustrating the states of the node 30-A to the node 30-D after short-term synchronization processing is performed upon completion of the long-term synchronization processing explained with reference toFIG. 23 . Upon being called by thecluster managing unit 24 after the completion of the long-term synchronization processing, the first assigningunit 22 reassigns, with respect to each of a plurality of partitions, thenodes 30 to serve as the owner node and the backup node in such a way that the processing load becomes more equal among thenodes 30 using the replicas generated in the long-term synchronization processing. - In the present example, as a result of the reassignment, as illustrated in the next-role table in
FIG. 24 ; with respect to thepartition # 1, the node 30-A is assigned to serve as the owner node and the node 30-B is assigned to serve as the backup node. Moreover, with respect to thepartition # 2, the node 30-B is assigned to serve as the owner node and the node 30-D is assigned to serve as the backup node. Furthermore, with respect to thepartition # 3, the node 30-A is assigned to serve as the owner node and the node 30-D is assigned to serve as the backup node. - Then, the
cluster managing unit 24 distributes the next-role table to the node 30-A, the node 30-B, and the node 30-D. Upon receiving the next-role table, thenode managing unit 34 of each of the node 30-A, the node 30-B, and the node 30-D performs short-term synchronization processing and rewrites the contents of the current-role table with the contents of the next-role table. With that, each of the node 30-A, the node 30-B, and the node 30-D becomes able to perform operations according to its newly-assigned role. - In this way, in the
database system 10 according to the present embodiment, anode 30 that would serve as the backup candidate node is assigned, and the data pieces are transferred from anode 30 assigned to serve as the owner node to thenode 30 assigned to serve as the backup candidate node. As a result, it becomes possible to ensure redundancy of the database and to even out the processing load of each of a plurality ofnodes 30. Besides, in thedatabase system 10, such operations can be performed without discontinuing the operations with respect to the access request from the client. Hence, even if anode 30 is separated off or anode 30 is newly added, data relocation can be done without stopping the system. - Meanwhile, computer programs executed in the
management device 20 and thenodes 30 according to the present embodiment are recorded in the form of installable or executable files in a computer-readable recording medium such as a CD-ROM, a flexible disk (FD), a CD-R, or a DVD (Digital Versatile Disk). - Alternatively, the computer programs executed in the
management device 20 and thenodes 30 according to the present embodiment can be saved as downloadable files on a computer connected to the Internet or can be made available for distribution through a network such as the Internet. - Still alternatively, the computer programs executed in the
management device 20 and thenodes 30 according to the present embodiment can be stored in advance in a ROM or the like. - The computer program executed in the
management device 20 according to the present embodiment contains a module for each of the abovementioned constituent elements (thetable memory unit 21, the first assigningunit 22, the second assigning unit 23, and the cluster managing unit 24). As the actual hardware, for example, a CPU (processor) reads the computer program from the above-mentioned recording medium and runs it such that the computer program is loaded in a main memory device. As a result, thetable memory unit 21, the first assigningunit 22, the second assigning unit 23, and thecluster managing unit 24 are generated in the main memory device. - Similarly, the computer program executed in the
nodes 30 according to the present embodiment contains a module for each of the abovementioned constituent elements (thedata storing unit 31, thetable memory unit 32, theaccess processing unit 33, thenode managing unit 34, and the transferring unit 35). As the actual hardware, for example, a CPU (processor) reads the computer program from the abovementioned recording medium and runs it such that the computer program is loaded in a main memory device. As a result, thedata storing unit 31, thetable memory unit 32, theaccess processing unit 33, thenode managing unit 34, and the transferringunit 35 are generated in the main memory device. - While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel embodiments described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the embodiments described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.
