WO2015170386A1

WO2015170386A1 - Database management system, computer system, and database management method

Info

Publication number: WO2015170386A1
Application number: PCT/JP2014/062378
Authority: WO
Inventors: 有哉礒田; 敦友田; 知広花井; 一智牛嶋
Original assignee: 株式会社日立製作所
Priority date: 2014-05-08
Filing date: 2014-05-08
Publication date: 2015-11-12

Abstract

This computer system receives a transaction processing instruction for updating one or more data regions of a plurality of data regions, and generates, on the basis of the transaction processing instruction, a data structure including Tx states, which are transaction processing states assumed according to the transaction processing instruction. The computer system then associates each of the one or more data regions, which are to be updated through transaction processing, with the data structure. The computer system then performs transaction processing on the basis of the data structure.

Description

Database management system, computer system, and database management method

The present invention generally relates to transaction processing, for example, database consistency control technology.

Transaction processing is generally performed in a computer system. Specifically, for example, a DBMS (database management system) is known as a system for processing transactions. The DBMS controls the consistency of records that are input and output data. As a result, the user can always input / output the intended record. Such DBMS consistency control is generally referred to as Tx (transaction) processing. The Tx process is used when a plurality of reference and update instructions are realized atomically. In Tx processing, a plurality of instructions in a specified section are processed atomically by declaring the beginning and end of the instruction section to be executed atomically (Tx processing section).

To realize Tx processing, it is necessary to satisfy ACID characteristics. The ACID property indicates atomicity, consistency, isolation, and durability. Atomicity means that a plurality of instructions in the Tx processing section are all executed or not executed at all. Consistency means not making a transition to a state other than the rules given to the database. Independence means that the result of Tx processing is always the same as when Tx processing is executed sequentially. Persistence means that if the Tx process is committed, the result of the Tx process can always be restored even if a failure occurs in the DBMS. As described above, the DBMS maintains the consistency of the Tx processing by satisfying the ACID characteristic.

However, there is a problem that the processing capacity of the DBMS decreases due to the high consistency control. For example, a lock mechanism for maintaining independence and a log mechanism for maintaining persistence cause a reduction in processing capability.

For the problem of the lock mechanism, in Patent Document 1, the state before and after the record update is managed by a switch. Thereby, the state before and after the update of the record can be changed with a small number of processes. Further, according to Non-Patent Document 1, it is possible to make a reference lock unnecessary among two types of locks, that is, a reference lock and an update lock, by managing the visibility of records in time. In Patent Document 2, the processing speed is increased by providing a lock-free database management method.

For the problem of the log mechanism, according to Patent Document 3, it is possible to refer to an uncommitted result by defining an uncommitted state before committing. Thus, Tx2 can refer to the commit result of Tx1 without waiting for the completion of the log process of Tx1, and the log process can be concealed.

Patent No. 4027055 US2011 / 0252000 US8,407,195

The techniques disclosed in

Patent Documents

1 and 2 and Non-Patent Document 1 can be expected to improve the problem of reference lock and update lock. However, the addition of a new state as in Patent Document 3 complicates the problem of reference lock and update lock. For this reason, a new lock mechanism for efficiently managing a plurality of states is required.

Not limited to the above, it is desirable to visualize the updated record at high speed (in a referable state). Generally, a lock is acquired for each record, and when a plurality of records are updated, each updated record is visualized after releasing all locks of the plurality of records.

The above-described problems are not limited to DBMS, but may be present in other computer systems that update data areas in transaction processing.

An object of the present invention is to visualize the updated data area at high speed.

The computer system receives a transaction processing instruction for updating one or more data areas of a plurality of data areas, and uses a data structure including a Tx state, which is a transaction processing state according to the transaction processing instruction, as the transaction processing instruction. Generate based on. The computer system associates each of one or more data areas updated by transaction processing with a data structure. The computer system executes transaction processing based on the data structure.

・ The updated data area can be visualized at high speed.

It is a functional block diagram of the whole system outline concerning one embodiment. It is a flowchart of a Tx whole process. It is a flowchart of Tx end. It is a flowchart of a commit determination. It is a block diagram of TxID, a record, and an index. It is a block diagram of a priority. It is a figure which shows the relationship between TxID, a record, and an index. It is a block diagram of a priority determination system management table. It is a block diagram of a locking system management table. It is an outline explanatory view of a lock system. It is a block diagram of an example of a Tx instruction. It is a block diagram of an example of a system selection management table. It is a block diagram of DBMS.

Hereinafter, a computer system that executes a DBMS to which an embodiment of the present invention is applied will be described with reference to the drawings.

In the following description, information may be described using the expression “xxx table”, but the information may be expressed in any data structure. That is, “xxx table” can be referred to as “xxx information” to indicate that the information does not depend on the data structure.

Further, in the following description, there is a case where the processing is described by using a functional unit configured by executing the program as a subject, but the program is executed by a processor (for example, a CPU (Central Processing Unit)), The subject of the process may be a processor in order to perform the determined process appropriately using a storage resource (for example, memory) and / or a communication interface. The processing described with the functional unit as the subject may be processing performed by an apparatus including a processor. The processor may include a hardware circuit that performs part or all of the processing. The computer program may be installed on the device from a program source. The program source may be, for example, a program distribution server or a storage medium that can be read by a computer.

Further, in the following description, when the same element is described separately, the entire reference code is used (for example, TxID200A), but when the same element is described without being distinguished, only the common code among the reference codes is used. (For example, TxID200).

FIG. 1 is a functional block diagram of the overall outline of the system according to the embodiment.

The user 1000 sends an instruction to the computer system 1100 by accessing the computer system (for example, a large-scale computer system realized by a plurality of nodes 1200) 1100 from the computer system (for example, personal computer) 1110 through the network 1400. it can. The user 1000 can also directly operate the computer system 1100.

The computer system 1100 has one or more nodes 1200. The node 1200 is an example of a computer. The nodes 1200 may be connected by a system bus 1101. The node 1200 includes a storage device 1202, an input / output device 1203, and a processor 1201 connected to them. The storage device 1202 includes a main storage device (for example, a DRAM (Dynamic Random Access Memory) and at least a main storage device) .The input / output device 1203 may be a communication interface device or a man-machine interface device. (For example, an input device such as a keyboard and a pointing device and an output device such as a display device.) Further, an external storage device 1300 capable of communicating via the network 1400 may be provided outside the computer system 1100.

The DBMS may be executed by one or more nodes 1200 (one or more processors 1201). For example, as illustrated in FIG. 13, the DBMS 1301 includes a query reception unit 1311 that receives a query, a query plan generation unit 1312 that generates a query plan that is an execution plan of the query, and a query execution that executes a query according to the query plan. Part 1313. Tx executed in the Tx process corresponds to Tx of one query according to the query plan. That is, in this embodiment, one TxID (transaction ID) is allocated by the query execution unit 1313 for one query (one query plan). At least two of these

functional units

1311, 1312, and 1313 may be integrated, or at least one of these

functional units

1311, 1312, and 1313 may be subdivided. These

functional units

1311, 1312, and 1313 are realized by executing the DBMS 1301 by one or more processors 1201.

A database referenced or updated by the DBMS 1301 exists in one or more storage devices 1202. In the present embodiment, a so-called in-memory database is realized. The storage device 1202 can acquire data faster than the external storage device 1300.
Therefore, the storage device 1202 can acquire the lock on the acquired data at a higher speed than the external storage device 1300. Note that all or part of the database may exist in the external storage device 1300.

This embodiment is based on the following way of thinking, for example. That is, when changing one Tx state atomically, no lock is required. The record state and the processing phase for each Tx can be synchronized (that is, the record can be managed for each Tx). Based on these, the state of Tx and the record are associated with each other, and the state of the plurality of records can be immediately changed as Tx intended by manipulating the state of Tx atomically. That is, the speed of visualization of the updated record can be realized. Hereinafter, this embodiment will be described in more detail.