Claims (16)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
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 (en) | 2013-03-12 | 2013-03-12 | Database system, program, and data processing method |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/056868 Continuation WO2014141393A1 (en) | 2013-03-12 | 2013-03-12 | Database system, program, and data processing method |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US18/167,959 Continuation US20230244694A1 (en) | 2013-03-12 | 2023-02-13 | Database system, computer program product, and data processing method |
Publications (1)
Publication Number | Publication Date |
---|---|
US20140279902A1 true US20140279902A1 (en) | 2014-09-18 |
Family
ID=51532972
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US14/206,819 Abandoned US20140279902A1 (en) | 2013-03-12 | 2014-03-12 | Database system, computer program product, and data processing method |
US18/167,959 Pending US20230244694A1 (en) | 2013-03-12 | 2023-02-13 | Database system, computer program product, and data processing method |
Family Applications After (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US18/167,959 Pending US20230244694A1 (en) | 2013-03-12 | 2023-02-13 | Database system, computer program product, and data processing method |
Country Status (6)
Country | Link |
---|---|
US (2) | US20140279902A1 (en) |
EP (2) | EP2975523A4 (en) |
JP (1) | JP5698865B2 (en) |
CN (1) | CN104185841B (en) |
AU (1) | AU2013381504B2 (en) |
WO (1) | WO2014141393A1 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3035595A1 (en) * | 2014-12-17 | 2016-06-22 | Alcatel Lucent | Routable distributed database for managing a plurality of entities of a telecommunication network |
US20170262346A1 (en) * | 2016-03-09 | 2017-09-14 | Commvault Systems, Inc. | Data management and backup of distributed storage environment |
CN107800551A (en) * | 2016-08-31 | 2018-03-13 | 北京优朋普乐科技有限公司 | Redis group systems and its method, the client for improving reliability |
US10162875B2 (en) | 2013-08-27 | 2018-12-25 | Kabushiki Kaisha Toshiba | Database system including a plurality of nodes |
US10685041B2 (en) | 2013-08-21 | 2020-06-16 | Kabushiki Kaisha Toshiba | Database system, computer program product, and data processing method |
US10977276B2 (en) * | 2015-07-31 | 2021-04-13 | International Business Machines Corporation | Balanced partition placement in distributed databases |
US11868333B2 (en) | 2019-12-13 | 2024-01-09 | Huawei Technologies Co., Ltd. | Data read/write method and apparatus for database |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110300188B (en) * | 2019-07-25 | 2022-03-22 | 中国工商银行股份有限公司 | Data transmission system, method and device |
Citations (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5970495A (en) * | 1995-09-27 | 1999-10-19 | International Business Machines Corporation | Method and apparatus for achieving uniform data distribution in a parallel database system |
USRE37600E1 (en) * | 1989-06-09 | 2002-03-19 | Ralph Weinger | Apparatus and method for data access |
US20040109436A1 (en) * | 2002-11-05 | 2004-06-10 | Microsoft Corporation | User-input scheduling of synchronization operation on a mobile device based on user activity |
US20050278392A1 (en) * | 2000-07-13 | 2005-12-15 | Microsoft Corporatioan | System and method for synchronizing multiple database files |
US20060117154A1 (en) * | 2003-09-09 | 2006-06-01 | Hitachi, Ltd. | Data processing system |
US20060203718A1 (en) * | 2005-03-14 | 2006-09-14 | Benhase Michael T | Method, apparatus and program storage device for providing a triad copy of storage data |
US20070198802A1 (en) * | 2004-04-30 | 2007-08-23 | Srinivas Kavuri | System and method for allocation of organizational resources |
US20080147673A1 (en) * | 2006-12-19 | 2008-06-19 | Aster Data Systems, 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 |