FIG. 2 is a flowchart of the Tx process.

The DBMS 1301 performs Tx overall processing based on a Tx processing command (query) from the user 1000. The DBMS 1301 declares Tx start (S100), performs Tx main processing (for example, a series of processes of reference, update, insertion, and aggregation) based on the Tx processing instruction (S120), and declares Tx end (S190). . In the Tx main process of S120, in the case of an update or deletion process, a lock “reservation” may be acquired by the query execution unit 1313 of the DBMS 1301. It is not always necessary to acquire a lock reservation. Whether or not a lock reservation is acquired is determined by a designated lock method (described later with reference to FIG. 9).

FIG. 3 is a flowchart of the end of Tx (S190 in FIG. 2).

The Tx end (S190) is a start (S150), a commit determination (S160) for confirming whether the Tx process maintains correct consistency, a persistence process (S170) for perpetuating the result of the Tx process, And end (S180). The “permanent processing” is to transmit data before and after data update as a log to at least one of the storage device 1202, the external storage device 1300, and the different computer system 1110. As a result, data is not lost even when a failure occurs in the current DBMS 1301.

FIG. 4 is a flowchart of control related to commit determination (S160 in FIG. 3).

The commit determination (S160) starts (S161), the update lock determination for locking the record before the update (S162), and the consistency determination for checking whether the consistency of the Tx main process (S120) is maintained ( S163) and end (S164). In the update lock determination (S162), the query execution unit 1313 of the DBMS 1301 determines the lock (the lock is acquired). This is an indispensable process, unlike acquisition of a “reservation” of a lock.

The DBMS 1301 performs the Tx process by combining these processes S100, S120, and S190 so as to satisfy the ACID characteristic specified by the user 1000. The Tx process is performed by the query execution unit 1313 of the DBMS 1301. One Tx process is performed for one Tx processing instruction (query). It is also possible to execute a plurality of Tx processes in parallel. One Tx process is performed by, for example, one processor core (or one thread assigned to one processor core).

FIG. 5 is a configuration diagram of TxID, record, and index.

At least one of the storage device 1202 and the external storage device 1300 stores the TxID 200, the record 250, and the index 450. These are managed by the DBMS 1301. Each of the record 250 and the index 450 is an example of a data area.

TxID 200 is a data structure. The TxID 200 is a Tx processing command from the user 1000, Tx information 201 for storing the progress of the Tx processing, a Tx state 202 indicating the state of the Tx processing, and a priority indicating an index group for determining the priority of the Tx processing A degree index group 203 is included. The TxID 200 is generated, for example, when a query plan is generated. Examples of the Tx state 202 include abort, lock, and commit. The value of the Tx state 202 is a value corresponding to the current process among the reference numerals 661 to 666 shown in FIG.

The record 250 includes a header 251 having record management information, record data 252, and TxID information 253 and 254 for uniquely identifying the TxID 200. The generated TxID information 253 is TxID 200 of Tx that generated the record 250 or reference information (for example, a pointer) to the TxID 200. The update TxID information 254 is TxID 200 of Tx for updating this record 250 or reference information to the TxID 200. When the Tx for updating the record 250 changes to another Tx, the updated TxID information 254 is updated to the TxID information of the other Tx.

The index 450 includes a header 451 having index management information, a key 452 which is a key value of the index, and TxID information 453 and 454 for uniquely specifying the TxID 200, respectively. The generated TxID information 453 is TxID 200 of Tx that generated the index 450 or reference information (for example, a pointer) to the TxID 200. The update TxID 454 is TxID 200 of Tx for updating the index 450 or reference information to the TxID 200.

The Tx state 202 of the TxID 200 serves as two or more states of the TxID 200 state, the record 250 state, and the index 450 state. If the Tx state 202 of the TxID 200 indicated by the generated TxID information 253 (453) is a state before “commit”, the record (index) having the generated TxID information 253 (453) is the target Tx (this generated TxID information). Reference from Tx other than Tx) corresponding to TxID200 indicated by H.253 (453) is impossible. However, depending on the consistency guarantee level (separation level) and speculative execution of Tx, even if the Tx state 202 of TxID 200 indicated by the generated TxID information 253 (453) is in a state other than “commit”, the Tx other than the target Tx Can also be referenced.

Each of the TxID information 253, 254, 453, and 454 has information such as an address where the corresponding TxID 200 is arranged (for example, an address in the storage device 1202), or a unique ID of the TxID 200, so that the TxID information 253 is obtained. TxID 200 can be specified based on at least one of H.254, 453, and 454.

FIG. 6 is a configuration diagram of the priority index group 203.

The priority index group 203 is a set of a plurality of indices (items), and each of the plurality of indices is used to determine the priority of Tx. The priority index group 203 is used when an acquisition conflict (lock acquisition conflict) of the record 250 occurs in a plurality of Tx processes. The priority index group 203 includes, for example, a priority (310) indicating the high priority of the Tx process, a retry count (311) that is the number of times that the Tx process has failed due to a record acquisition conflict with another Tx process, Tx The core number (312) of the processor core that executed the process, the thread number (313) of the thread that executed the Tx process, the time when the Tx process was started, resumed or designated, or the time (314) indicating the ID unique to the Tx process ), Number of connected records (315) which is the number of records related to TxID, number of reference records (316) which is the number of records referred to in Tx processing, and number of inserted records which is the number of records inserted in Tx processing (317) and the number of updated records (318) that is the number of records updated in the Tx process. At least one of the plurality of indexes included in the priority index group 203 may be omitted or may be another index.

FIG. 7 is a diagram showing the relationship between TxID, records, and indexes.

The record 250 and the index 450 are associated with the TxID 200. Specifically, for example, the TxID 200A specified from the TxID information 253 or 254 in each of the

records

250A and 250C is associated with the

records

250A and 250C, and the TxID 200A specified from the TxID information 453 or 454 in the index 450 is specified. , Associated with the index 450. Further, for example, the TxID 200B specified from the TxID information 253 or 254 in the record 250B is associated with the record 250B. Generally, since a lock is acquired for each of a plurality of records (and indexes) updated by Tx processing, the number of lock acquisitions (number of operations) depends on the number of records (and indexes). By updating the Tx state 202 in one TxID 200, a lock of a plurality (or one) record (index) associated with the one TxID 200 is acquired. Thereby, the required number of calculations can be reduced.

FIG. 8 is a configuration diagram of the priority determination method management table.

The priority determination method management table 801 is stored in one or more storage devices 1202 (see FIG. 1). The priority determination method management table 801 has a column 500 and

columns

501A, 501B,. The indicators (items) 510 to 518 in the column 500 have the same definitions (meaning) as the indicators (items) 310 to 318 in the column 300 shown in FIG.

Each of the

columns

501A, 501B,... Indicates a priority determination method (hereinafter also referred to as “P method”). In each of the

columns

501A, 501B,..., The number corresponding to the item indicates the weight of the item in the priority determination. The smaller the number, the higher the weight. Therefore, for example, “P method 1” in the column 501A is a method for performing priority determination in the order of priority, time, and core number. When the plurality of priorities corresponding to the plurality of TxIDs (Tx processing) are all different, in other words, when the plurality of priorities corresponding to the plurality of TxIDs do not have the same priority, the priority determination ends. Specifically, for example, when “P method 1” is adopted, first, the query execution unit 1313 compares priorities 301 of a plurality of TxIDs 200. If all of the priorities 301 are different, the priority determination ends. On the other hand, when the priorities 301 of at least two TxIDs 200 are the same, the query execution unit 1313 compares the times 314 of the at least two TxIDs 200 next. In this way, according to the definition of the adopted P method, the indicators are compared until the same priority does not exist for a plurality of priorities corresponding to a plurality of TxIDs.

FIG. 9 is a configuration diagram of the lock method management table.