US20090049240A1 (en) * | 2007-08-17 | 2009-02-19 | Fujitsu Limited | Apparatus and method for storage management system |
US20090210642A1 (en) * | 2004-03-08 | 2009-08-20 | Hitachi, Ltd. | Point in time remote copy for multiple sites |
US20100076939A1 (en) * | 2008-09-05 | 2010-03-25 | Hitachi, Ltd. | Information processing system, data update method and data update program |
US20100114949A1 (en) * | 2008-11-06 | 2010-05-06 | Fujitsu Limited | Contents deletion/update apparatus, contents deletion/update method and recording medium |
US20100306495A1 (en) * | 2009-05-28 | 2010-12-02 | Fujitsu Limited | Recording medium storing management program, management device and management method |
US20100315946A1 (en) * | 2009-06-10 | 2010-12-16 | Cisco Technology,Inc. | Failure protection for access ring topology |
US20110055182A1 (en) * | 2009-09-02 | 2011-03-03 | Microsoft Corporation | File system |
US20110283277A1 (en) * | 2010-05-11 | 2011-11-17 | International Business Machines Corporation | Virtualization and dynamic resource allocation aware storage level reordering |
US20110289049A1 (en) * | 2010-05-19 | 2011-11-24 | Microsoft Corporation | Scaleable fault-tolerant metadata service |
US20120137094A1 (en) * | 2010-11-30 | 2012-05-31 | International Business Machines Corporation | Snapshot based replication |
US20120166390A1 (en) * | 2010-12-23 | 2012-06-28 | Dwight Merriman | Method and apparatus for maintaining replica sets |
US20130031051A1 (en) * | 2011-07-29 | 2013-01-31 | International Business Machines Corporation | Adding a kew 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 |
US20130166606A1 (en) * | 2011-12-23 | 2013-06-27 | Lars Fricke | Table creation for partitioned tables |
Family Cites Families (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0392942A (en) * | 1989-09-06 | 1991-04-18 | Hitachi Ltd | Storing method and accessing method for file |
US5555404A (en) * | 1992-03-17 | 1996-09-10 | Telenor As | Continuously available database server having multiple groups of nodes with minimum intersecting sets of database fragment replicas |
JPH09146812A (en) * | 1995-11-27 | 1997-06-06 | Sanyo Electric Co Ltd | Data base device |
EP0854423A1 (en) * | 1997-01-20 | 1998-07-22 | TELEFONAKTIEBOLAGET L M ERICSSON (publ) | Data partitioning and duplication in a distributed data processing system |
JPH10247181A (en) * | 1997-03-05 | 1998-09-14 | Mitsubishi Electric Corp | Base backup computer system |
DE19836347C2 (en) * | 1998-08-11 | 2001-11-15 | Ericsson Telefon Ab L M | Fault-tolerant computer system |
JP2003345640A (en) * | 2002-05-28 | 2003-12-05 | Mitsubishi Electric Corp | Data backup system |
CN1317658C (en) * | 2002-12-31 | 2007-05-23 | 联想(北京)有限公司 | Fault-tolerance approach using machine group node interacting buckup |
US20050144316A1 (en) * | 2003-12-06 | 2005-06-30 | William Loo | Method and system for service node redundancy |
JP2005196602A (en) | 2004-01-09 | 2005-07-21 | Hitachi Ltd | System configuration changing method in unshared type database management system |
JP4107676B2 (en) * | 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 | Transaction takeover system |
US7725764B2 (en) * | 2006-08-04 | 2010-05-25 | Tsx Inc. | Failover system and method |
JP5192226B2 (en) | 2007-12-27 | 2013-05-08 | 株式会社日立製作所 | Method for adding standby computer, computer and computer system |
JP5222617B2 (en) * | 2008-04-28 | 2013-06-26 | 株式会社日立製作所 | Information system and I / O processing method |
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 |
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 |
-
2013
- 2013-03-12 CN CN201380003048.1A patent/CN104185841B/en active Active
- 2013-03-12 JP JP2014504113A patent/JP5698865B2/en active Active
- 2013-03-12 EP EP13834378.5A patent/EP2975523A4/en not_active Ceased
- 2013-03-12 AU AU2013381504A patent/AU2013381504B2/en active Active
- 2013-03-12 EP EP23152630.2A patent/EP4191431A1/en active Pending