The lock method management table 901 is stored in one or more storage devices 1202 (see FIG. 1). The lock method management table 901 has a column 600 and

columns

601A, 601B,.

Column 600 shows the states of S100 to S190. That is, the

rows

611A, 611B,... Correspond to the states of S100 to S190, respectively.

Each of the

columns

601A, 601B,... Indicates a lock method (hereinafter also referred to as “L method”).

For example, according to “L method 1”, confirmation and connection are performed during the process of the Tx main process (S120). “Confirmation” refers to determining the priority of whether the connection between the TxID 200 and the record 250 is possible. For example, when the first TxID 200A and the second TxID 200B are connected to the record 250A (see, for example, FIG. 7), the query execution unit 1313 performs the priority determination. When the priority of TxID 200A is higher (when TxID 200A wins the priority determination), query execution unit 1313 records information for specifying TxID 200A in TxID information 253 of record 250A. The record of such information is “connection”. That is, the series of processes corresponds to “confirmation” and “connection”.

Further, according to the “L method 1”, the state transition is performed during the process of Tx termination (S190). “State transition” is to update the Tx state 202 of the TxID 200. For example, the Tx state 202 is updated from the Tx main process (S120) to the Tx end (S190). As a result, the state of the plurality of

records

250A and 250C and the index 450 connected to the TxID 200A can be changed simultaneously in one process.

There are a plurality of combinations of these “confirmation”, “connection”, and “state transition”, and therefore, a plurality of L methods can be generated. Since a desired L method can be specified from a plurality of L methods, lock acquisition according to the lock method according to the attribute of the Tx process (for example, the time length required from the start to the end), the load of the computer system 1100, and the like can be expected. In the “L method 2”, since the priority confirmation determination is not performed, “confirmation” (that is, priority determination) is unnecessary. In this embodiment, the L method 2 is selected for Tx processing that is to be preferentially executed. This is because executing the state transition before the start of the Tx main process 611B can prevent the target Tx process from being aborted from another Tx process.

FIG. 10 is a schematic explanatory diagram of the locking method.

Here, the target of priority determination performed between a plurality of Tx processes is clarified. Here, the Tx processing of TxN (N = 1, 2, 3) is referred to as “Tx processing N”. The Tx processes 1 to 3 proceed as time (t) elapses. Here, states 661 to 666 are shown as examples of states.

The priority determination may be performed only for a state in which the Tx processes are the same or a Tx process having a lower comparison target. Specifically, for example, when the designated (selected) L method indicates a procedure including “confirmation”, the priority determination is performed based on whether the Tx process has the same progress state or the comparison target. It is performed only for the lower Tx process, and the comparison target may not be performed for the upper Tx process. As a result, it is possible to avoid that the upper Tx process is aborted by the lower Tx process and cannot be committed. “Lower” means that the current progress state is reached by the end of Tx, for example, Tx1 and Tx3 viewed from Tx2. Further, “upper” means that the state is in a new progress state due to the end of Tx, for example, Tx2 viewed from Tx3. For example, if the comparison target is the upper Tx process 3, even if the priority of the lower Tx process 1 is high, the lower Tx process 1 is lost in the priority determination.

FIG. 11 is a configuration diagram of an example of a Tx instruction.

The Tx processing command 700 is a command issued by the computer system 1110 operated by the user 1000 to the DBMS 1301 executed by the computer system 1100. The Tx processing instruction 700 includes an instruction body 701. The instruction body 701 shows details of Tx processing.

Further, the Tx processing instruction 700 may include at least one of the P method 702 and the L method 703. The P method 702 indicates a priority determination method designated by the user 1000 (or an application program executed by the computer system 1110), and the L method 703 is determined by the user 1000 (or an application program executed by the computer system 1110). Indicates the specified lock method. That is, the user 1000 (or an application program executed on the computer system 1110) can designate at least one of the P method and the L method for the DBMS 1301 for each Tx. The designation is performed by the Tx processing instruction 700. Therefore, it is not necessary to perform an operation different from the issuance of the Tx processing instruction 700.

The DBMS 1301 executes the processing described with reference to FIGS. 2 to 4 based on the Tx processing instruction 700. In this process, the association between the TxID and the record is controlled based on the relationship between the Tx processes (same state, lower level or higher level) and the designated P mode and L mode.

FIG. 12 is a configuration diagram of an example of the method selection management table.

The method selection management table 1201 is stored in one or more storage devices 1202 (see FIG. 1). In the method selection management table 1201, the designated P method and L method are registered for each of the application ranges 811A, 811B,.

Scope of application includes “user”, “database” and “Tx”. That is, a set of P method and L method can be specified in units of each user, each database, and each Tx.

For example, when a combination of the P method and the L method is specified for each user, the Tx processing according to the Tx processing instruction from the user always specifies the user regardless of the database to be processed and the Tx to be executed. Tx processing is performed on the basis of the completed P method and L method. In this case, different P methods and different L methods can coexist for one database. In the Tx process, the index for comparison is specified by referring to the priority determination method management table 801 using the designated P method as a key, and the lock method management table using the designated L method as a key. By referring to 901, the lock acquisition procedure is specified.

Also, for example, when a combination of P method and L method is specified for each database, Tx processing for that database is always specified for that database regardless of the user who issued the Tx processing command or Tx to be executed. Tx processing is performed on the basis of the completed P method and L method.

Also, for example, when a set of P method and L method is specified in units of Tx, the Tx processing is performed based on the specified P method and L method. Even in this case, different P methods and different L methods can coexist for one database.

The specification of the P method and the L method in Tx units is dynamically performed (for example, specified in the Tx processing instruction), and the specification of the P method and the L method in any unit of the user and the database is static. (For example, registered in the method selection management table 1201 in advance). If a dynamic designation is made although a static designation is made, the dynamic designation may always be adopted by the query execution unit 1313, or the dynamic designation and the static designation may be adopted. The query execution unit 1313 may determine which of the dynamic designation and the static designation is adopted based on a policy for determining which one is to be adopted.

Also, the P method and the L method may be created and specified based on a predetermined standard.

For example, suppose that there is a Tx process with many references, specifically, a Tx process performed in accordance with a Tx process instruction 700 that meets a condition that there are many references (a condition defined in advance). When Tx processing with many references is detected by the DBMS 1301 (for example, the query execution unit 1313), the following processing may be performed by the DBMS 1301 (for example, the query execution unit 1313). That is, a setting may be made so that the number of reference records per unit time is maximized. Specifically, for example, as a P method, a P method in which 1: time, 2: reference record number, 3: core number, and 4: thread number are set in descending order of weight may be specified. The combination of the core number and the thread number is a set for making the priority unique, and can be substituted if there is another index for making the priority unique. Further, L method 1 (see FIG. 9) may be designated as the L method. This is because priority determination can be performed (for example, when “L method 2” is selected, another Tx process is forcibly aborted (so-called first come first served), and priority determination is performed. Not) According to the index “number of reference records”, the TxID having the larger number of reference records 316 (see FIG. 6) wins the priority determination.

Also, for example, Tx processing with many updates, specifically, Tx processing (typically committing many updates) performed in accordance with a Tx processing instruction 700 that conforms to a condition with many updates (predefined conditions). Suppose that there is a Tx process). When Tx processing with many updates is detected by the DBMS 1301 (for example, the query execution unit 1313), the following processing may be performed by the DBMS 1301 (for example, the query execution unit 1313). That is, a setting that maximizes the number of updates and inserted records per unit time may be performed. For example, as the P method, a P method in which 1: time, 2: update record number, 3: insert record number, 4: core number, and 5: thread number may be specified in descending order of weight. The combination of the core number and the thread number is a set for making the priority unique, and can be substituted if there is another index for making the priority unique. Further, L method 1 (see FIG. 9) may be designated as the L method. This is because priority determination can be performed. According to the index “number of reference records”, the TxID having the larger number of reference records wins the priority determination. According to the index “number of update records”, the TxID with the larger number of update records 318 (see FIG. 6) wins the priority determination. Further, according to the index “number of inserted records”, the TxID having the larger number of inserted records 317 (see FIG. 6) wins in the priority determination. In Tx processing with many insertions, for example, as a P method, 1: time, 2: number of inserted records, 3: number of updated records, 4: core number, and 5: thread number are set in descending order of weight. The P method may be specified.