- 2013-03-12 WO PCT/JP2013/056868 patent/WO2014141393A1/en active Application Filing
-
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 (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
USRE37600E1 (en) * | 1989-06-09 | 2002-03-19 | Ralph Weinger | Apparatus and method for data access |
US5970495A (en) * | 1995-09-27 | 1999-10-19 | International Business Machines Corporation | Method and apparatus for achieving uniform data distribution in a parallel database system |
US20050278392A1 (en) * | 2000-07-13 | 2005-12-15 | Microsoft Corporatioan | System and method for synchronizing multiple database files |
US20040109436A1 (en) * | 2002-11-05 | 2004-06-10 | Microsoft Corporation | User-input scheduling of synchronization operation on a mobile device based on user activity |
US20060117154A1 (en) * | 2003-09-09 | 2006-06-01 | Hitachi, Ltd. | Data processing system |
US20090210642A1 (en) * | 2004-03-08 | 2009-08-20 | Hitachi, Ltd. | Point in time remote copy for multiple sites |
US20070198802A1 (en) * | 2004-04-30 | 2007-08-23 | Srinivas Kavuri | System and method for allocation of organizational resources |
US20060203718A1 (en) * | 2005-03-14 | 2006-09-14 | Benhase Michael T | Method, apparatus and program storage device for providing a triad copy of storage data |
US20080147673A1 (en) * | 2006-12-19 | 2008-06-19 | Aster Data Systems, 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 |
US20090049240A1 (en) * | 2007-08-17 | 2009-02-19 | Fujitsu Limited | Apparatus and method for storage management system |
US20100076939A1 (en) * | 2008-09-05 | 2010-03-25 | Hitachi, Ltd. | Information processing system, data update method and data update program |
US20100114949A1 (en) * | 2008-11-06 | 2010-05-06 | Fujitsu Limited | Contents deletion/update apparatus, contents deletion/update method and recording medium |
US20100306495A1 (en) * | 2009-05-28 | 2010-12-02 | Fujitsu Limited | Recording medium storing management program, management device and management method |
US20100315946A1 (en) * | 2009-06-10 | 2010-12-16 | Cisco Technology,Inc. | Failure protection for access ring topology |
US20110055182A1 (en) * | 2009-09-02 | 2011-03-03 | Microsoft Corporation | File system |
US20110283277A1 (en) * | 2010-05-11 | 2011-11-17 | International Business Machines Corporation | Virtualization and dynamic resource allocation aware storage level reordering |
US20110289049A1 (en) * | 2010-05-19 | 2011-11-24 | Microsoft Corporation | Scaleable fault-tolerant metadata service |
US20120137094A1 (en) * | 2010-11-30 | 2012-05-31 | International Business Machines Corporation | Snapshot based replication |
US20120166390A1 (en) * | 2010-12-23 | 2012-06-28 | Dwight Merriman | Method and apparatus for maintaining replica sets |
US20130031051A1 (en) * | 2011-07-29 | 2013-01-31 | International Business Machines Corporation | Adding a kew 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 |
US20130166606A1 (en) * | 2011-12-23 | 2013-06-27 | Lars Fricke | Table creation for partitioned tables |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10685041B2 (en) | 2013-08-21 | 2020-06-16 | Kabushiki Kaisha Toshiba | Database system, computer program product, and data processing method |
US10162875B2 (en) | 2013-08-27 | 2018-12-25 | Kabushiki Kaisha Toshiba | Database system including a plurality of nodes |
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 |
US20170262346A1 (en) * | 2016-03-09 | 2017-09-14 | Commvault Systems, Inc. | Data management and backup of distributed storage environment |
US10452490B2 (en) * | 2016-03-09 | 2019-10-22 | Commvault Systems, Inc. | Data management and backup of distributed storage environment |
US11237919B2 (en) | 2016-03-09 | 2022-02-01 | Commvault Systems, Inc. | Data transfer to a distributed storage environment |
US11301334B2 (en) * | 2016-03-09 | 2022-04-12 | Commvault Systems, Inc. | Monitoring of nodes within a distributed storage environment |