As described above, the association between Tx and record is changed based on the result of the priority determination. Thereby, high-speed lock detection can be realized. That is, in general, the lock determination is performed in S162, but in this embodiment, it is determined at the time of S120 whether or not there is a possibility that a lock can be acquired by the lock reservation that can be performed in the Tx main process of S120. it can. In addition, since there is no overlap among the plurality of priorities corresponding to the plurality of TxIDs (since the priorities are uniquely set), deadlock can be avoided.

The outline of the processing performed in this embodiment is as follows, for example (in the following description, the TxID 200 of the target Tx is described as “TxID200F”).

That is, the query receiving unit 1311 receives the Tx processing instruction 700 (query). Thereby, Tx processing is started (S100). The query plan generation unit 1312 generates a query plan based on the Tx processing instruction 700, and at that time, generates a TxID (structure) 200 (hereinafter, 200F). The query execution unit 1313 uses the TxID 200F based on the query plan, and executes the processes of FIGS. 2 to 4 according to the TxID 200F.

At that time, if the query execution unit 1313 is connected to another TxID in one or more records (specifically, updated TxID information 254 of each of the one or more records) among a plurality of records to be connected to the TxID 200F, For example, the following processing is performed for each of the one or more records. In the following description, one record (hereinafter, target record) is taken as an example.
The query execution unit 1313 determines whether the state of TxID 200 (hereinafter, 200T) of the connection destination (update TxID information 254) of the target record is the same as TxID 200F or whether TxID 200T is lower than TxID 200F.
If the result of the determination is affirmative, the query execution unit 1313 determines whether the TxID 200F and TxID 200T are in accordance with the P method specified for the TxID 200F (for example, the P method specified for each user, each database, or each Tx). Priority determination of which priority is higher is performed.
When TxID 200F wins in the priority determination, the query execution unit 1313 acquires a lock according to the L method specified for TxID 200F (for example, the L method specified for each user, each database, or each Tx) ( Get lock reservation). Specifically, for example, the query execution unit 1313 changes the Tx state 202 of the TxID 200T lost in the priority determination to “Abort”, and uses the updated TxID information 254 in the target record as information for specifying the TxID 200F. Update.

The process from the start of the Tx process to this step is the Tx main process (S120).

When all the Tx main processes (S120) based on the query plan are finished, the query execution unit 1313 proceeds to the update lock determination (S162 in FIG. 4). Here, the Tx state 202 of the TxID 200F is changed to “update lock determination 663”. If the Tx state 202 of the TxID 200F is “Abort”, the query execution unit 1313 performs an abort process.

Next, the query execution unit 1313 performs consistency determination (S163 in FIG. 4). If the result of the consistency determination (S163) is positive, the query execution unit 1313 changes the Tx state 202 of the TxID 200F to “consistency determination 664”. If the Tx state 202 of the TxID 200F is “Abort”, the query execution unit 1313 performs an abort process.

The commit determination including S162 and S163 (S160 in FIG. 3) ends, and then the query execution unit 1313 performs a persistence process (S170 in FIG. 3). After confirming that the log has been made permanent, the query execution unit 1313 changes the Tx state 202 of the TxID 200F to “permanent processing 665”.

Finally, the query execution unit 1313 performs Tx termination (S190 in FIG. 2) of Tx corresponding to TxID 200F. Here, the query execution unit 1313 changes the Tx state 202 of the TxID 200F to “Tx end 666 (commit)”.

As mentioned above, although one embodiment was described, it cannot be overemphasized that this invention can be variously changed in the range which is not limited to this embodiment and does not deviate from the summary.

1100: Computer system

Claims

A reception unit that receives a transaction processing instruction for updating one or more data areas of a plurality of data areas of the database;
A data structure including a Tx state that is a state of transaction processing according to the transaction processing instruction is generated based on the transaction processing instruction, and each of the one or more data areas updated by the transaction processing is stored in the data structure. And a database management system having an execution unit that executes transaction processing based on the data structure.
When the first data area among the one or more data areas updated by the first transaction process is associated with the generated second data structure, the execution unit executes the second data Based on the second Tx state which is the Tx state included in the structure, the first data area is replaced with the second data structure in the first data structure corresponding to the first transaction processing. Control whether to associate,
The database management system according to claim 1.
The execution unit executes associating each of the one or more data areas with the data structure according to a designated association method among a plurality of association methods,
Each of the plurality of association methods is a method that defines a procedure for associating each of the one or more data areas with the data structure.
The database management system according to claim 1.
A procedure in which a first data area among one or more data areas updated by a first transaction process is associated with a generated second data structure, and indicated by the designated association method Is included in determining the priority, the execution unit,
Based on the second data structure and the first data structure corresponding to the first transaction process, the first transaction process and the second transaction process corresponding to the second data structure. Priority determination to determine the priority of
Associating the first data area with the first data structure instead of the second data structure when the priority of the first transaction is higher;
The database management system according to claim 3.
When the progress state of the second transaction process is the same as the progress state of the first transaction process, or when the progress state is older than the end state than the first transaction process, the execution unit sets the priority Judgment degree,
When the progress state of the second transaction process is a new progress state from the end state than the first transaction process, the execution unit does not perform the priority determination;
The database management system according to claim 4.
The data structure includes a priority index group that is a plurality of indexes that can be used for priority determination,
The execution unit executes the priority determination according to a specified priority determination method among a plurality of priority determination methods.
The database management system according to claim 4.
The designation of the association method is performed in at least one unit of the user unit and the transaction processing unit.
The database management system according to claim 3.
The transaction processing instruction includes the specified association method;
The database management system according to claim 7.
When the first data area among the one or more data areas updated by the first transaction process is associated with the generated second data structure, the execution unit includes:
Based on the second data structure and the first data structure corresponding to the first transaction process, the first transaction process and the second transaction process corresponding to the second data structure. Priority determination to determine the priority of
Associating the first data area with the first data structure instead of the second data structure when the priority of the first transaction is higher;
The database management system according to claim 1.
The data area associated with the data structure is associated with at least one of a data structure of a transaction process that generated the data area and a data structure of a transaction process that recently updated the data area,
The database management system according to claim 1.
The data structure includes a priority index group that is a plurality of indexes that can be used for priority determination,
The execution unit executes the priority determination according to a designated priority determination method among a plurality of priority determination methods,
In each of the plurality of priority determination methods, a weight of an index to be used is defined,
The execution unit performs the priority determination in the order of index weights defined by the designated priority determination method, and the index value and the second data structure included in the first data structure. Compare the values of the indicators contained in the body,
The database management system according to claim 9.
The specification of the priority determination method is performed in at least one of the user unit and the transaction processing unit.
The database management system according to claim 11.
The transaction processing instruction includes the designated priority determination method;
The database management system according to claim 12.
An input / output device that receives a transaction processing instruction for updating one or more data areas of the plurality of data areas;
A data structure including a Tx state that is a state of transaction processing according to the transaction processing instruction is generated based on the transaction processing instruction, and each of the one or more data areas updated by the transaction processing is stored in the data structure. And a processor for executing transaction processing based on the data structure.
Accepts a transaction processing instruction for updating one or more data areas of a plurality of data areas of a database;
Generating a data structure including a Tx state which is a state of transaction processing according to the transaction processing instruction based on the transaction processing instruction;
Associating each of the one or more data areas updated by the transaction process with the data structure;
Performing transaction processing based on the data structure;
Database management method.