CN107800551A (en) * | 2016-08-31 | 2018-03-13 | 北京优朋普乐科技有限公司 | Redis group systems and its method, the client for improving reliability |
US11868333B2 (en) | 2019-12-13 | 2024-01-09 | Huawei Technologies Co., Ltd. | Data read/write method and apparatus for database |
Also Published As
Publication number | Publication date |
---|---|
EP4191431A1 (en) | 2023-06-07 |
WO2014141393A1 (en) | 2014-09-18 |
JPWO2014141393A1 (en) | 2017-02-16 |
EP2975523A1 (en) | 2016-01-20 |
AU2013381504A1 (en) | 2015-02-12 |
EP2975523A4 (en) | 2017-02-08 |
JP5698865B2 (en) | 2015-04-08 |
CN104185841B (en) | 2018-08-17 |
AU2013381504B2 (en) | 2016-06-23 |
CN104185841A (en) | 2014-12-03 |
US20230244694A1 (en) | 2023-08-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10685041B2 (en) | Database system, computer program product, and data processing method | |
US20230244694A1 (en) | Database system, computer program product, and data processing method | |
US10162875B2 (en) | Database system including a plurality of nodes | |
US10997163B2 (en) | Data ingestion using file queues | |
CN106506605B (en) | SaaS application construction method based on micro-service architecture | |
US11093468B1 (en) | Advanced metadata management | |
US9959332B2 (en) | System and method for massively parallel processor database | |
CN108121782B (en) | Distribution method of query request, database middleware system and electronic equipment | |
CN112507023A (en) | Replicated database distribution for workload balancing after cluster reconfiguration | |
US9323791B2 (en) | Apparatus and method for expanding a shared-nothing system | |
JP2014232483A (en) | Database system, retrieval method and program | |
CN111386521B (en) | Redistributing table data in a database cluster | |
US10437797B1 (en) | In-memory distributed database with a remote data store | |
US10289723B1 (en) | Distributed union all queries | |
US20140324928A1 (en) | Large-scale data transfer | |
CN108153759B (en) | Data transmission method of distributed database, intermediate layer server and system | |
Arrieta-Salinas et al. | Classic replication techniques on the cloud | |
KR102124897B1 (en) | Distributed Messaging System and Method for Dynamic Partitioning in Distributed Messaging System | |
KR101654969B1 (en) | Method and apparatus for assigning namenode in virtualized cluster environments | |
CN105518664A (en) | Managing database nodes | |
Karlapalem et al. | A Multi-agent Simulation Framework on Small Hadoop Clouds |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: KABUSHIKI KAISHA TOSHIBA, JAPAN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HATTORI, MASAKAZU;REEL/FRAME:032961/0095 Effective date: 20140426 Owner name: TOSHIBA SOLUTIONS CORPORATION, JAPAN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HATTORI, MASAKAZU;REEL/FRAME:032961/0095 Effective date: 20140426 |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: FINAL REJECTION MAILED |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: NON FINAL ACTION MAILED |
|
STCV | Information on status: appeal procedure |
Free format text: NOTICE OF APPEAL FILED |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: FINAL REJECTION MAILED |
|
STCV | Information on status: appeal procedure |
Free format text: NOTICE OF APPEAL FILED |
|
STCV | Information on status: appeal procedure |
Free format text: APPEAL BRIEF (OR SUPPLEMENTAL BRIEF) ENTERED AND FORWARDED TO EXAMINER |
|
STCV | Information on status: appeal procedure |
Free format text: EXAMINER'S ANSWER TO APPEAL BRIEF MAILED |
|
STCV | Information on status: appeal procedure |
Free format text: ON APPEAL -- AWAITING DECISION BY THE BOARD OF APPEALS |
|
STCV | Information on status: appeal procedure |
Free format text: BOARD OF APPEALS DECISION RENDERED |
|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- AFTER EXAMINER'S ANSWER OR BOARD OF APPEALS DECISION |